<?xml version="1.0" encoding="UTF-8"?>
<?xml-stylesheet href="/stylesheet.css" type="text/css"?>
<?xml-stylesheet href="data:text/css,%3Aroot%7B--art%3Aurl%28%22https%3A%2F%2Fimg.transistorcdn.com%2FEDBdfQZ84XNPurTV_DnBqeugA25dI9ba4FQNk3_1lSQ%2Frs%3Afill%3A0%3A0%3A1%2Fw%3A400%2Fh%3A400%2Fq%3A60%2Fmb%3A500000%2FaHR0cHM6Ly9pbWct%2FdXBsb2FkLXByb2R1%2FY3Rpb24udHJhbnNp%2Fc3Rvci5mbS9kMDY0%2FZWFlNjUwZGJiZjhl%2FZjRjNGJkZWU5Zjhi%2FZjNjOC5qcGc.webp%22%29%3B--art-scale%3A1%7D" type="text/css"?>
<rss version="2.0" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:sy="http://purl.org/rss/1.0/modules/syndication/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:itunes="http://www.itunes.com/dtds/podcast-1.0.dtd" xmlns:podcast="https://podcastindex.org/namespace/1.0">
  <channel>
    <atom:link rel="self" type="application/rss+xml" href="https://feeds.transistor.fm/ai-news-daily" title="MP3 Audio"/>
    <atom:link rel="hub" href="https://pubsubhubbub.appspot.com/"/>
    <podcast:podping usesPodping="true"/>
    <title>AI News Daily</title>
    <link>https://researchnode.dev</link>
    <generator>Transistor (https://transistor.fm)</generator>
    <itunes:new-feed-url>https://feeds.transistor.fm/ai-news-daily</itunes:new-feed-url>
    <description>A concise English-language briefing on artificial intelligence: new models and research, AI security, documented threat-actor activity, and the business of AI. Approximately five minutes daily, with an approximately twenty-minute Sunday review of the week's most important substantive developments. Produced by ResearchNode.dev using AI-generated research and narration, with source checks before release and links to original reporting and research in the episode notes.</description>
    <copyright>© 2026 ResearchNode.dev</copyright>
    <podcast:guid>eaa4dba5-3db8-5259-9ef4-927af803f67d</podcast:guid>
    <podcast:locked>yes</podcast:locked>
    <language>en</language>
    <pubDate>Fri, 20 Nov 2026 07:06:40 +0100</pubDate>
    <lastBuildDate>Fri, 20 Nov 2026 07:07:19 +0100</lastBuildDate>
    <image>
      <url>https://img.transistorcdn.com/7d_tinem0ujE0swVtuUZZ8ZT5cVkjmIrmtYQva1pG6U/rs:fill:0:0:1/w:1400/h:1400/q:60/mb:500000/aHR0cHM6Ly9pbWct/dXBsb2FkLXByb2R1/Y3Rpb24udHJhbnNp/c3Rvci5mbS9kMDY0/ZWFlNjUwZGJiZjhl/ZjRjNGJkZWU5Zjhi/ZjNjOC5qcGc.jpg</url>
      <title>AI News Daily</title>
      <link>https://researchnode.dev</link>
    </image>
    <itunes:category text="News">
      <itunes:category text="Tech News"/>
    </itunes:category>
    <itunes:category text="Technology"/>
    <itunes:type>episodic</itunes:type>
    <itunes:author>ResearchNode.dev</itunes:author>
    <itunes:image href="https://img.transistorcdn.com/7d_tinem0ujE0swVtuUZZ8ZT5cVkjmIrmtYQva1pG6U/rs:fill:0:0:1/w:1400/h:1400/q:60/mb:500000/aHR0cHM6Ly9pbWct/dXBsb2FkLXByb2R1/Y3Rpb24udHJhbnNp/c3Rvci5mbS9kMDY0/ZWFlNjUwZGJiZjhl/ZjRjNGJkZWU5Zjhi/ZjNjOC5qcGc.jpg"/>
    <itunes:summary>A concise English-language briefing on artificial intelligence: new models and research, AI security, documented threat-actor activity, and the business of AI. Approximately five minutes daily, with an approximately twenty-minute Sunday review of the week's most important substantive developments. Produced by ResearchNode.dev using AI-generated research and narration, with source checks before release and links to original reporting and research in the episode notes.</itunes:summary>
    <itunes:subtitle>A concise English-language briefing on artificial intelligence: new models and research, AI security, documented threat-actor activity, and the business of AI.</itunes:subtitle>
    <itunes:keywords></itunes:keywords>
    <itunes:owner>
      <itunes:name>Mikael Eriksson</itunes:name>
    </itunes:owner>
    <itunes:complete>No</itunes:complete>
    <itunes:explicit>No</itunes:explicit>
    <item>
      <title>Gemini Neon, science agents and an AI acquisition</title>
      <itunes:title>Gemini Neon, science agents and an AI acquisition</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">c9f0c5a4-a8e8-4583-acd8-ef5a707de886</guid>
      <link>https://share.transistor.fm/s/fd9803ec</link>
      <description>
        <![CDATA[<p>New Google and RadixArk models, NVIDIA’s scientific AI resources, Microsoft’s security agents, an EU attack report and Palo Alto Networks’ proposed acquisition.</p>
<p><b>Stories and sources</b></p>
<p><b>Google begins Gemini 4 Neon rollout</b></p>
<p>Hosted access expands, with separate restricted cyber access planned.</p>
<ul>
<li><a href="https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-neon/">Introducing Gemini 4 Neon</a> — Google</li>
</ul>
<p><b>RadixArk releases DaisyRose-27B</b></p>
<p>Downloadable weights accompany training details and a benchmark-version qualification.</p>
<ul>
<li><a href="https://huggingface.co/blog/RadixArk/daisyrose-27b-pushing-the-frontier-of-post-trainin">DaisyRose-27B: Pushing the Frontier of Post-Training Efficiency</a> — RadixArk on Hugging Face</li>
<li><a href="https://huggingface.co/RadixArk/DaisyRose-27B">RadixArk/DaisyRose-27B</a> — RadixArk on Hugging Face</li>
</ul>
<p><b>NVIDIA introduces Nemotron for Science</b></p>
<p>Chemistry data and adaptation recipes accompany a reported simulation-repair experiment.</p>
<ul>
<li><a href="https://nvidianews.nvidia.com/news/nvidia-introduces-nemotron-for-science-to-accelerate-discoveries-with-agents">NVIDIA Introduces Nemotron for Science to Accelerate Discoveries With Agents</a> — NVIDIA</li>
<li><a href="https://blogs.nvidia.com/blog/ian-buck-sc26/">Ian Buck Kicks Off NVIDIA’s Week at SC26 as Science Enters the Age of Agents</a> — NVIDIA</li>
</ul>
<p><b>Microsoft expands Perception security agents</b></p>
<p>General availability adds investigation inputs and reviewed detection workflows.</p>
<ul>
<li><a href="https://techcommunity.microsoft.com/blog/perception-blog/whats-new-with-microsoft-perception-at-ignite-2026/4571834">What's new with Microsoft Perception at Ignite 2026</a> — Microsoft</li>
</ul>
<p><b>EU confirms AI-assisted attack</b></p>
<p>AP reports official confirmation.</p>
<ul>
<li><a href="https://apnews.com/article/cyberattack-european-union-artificial-intelligence-ai-79705095e8141cd650196dbff1cf3c70">A cyberattack using artificial intelligence targeted the EU in October, official says</a> — Associated Press</li>
</ul>
<p><b>Palo Alto Networks agrees to acquire Applied Compute</b></p>
<p>The $4.5 billion all-stock deal remains conditional.</p>
<ul>
<li><a href="https://investors.paloaltonetworks.com/static-files/c70b2bd5-d86b-4a4f-8380-c7d6c626b191">Form 8-K for Palo Alto Networks Inc. filed November 18, 2026</a> — Palo Alto Networks</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>New Google and RadixArk models, NVIDIA’s scientific AI resources, Microsoft’s security agents, an EU attack report and Palo Alto Networks’ proposed acquisition.</p>
<p><b>Stories and sources</b></p>
<p><b>Google begins Gemini 4 Neon rollout</b></p>
<p>Hosted access expands, with separate restricted cyber access planned.</p>
<ul>
<li><a href="https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-neon/">Introducing Gemini 4 Neon</a> — Google</li>
</ul>
<p><b>RadixArk releases DaisyRose-27B</b></p>
<p>Downloadable weights accompany training details and a benchmark-version qualification.</p>
<ul>
<li><a href="https://huggingface.co/blog/RadixArk/daisyrose-27b-pushing-the-frontier-of-post-trainin">DaisyRose-27B: Pushing the Frontier of Post-Training Efficiency</a> — RadixArk on Hugging Face</li>
<li><a href="https://huggingface.co/RadixArk/DaisyRose-27B">RadixArk/DaisyRose-27B</a> — RadixArk on Hugging Face</li>
</ul>
<p><b>NVIDIA introduces Nemotron for Science</b></p>
<p>Chemistry data and adaptation recipes accompany a reported simulation-repair experiment.</p>
<ul>
<li><a href="https://nvidianews.nvidia.com/news/nvidia-introduces-nemotron-for-science-to-accelerate-discoveries-with-agents">NVIDIA Introduces Nemotron for Science to Accelerate Discoveries With Agents</a> — NVIDIA</li>
<li><a href="https://blogs.nvidia.com/blog/ian-buck-sc26/">Ian Buck Kicks Off NVIDIA’s Week at SC26 as Science Enters the Age of Agents</a> — NVIDIA</li>
</ul>
<p><b>Microsoft expands Perception security agents</b></p>
<p>General availability adds investigation inputs and reviewed detection workflows.</p>
<ul>
<li><a href="https://techcommunity.microsoft.com/blog/perception-blog/whats-new-with-microsoft-perception-at-ignite-2026/4571834">What's new with Microsoft Perception at Ignite 2026</a> — Microsoft</li>
</ul>
<p><b>EU confirms AI-assisted attack</b></p>
<p>AP reports official confirmation.</p>
<ul>
<li><a href="https://apnews.com/article/cyberattack-european-union-artificial-intelligence-ai-79705095e8141cd650196dbff1cf3c70">A cyberattack using artificial intelligence targeted the EU in October, official says</a> — Associated Press</li>
</ul>
<p><b>Palo Alto Networks agrees to acquire Applied Compute</b></p>
<p>The $4.5 billion all-stock deal remains conditional.</p>
<ul>
<li><a href="https://investors.paloaltonetworks.com/static-files/c70b2bd5-d86b-4a4f-8380-c7d6c626b191">Form 8-K for Palo Alto Networks Inc. filed November 18, 2026</a> — Palo Alto Networks</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Fri, 20 Nov 2026 07:06:40 +0100</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/fd9803ec/bde671cf.mp3" length="4795706" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>300</itunes:duration>
      <itunes:summary>New Google and RadixArk models, NVIDIA’s scientific AI resources, Microsoft’s security agents, an EU attack report and Palo Alto Networks’ proposed acquisition.</itunes:summary>
      <itunes:subtitle>New Google and RadixArk models, NVIDIA’s scientific AI resources, Microsoft’s security agents, an EU attack report and Palo Alto Networks’ proposed acquisition.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/fd9803ec/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>Genie 4 access, open speech recognition and Anthropic’s IPO pricing</title>
      <itunes:title>Genie 4 access, open speech recognition and Anthropic’s IPO pricing</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">07e5731b-ea61-4091-998d-41c8feaa896f</guid>
      <link>https://share.transistor.fm/s/3ae2ea1a</link>
      <description>
        <![CDATA[<p>Google expands Project Genie access, Reson8 releases speech-model weights, IBM and NVIDIA announce deployment options, Bitdefender documents malicious agent configurations, and Anthropic prices its offering.</p>
<p><b>Stories and sources</b></p>
<p><b>Google expands Genie 4 access through Project Genie</b></p>
<p>Eligible Ultra subscribers gain access; Pro members can join a waitlist.</p>
<ul>
<li><a href="https://blog.google/innovation-and-ai/models-and-research/google-deepmind/genie-4-now-available-project-genie/">Genie 4 is now available</a> — Google</li>
<li><a href="https://deepmind.google/blog/genie-4-a-new-generation-of-world-simulation/">Genie 4: A new generation of world simulation</a> — Google DeepMind</li>
</ul>
<p><b>Reson8 releases Rumble-2-Flash weights</b></p>
<p>The compact multilingual speech model includes transcription controls and reported benchmark results.</p>
<ul>
<li><a href="https://huggingface.co/blog/rumble-2-flash">Rumble-2-Flash: an open 0.5B speech model for 40+ languages</a> — Reson8 on Hugging Face</li>
<li><a href="https://huggingface.co/Reson8/Rumble-2-Flash">Reson8/Rumble-2-Flash</a> — Reson8 on Hugging Face</li>
</ul>
<p><b>IBM announces time-series models inside Confluent</b></p>
<p>Streaming forecasts and anomaly detection run on CPUs.</p>
<ul>
<li><a href="https://huggingface.co/blog/ibm-research/time-series-powered-streamhouse">Time-Series Models in the Streamhouse: A New Frontier and Use Cases</a> — IBM Research on Hugging Face</li>
</ul>
<p><b>NVIDIA announces RTX Spark desktop availability</b></p>
<p>ASUS, Dell and MSI systems support local AI workflows.</p>
<ul>
<li><a href="https://blogs.nvidia.com/blog/rtx-spark-desktop-local-ai-available-now/">NVIDIA RTX Spark Compact Desktops Out Now</a> — NVIDIA</li>
</ul>
<p><b>Bitdefender reports malicious coding-agent configurations</b></p>
<p>Malicious repositories target assistant execution paths.</p>
<ul>
<li><a href="https://www.bitdefender.com/en-au/blog/labs/how-ai-agents-become-new-targets-in-github-malware-campaign">How AI Agents Become New Targets in GitHub Malware Campaign</a> — Bitdefender Labs</li>
</ul>
<p><b>Anthropic prices its initial public offering</b></p>
<p>Pricing precedes expected trading and conditional closing.</p>
<ul>
<li><a href="https://www.anthropic.com/news/anthropic-prices-initial-public-offering">Anthropic prices initial public offering</a> — Anthropic</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>Google expands Project Genie access, Reson8 releases speech-model weights, IBM and NVIDIA announce deployment options, Bitdefender documents malicious agent configurations, and Anthropic prices its offering.</p>
<p><b>Stories and sources</b></p>
<p><b>Google expands Genie 4 access through Project Genie</b></p>
<p>Eligible Ultra subscribers gain access; Pro members can join a waitlist.</p>
<ul>
<li><a href="https://blog.google/innovation-and-ai/models-and-research/google-deepmind/genie-4-now-available-project-genie/">Genie 4 is now available</a> — Google</li>
<li><a href="https://deepmind.google/blog/genie-4-a-new-generation-of-world-simulation/">Genie 4: A new generation of world simulation</a> — Google DeepMind</li>
</ul>
<p><b>Reson8 releases Rumble-2-Flash weights</b></p>
<p>The compact multilingual speech model includes transcription controls and reported benchmark results.</p>
<ul>
<li><a href="https://huggingface.co/blog/rumble-2-flash">Rumble-2-Flash: an open 0.5B speech model for 40+ languages</a> — Reson8 on Hugging Face</li>
<li><a href="https://huggingface.co/Reson8/Rumble-2-Flash">Reson8/Rumble-2-Flash</a> — Reson8 on Hugging Face</li>
</ul>
<p><b>IBM announces time-series models inside Confluent</b></p>
<p>Streaming forecasts and anomaly detection run on CPUs.</p>
<ul>
<li><a href="https://huggingface.co/blog/ibm-research/time-series-powered-streamhouse">Time-Series Models in the Streamhouse: A New Frontier and Use Cases</a> — IBM Research on Hugging Face</li>
</ul>
<p><b>NVIDIA announces RTX Spark desktop availability</b></p>
<p>ASUS, Dell and MSI systems support local AI workflows.</p>
<ul>
<li><a href="https://blogs.nvidia.com/blog/rtx-spark-desktop-local-ai-available-now/">NVIDIA RTX Spark Compact Desktops Out Now</a> — NVIDIA</li>
</ul>
<p><b>Bitdefender reports malicious coding-agent configurations</b></p>
<p>Malicious repositories target assistant execution paths.</p>
<ul>
<li><a href="https://www.bitdefender.com/en-au/blog/labs/how-ai-agents-become-new-targets-in-github-malware-campaign">How AI Agents Become New Targets in GitHub Malware Campaign</a> — Bitdefender Labs</li>
</ul>
<p><b>Anthropic prices its initial public offering</b></p>
<p>Pricing precedes expected trading and conditional closing.</p>
<ul>
<li><a href="https://www.anthropic.com/news/anthropic-prices-initial-public-offering">Anthropic prices initial public offering</a> — Anthropic</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Thu, 19 Nov 2026 07:07:34 +0100</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/3ae2ea1a/c4fe3519.mp3" length="5681362" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>356</itunes:duration>
      <itunes:summary>Google expands Project Genie access, Reson8 releases speech-model weights, IBM and NVIDIA announce deployment options, Bitdefender documents malicious agent configurations, and Anthropic prices its offering.</itunes:summary>
      <itunes:subtitle>Google expands Project Genie access, Reson8 releases speech-model weights, IBM and NVIDIA announce deployment options, Bitdefender documents malicious agent configurations, and Anthropic prices its offering.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/3ae2ea1a/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>GPT-6.2, new coding models and automated exploitation</title>
      <itunes:title>GPT-6.2, new coding models and automated exploitation</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">9b0b4a30-be9d-4a0e-822c-85f6701c11cd</guid>
      <link>https://share.transistor.fm/s/77a41e45</link>
      <description>
        <![CDATA[<p>New models from OpenAI, Microsoft and Alibaba; Anthropic’s disclosure deadlines; Google’s attacker-infrastructure findings; and ElevenLabs’ Kotoba acquisition announcement.</p>
<p><b>Stories and sources</b></p>
<p><b>OpenAI releases GPT-6.2 Astra and Sol</b></p>
<p>The models reach Work, Codex and the API, with lower Astra token prices and attributed coding results.</p>
<ul>
<li><a href="https://learn.chatgpt.com/docs/whats-new/gpt-62-astra-sol">GPT-6.2</a> — OpenAI</li>
<li><a href="https://openai.com/index/introducing-gpt-6-2-astra-sol/">Introducing GPT-6.2 Astra and Sol</a> — OpenAI</li>
<li><a href="https://help.openai.com/en/articles/9624314-model-release-notes">Model Release Notes</a> — OpenAI</li>
</ul>
<p><b>Microsoft releases MAI Code 2 models</b></p>
<p>Copilot availability accompanies benchmark comparisons using promotional pricing.</p>
<ul>
<li><a href="https://microsoft.ai/news/microsoft-ignite-2026/">Seven new world class models at scalable price-performance</a> — Microsoft AI</li>
<li><a href="https://microsoft.ai/news/introducing-mai-code-2-and-mai-code-2-flash/">MAI Code 2 and Code 2 Flash: Everyday coding at scalable prices</a> — Microsoft AI</li>
</ul>
<p><b>Alibaba releases Qwen4-Max beyond its preview</b></p>
<p>API access accompanies multimodal capabilities and vendor-reported efficiency results.</p>
<ul>
<li><a href="https://www.alibabacloud.com/en/press-room/alibaba-unveils-qwen4-max?_p_lc=1">Alibaba Unveils Qwen4-Max: New AI Teammate for Work, Creativity, and Discovery</a> — Alibaba Cloud</li>
<li><a href="https://docs.qwencloud.com/changelog/models">Model releases</a> — QwenCloud</li>
</ul>
<p><b>Anthropic shortens vulnerability-disclosure deadlines</b></p>
<p>New deadlines begin in January, with exceptions.</p>
<ul>
<li><a href="https://www.anthropic.com/research/coordinated-vulnerability-disclosure-timelines">Updating our coordinated vulnerability disclosure policy timelines</a> — Anthropic</li>
</ul>
<p><b>Google reports automated exploitation infrastructure</b></p>
<p>Investigators found exploit_agent alongside stolen data.</p>
<ul>
<li><a href="https://cloud.google.com/blog/topics/threat-intelligence/agentic-threat-activity-crosses-a-significant-threshold?hl=en">GTIG AI Threat Tracker: Agentic Threat Activity Crosses a Significant Threshold</a> — Google Cloud</li>
</ul>
<p><b>ElevenLabs announces Kotoba acquisition</b></p>
<p>The announcement targets Japanese voice technology integration.</p>
<ul>
<li><a href="https://elevenlabs.io/blog/kotoba">Kotoba is joining ElevenLabs</a> — ElevenLabs</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>New models from OpenAI, Microsoft and Alibaba; Anthropic’s disclosure deadlines; Google’s attacker-infrastructure findings; and ElevenLabs’ Kotoba acquisition announcement.</p>
<p><b>Stories and sources</b></p>
<p><b>OpenAI releases GPT-6.2 Astra and Sol</b></p>
<p>The models reach Work, Codex and the API, with lower Astra token prices and attributed coding results.</p>
<ul>
<li><a href="https://learn.chatgpt.com/docs/whats-new/gpt-62-astra-sol">GPT-6.2</a> — OpenAI</li>
<li><a href="https://openai.com/index/introducing-gpt-6-2-astra-sol/">Introducing GPT-6.2 Astra and Sol</a> — OpenAI</li>
<li><a href="https://help.openai.com/en/articles/9624314-model-release-notes">Model Release Notes</a> — OpenAI</li>
</ul>
<p><b>Microsoft releases MAI Code 2 models</b></p>
<p>Copilot availability accompanies benchmark comparisons using promotional pricing.</p>
<ul>
<li><a href="https://microsoft.ai/news/microsoft-ignite-2026/">Seven new world class models at scalable price-performance</a> — Microsoft AI</li>
<li><a href="https://microsoft.ai/news/introducing-mai-code-2-and-mai-code-2-flash/">MAI Code 2 and Code 2 Flash: Everyday coding at scalable prices</a> — Microsoft AI</li>
</ul>
<p><b>Alibaba releases Qwen4-Max beyond its preview</b></p>
<p>API access accompanies multimodal capabilities and vendor-reported efficiency results.</p>
<ul>
<li><a href="https://www.alibabacloud.com/en/press-room/alibaba-unveils-qwen4-max?_p_lc=1">Alibaba Unveils Qwen4-Max: New AI Teammate for Work, Creativity, and Discovery</a> — Alibaba Cloud</li>
<li><a href="https://docs.qwencloud.com/changelog/models">Model releases</a> — QwenCloud</li>
</ul>
<p><b>Anthropic shortens vulnerability-disclosure deadlines</b></p>
<p>New deadlines begin in January, with exceptions.</p>
<ul>
<li><a href="https://www.anthropic.com/research/coordinated-vulnerability-disclosure-timelines">Updating our coordinated vulnerability disclosure policy timelines</a> — Anthropic</li>
</ul>
<p><b>Google reports automated exploitation infrastructure</b></p>
<p>Investigators found exploit_agent alongside stolen data.</p>
<ul>
<li><a href="https://cloud.google.com/blog/topics/threat-intelligence/agentic-threat-activity-crosses-a-significant-threshold?hl=en">GTIG AI Threat Tracker: Agentic Threat Activity Crosses a Significant Threshold</a> — Google Cloud</li>
</ul>
<p><b>ElevenLabs announces Kotoba acquisition</b></p>
<p>The announcement targets Japanese voice technology integration.</p>
<ul>
<li><a href="https://elevenlabs.io/blog/kotoba">Kotoba is joining ElevenLabs</a> — ElevenLabs</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Wed, 18 Nov 2026 07:06:04 +0100</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/77a41e45/1bf2da71.mp3" length="5458590" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>342</itunes:duration>
      <itunes:summary>New models from OpenAI, Microsoft and Alibaba; Anthropic’s disclosure deadlines; Google’s attacker-infrastructure findings; and ElevenLabs’ Kotoba acquisition announcement.</itunes:summary>
      <itunes:subtitle>New models from OpenAI, Microsoft and Alibaba; Anthropic’s disclosure deadlines; Google’s attacker-infrastructure findings; and ElevenLabs’ Kotoba acquisition announcement.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/77a41e45/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>Agent safeguards, audio models and training benchmarks</title>
      <itunes:title>Agent safeguards, audio models and training benchmarks</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">cf191a29-4b7e-4ea9-a1ca-dee9cfd3dc30</guid>
      <link>https://share.transistor.fm/s/f61e8181</link>
      <description>
        <![CDATA[<p>California sues OpenAI over experimental-agent safeguards; Liquid AI adds audio input; Hugging Face changes inference routing; NVIDIA reports training results; and researchers examine privacy reviews and model identity checks.</p>
<p><b>Stories and sources</b></p>
<p><b>California sues OpenAI over experimental-agent safeguards</b></p>
<p>The state alleges unsafe testing; OpenAI disputes being singled out.</p>
<ul>
<li><a href="https://oag.ca.gov/news/press-releases/it%E2%80%99s-golden-state-not-wild-west-attorney-general-bonta-sues-openai-over">It’s the Golden State, not the Wild West: Attorney General Bonta Sues OpenAI Over Cybersecurity Incidents</a> — California Department of Justice</li>
<li><a href="https://www.kpbs.org/news/science-technology/2026/11/16/california-sues-openai-over-hacking-risks">California sues OpenAI over hacking risks</a> — NPR, published by KPBS</li>
</ul>
<p><b>Liquid AI adds audio to decision models</b></p>
<p>API releases add audio and structured extraction tasks.</p>
<ul>
<li><a href="https://www.liquid.ai/blog/d1-2">d1.2: State-of-the-Art Omnimodal Decision Models</a> — Liquid AI</li>
</ul>
<p><b>Hugging Face switches language-model routing to OpenRouter</b></p>
<p>Routing changes include retries and new billing constraints.</p>
<ul>
<li><a href="https://huggingface.co/blog/openrouter">Hugging Face and OpenRouter join forces to make open-weight models easy to run</a> — Hugging Face</li>
</ul>
<p><b>NVIDIA reports Vera Rubin training benchmark results</b></p>
<p>Reported gains vary by workload and configuration.</p>
<ul>
<li><a href="https://developer.nvidia.com/blog/nvidia-vera-rubin-nvl72-achieves-up-to-2x-blackwell-ultra-performance-in-mlperf-training-v6-1/">NVIDIA Vera Rubin NVL72 Achieves up to 2x Blackwell Ultra Performance in MLPerf Training v6.1</a> — NVIDIA</li>
<li><a href="https://mlcommons.org/benchmarks/training/">MLPerf Training</a> — MLCommons</li>
</ul>
<p><b>PrivacyPR evaluates privacy-focused code review</b></p>
<p>Explicit privacy prompts improve reported detection, with false alarms.</p>
<ul>
<li><a href="https://arxiv.org/abs/2611.18293">PrivacyPR: A Dataset of Real-World Privacy Fixes for Benchmarking Privacy-Aware Code Review</a> — arXiv</li>
<li><a href="https://arxiv.org/html/2611.18293v1">PrivacyPR: A Dataset of Real-World Privacy Fixes for Benchmarking Privacy-Aware Code Review</a> — arXiv</li>
</ul>
<p><b>Research tests history-aware model identity audits</b></p>
<p>Lineup compares responses using matching conversation histories.</p>
<ul>
<li><a href="https://arxiv.org/abs/2611.18104">Contagious Fingerprints: When LLMs Inherit Another Model's Identity</a> — arXiv</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>California sues OpenAI over experimental-agent safeguards; Liquid AI adds audio input; Hugging Face changes inference routing; NVIDIA reports training results; and researchers examine privacy reviews and model identity checks.</p>
<p><b>Stories and sources</b></p>
<p><b>California sues OpenAI over experimental-agent safeguards</b></p>
<p>The state alleges unsafe testing; OpenAI disputes being singled out.</p>
<ul>
<li><a href="https://oag.ca.gov/news/press-releases/it%E2%80%99s-golden-state-not-wild-west-attorney-general-bonta-sues-openai-over">It’s the Golden State, not the Wild West: Attorney General Bonta Sues OpenAI Over Cybersecurity Incidents</a> — California Department of Justice</li>
<li><a href="https://www.kpbs.org/news/science-technology/2026/11/16/california-sues-openai-over-hacking-risks">California sues OpenAI over hacking risks</a> — NPR, published by KPBS</li>
</ul>
<p><b>Liquid AI adds audio to decision models</b></p>
<p>API releases add audio and structured extraction tasks.</p>
<ul>
<li><a href="https://www.liquid.ai/blog/d1-2">d1.2: State-of-the-Art Omnimodal Decision Models</a> — Liquid AI</li>
</ul>
<p><b>Hugging Face switches language-model routing to OpenRouter</b></p>
<p>Routing changes include retries and new billing constraints.</p>
<ul>
<li><a href="https://huggingface.co/blog/openrouter">Hugging Face and OpenRouter join forces to make open-weight models easy to run</a> — Hugging Face</li>
</ul>
<p><b>NVIDIA reports Vera Rubin training benchmark results</b></p>
<p>Reported gains vary by workload and configuration.</p>
<ul>
<li><a href="https://developer.nvidia.com/blog/nvidia-vera-rubin-nvl72-achieves-up-to-2x-blackwell-ultra-performance-in-mlperf-training-v6-1/">NVIDIA Vera Rubin NVL72 Achieves up to 2x Blackwell Ultra Performance in MLPerf Training v6.1</a> — NVIDIA</li>
<li><a href="https://mlcommons.org/benchmarks/training/">MLPerf Training</a> — MLCommons</li>
</ul>
<p><b>PrivacyPR evaluates privacy-focused code review</b></p>
<p>Explicit privacy prompts improve reported detection, with false alarms.</p>
<ul>
<li><a href="https://arxiv.org/abs/2611.18293">PrivacyPR: A Dataset of Real-World Privacy Fixes for Benchmarking Privacy-Aware Code Review</a> — arXiv</li>
<li><a href="https://arxiv.org/html/2611.18293v1">PrivacyPR: A Dataset of Real-World Privacy Fixes for Benchmarking Privacy-Aware Code Review</a> — arXiv</li>
</ul>
<p><b>Research tests history-aware model identity audits</b></p>
<p>Lineup compares responses using matching conversation histories.</p>
<ul>
<li><a href="https://arxiv.org/abs/2611.18104">Contagious Fingerprints: When LLMs Inherit Another Model's Identity</a> — arXiv</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Tue, 17 Nov 2026 07:06:32 +0100</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/f61e8181/1b3dfd85.mp3" length="5214919" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>326</itunes:duration>
      <itunes:summary>California sues OpenAI over experimental-agent safeguards; Liquid AI adds audio input; Hugging Face changes inference routing; NVIDIA reports training results; and researchers examine privacy reviews and model identity checks.</itunes:summary>
      <itunes:subtitle>California sues OpenAI over experimental-agent safeguards; Liquid AI adds audio input; Hugging Face changes inference routing; NVIDIA reports training results; and researchers examine privacy reviews and model identity checks.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/f61e8181/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>Antigravity model handoffs and Cursor’s breach findings</title>
      <itunes:title>Antigravity model handoffs and Cursor’s breach findings</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">2aca5c43-68cf-40a3-b2aa-efc1e56e4af9</guid>
      <link>https://share.transistor.fm/s/df01867d</link>
      <description>
        <![CDATA[<p>Google introduces coding-model orchestration, Allen AI releases XYBench, Cursor details its breach, ZDI discloses a patched llama.cpp flaw, and Mirendil announces funding.</p>
<p><b>Stories and sources</b></p>
<p><b>Google introduces Orchestrator in Antigravity</b></p>
<p>Automatic handoffs pair planning models with Hydrogen for implementation, with plan-dependent access.</p>
<ul>
<li><a href="https://antigravity.google/blog/gemini-4-hydrogen-in-google-antigravity">Gemini 4 Hydrogen and Orchestrator</a> — Google Antigravity</li>
<li><a href="https://antigravity.google/docs/models">Choose a model</a> — Google Antigravity</li>
</ul>
<p><b>Allen AI announces XYBench release</b></p>
<p>The benchmark tests whether assistants identify misconceptions and address underlying goals.</p>
<ul>
<li><a href="https://huggingface.co/blog/allenai/xybench">XYBench: Helping with the problem behind the question</a> — Allen AI on Hugging Face</li>
<li><a href="https://arxiv.org/abs/2609.06842">XYBench: Can LLMs Respond Pragmatically to Queries with Misconceptions?</a> — arXiv</li>
<li><a href="https://github.com/allenai/XYBench">allenai/XYBench</a> — Allen AI on GitHub</li>
</ul>
<p><b>Cursor details the chain behind its security incident</b></p>
<p>Cursor links stolen historical agent traces and exposed credentials to subsequent access and malicious extensions.</p>
<ul>
<li><a href="https://cursor.com/blog/security-incident-update">An update on our recent security incident</a> — Cursor</li>
<li><a href="https://cursor.com/blog/claude-code-extension-security-notice">Security notice — malicious version of Anthropic.claude-code (2.1.312 and 2.1.313) identified in the Cursor VS Code Extension Marketplace</a> — Cursor</li>
</ul>
<p><b>ZDI discloses patched llama.cpp parsing flaw</b></p>
<p>A malicious audio projector file could trigger code execution when opened.</p>
<ul>
<li><a href="https://www.zerodayinitiative.com/advisories/ZDI-26-954/">llama.cpp MTMD Multimodal GGUF File Parsing Integer Overflow Remote Code Execution Vulnerability</a> — TrendAI Zero Day Initiative</li>
</ul>
<p><b>Mirendil announces one-billion-dollar Series A</b></p>
<p>The AI research-tools company reports financing while Lor remains in private beta.</p>
<ul>
<li><a href="https://mirendil.com/news/series-a/">Mirendil Raises $1B to Democratize Frontier AI R&amp;D</a> — Mirendil</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>Google introduces coding-model orchestration, Allen AI releases XYBench, Cursor details its breach, ZDI discloses a patched llama.cpp flaw, and Mirendil announces funding.</p>
<p><b>Stories and sources</b></p>
<p><b>Google introduces Orchestrator in Antigravity</b></p>
<p>Automatic handoffs pair planning models with Hydrogen for implementation, with plan-dependent access.</p>
<ul>
<li><a href="https://antigravity.google/blog/gemini-4-hydrogen-in-google-antigravity">Gemini 4 Hydrogen and Orchestrator</a> — Google Antigravity</li>
<li><a href="https://antigravity.google/docs/models">Choose a model</a> — Google Antigravity</li>
</ul>
<p><b>Allen AI announces XYBench release</b></p>
<p>The benchmark tests whether assistants identify misconceptions and address underlying goals.</p>
<ul>
<li><a href="https://huggingface.co/blog/allenai/xybench">XYBench: Helping with the problem behind the question</a> — Allen AI on Hugging Face</li>
<li><a href="https://arxiv.org/abs/2609.06842">XYBench: Can LLMs Respond Pragmatically to Queries with Misconceptions?</a> — arXiv</li>
<li><a href="https://github.com/allenai/XYBench">allenai/XYBench</a> — Allen AI on GitHub</li>
</ul>
<p><b>Cursor details the chain behind its security incident</b></p>
<p>Cursor links stolen historical agent traces and exposed credentials to subsequent access and malicious extensions.</p>
<ul>
<li><a href="https://cursor.com/blog/security-incident-update">An update on our recent security incident</a> — Cursor</li>
<li><a href="https://cursor.com/blog/claude-code-extension-security-notice">Security notice — malicious version of Anthropic.claude-code (2.1.312 and 2.1.313) identified in the Cursor VS Code Extension Marketplace</a> — Cursor</li>
</ul>
<p><b>ZDI discloses patched llama.cpp parsing flaw</b></p>
<p>A malicious audio projector file could trigger code execution when opened.</p>
<ul>
<li><a href="https://www.zerodayinitiative.com/advisories/ZDI-26-954/">llama.cpp MTMD Multimodal GGUF File Parsing Integer Overflow Remote Code Execution Vulnerability</a> — TrendAI Zero Day Initiative</li>
</ul>
<p><b>Mirendil announces one-billion-dollar Series A</b></p>
<p>The AI research-tools company reports financing while Lor remains in private beta.</p>
<ul>
<li><a href="https://mirendil.com/news/series-a/">Mirendil Raises $1B to Democratize Frontier AI R&amp;D</a> — Mirendil</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Mon, 16 Nov 2026 07:06:25 +0100</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/df01867d/1c9e4e7f.mp3" length="4169185" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>261</itunes:duration>
      <itunes:summary>Google introduces coding-model orchestration, Allen AI releases XYBench, Cursor details its breach, ZDI discloses a patched llama.cpp flaw, and Mirendil announces funding.</itunes:summary>
      <itunes:subtitle>Google introduces coding-model orchestration, Allen AI releases XYBench, Cursor details its breach, ZDI discloses a patched llama.cpp flaw, and Mirendil announces funding.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/df01867d/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>DeepSeek expands APIs; Inkling weights released</title>
      <itunes:title>DeepSeek expands APIs; Inkling weights released</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">713e18c4-9812-4f95-95da-2b93ee3227b3</guid>
      <link>https://share.transistor.fm/s/7fb38d9f</link>
      <description>
        <![CDATA[<p>Developer tools from DeepSeek and Perplexity, downloadable Inkling weights, agent-training research, and Cohere's security and coding acquisitions.</p>
<p><b>Stories and sources</b></p>
<p><b>DeepSeek opens application capabilities through its API</b></p>
<p>Managed search, document input and link reading become available to developers.</p>
<ul>
<li><a href="https://api-docs.deepseek.com/updates/">Change Log</a> — DeepSeek</li>
<li><a href="https://api-docs.deepseek.com/guides/build_a_chatbot/">Build a Chatbot</a> — DeepSeek</li>
</ul>
<p><b>Thinking Machines releases Inkling 2 Small weights</b></p>
<p>Downloadable weights arrive with substantial GPU-memory requirements.</p>
<ul>
<li><a href="https://thinkingmachines.ai/news/inkling-2-small/">Introducing Inkling 2 Small</a> — Thinking Machines Lab</li>
<li><a href="https://huggingface.co/thinkingmachines/Inkling-2-Small">thinkingmachines/Inkling-2-Small</a> — Thinking Machines Lab</li>
</ul>
<p><b>Perplexity adds extraction to Decider models</b></p>
<p>Decider v1.3 and Flash add Locate, with documented processing and access conditions.</p>
<ul>
<li><a href="https://www.perplexity.ai/en-GB/hub/blog/perplexity-decider">Perplexity Decider v1.3: Training, eval, and use cases</a> — Perplexity</li>
<li><a href="https://huggingface.co/perplexity-ai/pplx-decider-v1.3">perplexity-ai/pplx-decider-v1.3</a> — Perplexity</li>
</ul>
<p><b>ACE-RL trains agents to improve task playbooks</b></p>
<p>Researchers report gains in bounded experiments with small models and task-specific memory.</p>
<ul>
<li><a href="https://arxiv.org/abs/2611.16409">Experience-Handoff Reinforcement Learning for Recursive Self-Improvement</a> — arXiv</li>
<li><a href="https://arxiv.org/pdf/2611.16409">Experience-Handoff Reinforcement Learning for Recursive Self-Improvement</a> — arXiv</li>
</ul>
<p><b>Cohere adds AISLE and cubic</b></p>
<p>Completed additions support cybersecurity and planned coding integration.</p>
<ul>
<li><a href="https://cohere.com/blog/cohere-expands-its-sovereign-ai-platform-with-aisle-and-cubic">Cohere Expands Its Sovereign AI Platform with AISLE and cubic</a> — Cohere</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>Developer tools from DeepSeek and Perplexity, downloadable Inkling weights, agent-training research, and Cohere's security and coding acquisitions.</p>
<p><b>Stories and sources</b></p>
<p><b>DeepSeek opens application capabilities through its API</b></p>
<p>Managed search, document input and link reading become available to developers.</p>
<ul>
<li><a href="https://api-docs.deepseek.com/updates/">Change Log</a> — DeepSeek</li>
<li><a href="https://api-docs.deepseek.com/guides/build_a_chatbot/">Build a Chatbot</a> — DeepSeek</li>
</ul>
<p><b>Thinking Machines releases Inkling 2 Small weights</b></p>
<p>Downloadable weights arrive with substantial GPU-memory requirements.</p>
<ul>
<li><a href="https://thinkingmachines.ai/news/inkling-2-small/">Introducing Inkling 2 Small</a> — Thinking Machines Lab</li>
<li><a href="https://huggingface.co/thinkingmachines/Inkling-2-Small">thinkingmachines/Inkling-2-Small</a> — Thinking Machines Lab</li>
</ul>
<p><b>Perplexity adds extraction to Decider models</b></p>
<p>Decider v1.3 and Flash add Locate, with documented processing and access conditions.</p>
<ul>
<li><a href="https://www.perplexity.ai/en-GB/hub/blog/perplexity-decider">Perplexity Decider v1.3: Training, eval, and use cases</a> — Perplexity</li>
<li><a href="https://huggingface.co/perplexity-ai/pplx-decider-v1.3">perplexity-ai/pplx-decider-v1.3</a> — Perplexity</li>
</ul>
<p><b>ACE-RL trains agents to improve task playbooks</b></p>
<p>Researchers report gains in bounded experiments with small models and task-specific memory.</p>
<ul>
<li><a href="https://arxiv.org/abs/2611.16409">Experience-Handoff Reinforcement Learning for Recursive Self-Improvement</a> — arXiv</li>
<li><a href="https://arxiv.org/pdf/2611.16409">Experience-Handoff Reinforcement Learning for Recursive Self-Improvement</a> — arXiv</li>
</ul>
<p><b>Cohere adds AISLE and cubic</b></p>
<p>Completed additions support cybersecurity and planned coding integration.</p>
<ul>
<li><a href="https://cohere.com/blog/cohere-expands-its-sovereign-ai-platform-with-aisle-and-cubic">Cohere Expands Its Sovereign AI Platform with AISLE and cubic</a> — Cohere</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Sun, 15 Nov 2026 07:06:02 +0100</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/7fb38d9f/877f9471.mp3" length="4947007" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>310</itunes:duration>
      <itunes:summary>Developer tools from DeepSeek and Perplexity, downloadable Inkling weights, agent-training research, and Cohere's security and coding acquisitions.</itunes:summary>
      <itunes:subtitle>Developer tools from DeepSeek and Perplexity, downloadable Inkling weights, agent-training research, and Cohere's security and coding acquisitions.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/7fb38d9f/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>Local AI libraries, weather datasets and model backdoors</title>
      <itunes:title>Local AI libraries, weather datasets and model backdoors</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">0e0a3db7-debf-4e13-a384-5e4942aa0230</guid>
      <link>https://share.transistor.fm/s/07f970da</link>
      <description>
        <![CDATA[<p>New llama.cpp and speech-model releases, DeepMind weather datasets, Microsoft backdoor research, NVIDIA workload inventory and a planned inference infrastructure deployment.</p>
<p><b>Stories and sources</b></p>
<p><b>llama.cpp and ggml release version 1.0.0</b></p>
<p>The releases establish compatibility rules and add audio, vision and server changes.</p>
<ul>
<li><a href="https://huggingface.co/blog/ggml-org/llama-cpp-v1">llama.cpp v1.0.0 : A stronger foundation for the local AI ecosystem</a> — ggml-org</li>
<li><a href="https://github.com/ggml-org/llama.cpp/releases/tag/v1.0.0">Release v1.0.0</a> — ggml-org</li>
</ul>
<p><b>DeepMind releases SatWeave and PARDIG</b></p>
<p>Weather datasets become available as historical archives and live feeds.</p>
<ul>
<li><a href="https://deepmind.google/blog/opening-the-datasets-behind-our-ai-weather-models/">Opening the datasets behind our AI weather models</a> — Google DeepMind</li>
</ul>
<p><b>ElevenLabs launches Scribe v3 and v3 Live</b></p>
<p>API releases add live speaker labels and vendor-reported transcription improvements.</p>
<ul>
<li><a href="https://elevenlabs.io/blog/scribe-v3">Introducing Scribe v3 and v3 Live, our most accurate speech to text models</a> — ElevenLabs</li>
</ul>
<p><b>Microsoft demonstrates dormant-neuron backdoors</b></p>
<p>Lazarus research alters model weights; Microsoft reports no observed attacker weaponization.</p>
<ul>
<li><a href="https://www.microsoft.com/en-us/research/blog/lazare-veni-foras-reprogramming-dormant-neurons-to-backdoor-ai-agents/">Lazare, veni foras: Reprogramming Dormant Neurons to Backdoor AI Agents</a> — Microsoft Research</li>
<li><a href="https://www.microsoft.com/en-us/security/blog/2026/11/10/detecting-sleeper-agents-hiding-in-ai-model-neurons/">Detecting sleeper agents hiding in AI model neurons</a> — Microsoft Security</li>
</ul>
<p><b>NVIDIA adds runtime AI workload inventory</b></p>
<p>The integration records supported workloads with explicit coverage and data-access limits.</p>
<ul>
<li><a href="https://developer.nvidia.com/blog/validating-gpu-platforms-with-runtime-ai-inventory-via-the-nvidia-dsx-ai-factory/">Validating GPU Platforms with Runtime AI Inventory via the NVIDIA DSX AI Factory</a> — NVIDIA</li>
</ul>
<p><b>Rackspace and Together AI announce infrastructure agreement</b></p>
<p>Deployment is planned for 2027, subject to conditions.</p>
<ul>
<li><a href="https://ir.rackspace.com/node/13806/pdf">Rackspace and Together AI Announce Strategic Agreement to Support Enterprise AI Inference</a> — Rackspace</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>New llama.cpp and speech-model releases, DeepMind weather datasets, Microsoft backdoor research, NVIDIA workload inventory and a planned inference infrastructure deployment.</p>
<p><b>Stories and sources</b></p>
<p><b>llama.cpp and ggml release version 1.0.0</b></p>
<p>The releases establish compatibility rules and add audio, vision and server changes.</p>
<ul>
<li><a href="https://huggingface.co/blog/ggml-org/llama-cpp-v1">llama.cpp v1.0.0 : A stronger foundation for the local AI ecosystem</a> — ggml-org</li>
<li><a href="https://github.com/ggml-org/llama.cpp/releases/tag/v1.0.0">Release v1.0.0</a> — ggml-org</li>
</ul>
<p><b>DeepMind releases SatWeave and PARDIG</b></p>
<p>Weather datasets become available as historical archives and live feeds.</p>
<ul>
<li><a href="https://deepmind.google/blog/opening-the-datasets-behind-our-ai-weather-models/">Opening the datasets behind our AI weather models</a> — Google DeepMind</li>
</ul>
<p><b>ElevenLabs launches Scribe v3 and v3 Live</b></p>
<p>API releases add live speaker labels and vendor-reported transcription improvements.</p>
<ul>
<li><a href="https://elevenlabs.io/blog/scribe-v3">Introducing Scribe v3 and v3 Live, our most accurate speech to text models</a> — ElevenLabs</li>
</ul>
<p><b>Microsoft demonstrates dormant-neuron backdoors</b></p>
<p>Lazarus research alters model weights; Microsoft reports no observed attacker weaponization.</p>
<ul>
<li><a href="https://www.microsoft.com/en-us/research/blog/lazare-veni-foras-reprogramming-dormant-neurons-to-backdoor-ai-agents/">Lazare, veni foras: Reprogramming Dormant Neurons to Backdoor AI Agents</a> — Microsoft Research</li>
<li><a href="https://www.microsoft.com/en-us/security/blog/2026/11/10/detecting-sleeper-agents-hiding-in-ai-model-neurons/">Detecting sleeper agents hiding in AI model neurons</a> — Microsoft Security</li>
</ul>
<p><b>NVIDIA adds runtime AI workload inventory</b></p>
<p>The integration records supported workloads with explicit coverage and data-access limits.</p>
<ul>
<li><a href="https://developer.nvidia.com/blog/validating-gpu-platforms-with-runtime-ai-inventory-via-the-nvidia-dsx-ai-factory/">Validating GPU Platforms with Runtime AI Inventory via the NVIDIA DSX AI Factory</a> — NVIDIA</li>
</ul>
<p><b>Rackspace and Together AI announce infrastructure agreement</b></p>
<p>Deployment is planned for 2027, subject to conditions.</p>
<ul>
<li><a href="https://ir.rackspace.com/node/13806/pdf">Rackspace and Together AI Announce Strategic Agreement to Support Enterprise AI Inference</a> — Rackspace</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Sat, 14 Nov 2026 07:06:06 +0100</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/07f970da/3739e92e.mp3" length="5461933" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>342</itunes:duration>
      <itunes:summary>New llama.cpp and speech-model releases, DeepMind weather datasets, Microsoft backdoor research, NVIDIA workload inventory and a planned inference infrastructure deployment.</itunes:summary>
      <itunes:subtitle>New llama.cpp and speech-model releases, DeepMind weather datasets, Microsoft backdoor research, NVIDIA workload inventory and a planned inference infrastructure deployment.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/07f970da/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>Granite previews, driving models and AI intrusion findings</title>
      <itunes:title>Granite previews, driving models and AI intrusion findings</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">13de62f9-6bee-4f9f-b09f-57cb1efe59f8</guid>
      <link>https://share.transistor.fm/s/b2f250e4</link>
      <description>
        <![CDATA[<p>Experimental models from IBM, NVIDIA’s driving research, BotPsy, Forescout’s vulnerability findings, Unit 42’s intrusion report and Mercor’s financing.</p>
<p><b>Stories and sources</b></p>
<p><b>IBM releases experimental LLMON models</b></p>
<p>Granite previews explicitly separate instructions and data.</p>
<ul>
<li><a href="https://huggingface.co/blog/ibm-granite/llmon-an-llm-native-markup-language-now-in-granite">LLMON: An LLM-Native Markup Language, now in Granite 4.1 3B and 8B</a> — IBM Granite via Hugging Face</li>
<li><a href="https://huggingface.co/ibm-granite/granite-4.1-3b-llmon-preview">ibm-granite/granite-4.1-3b-llmon-preview</a> — IBM Granite via Hugging Face</li>
</ul>
<p><b>NVIDIA releases Alpamayo 2 Edge</b></p>
<p>A compact driving model becomes available with simulation results.</p>
<ul>
<li><a href="https://developer.nvidia.com/blog/running-av-reasoning-models-within-onboard-compute-and-latency-limits/">Running AV Reasoning Models Within Onboard Compute and Latency Limits</a> — NVIDIA</li>
<li><a href="https://huggingface.co/nvidia/Alpamayo2-Edge">nvidia/Alpamayo2-Edge</a> — NVIDIA via Hugging Face</li>
</ul>
<p><b>Tether introduces BotPsy</b></p>
<p>A small tool-calling model arrives with research-use limitations.</p>
<ul>
<li><a href="https://huggingface.co/blog/qvac/botpsy">BotPsy: On-Device Personal Assistant with a Sub-Billion-Parameter Agentic Language Model</a> — Tether AI Research via Hugging Face</li>
<li><a href="https://huggingface.co/qvac/BotPsy-0.8B">qvac/BotPsy-0.8B</a> — Tether AI Research via Hugging Face</li>
</ul>
<p><b>Forescout reports AI-assisted vulnerability discoveries</b></p>
<p>Researchers report 99 validated flaws, with setup costs excluded from their comparison.</p>
<ul>
<li><a href="https://wgem.marketminute.com/article/bizwire-2026-11-12-forescout-research-shows-ai-cut-vulnerability-hunting-from-nine-months-to-90-hours">Forescout Research Shows AI Cut Vulnerability Hunting from Nine Months to 90 Hours</a> — Forescout via Business Wire, syndicated by WGEM MarketMinute</li>
<li><a href="https://www.forescout.com/research-labs/finding-critical-infrastructure-bugs-with-off-the-shelf-ai/">Finding Critical Infrastructure Bugs with Off-the-Shelf AI</a> — Forescout</li>
</ul>
<p><b>Unit 42 reports Cairn intrusions</b></p>
<p>Observed operations carry qualified China-nexus attribution.</p>
<ul>
<li><a href="https://unit42.paloaltonetworks.com/cairn-framework-ai-autonomous-attacks/">Attackers Use Autonomous AI to Target Semiconductor, Defense and Civil Society Sectors</a> — Palo Alto Networks Unit 42</li>
</ul>
<p><b>Mercor announces Series D</b></p>
<p>Financing includes an employee tender.</p>
<ul>
<li><a href="https://www.mercor.com/blog/series-d/">Organizing human intelligence for the entire economy</a> — Mercor</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>Experimental models from IBM, NVIDIA’s driving research, BotPsy, Forescout’s vulnerability findings, Unit 42’s intrusion report and Mercor’s financing.</p>
<p><b>Stories and sources</b></p>
<p><b>IBM releases experimental LLMON models</b></p>
<p>Granite previews explicitly separate instructions and data.</p>
<ul>
<li><a href="https://huggingface.co/blog/ibm-granite/llmon-an-llm-native-markup-language-now-in-granite">LLMON: An LLM-Native Markup Language, now in Granite 4.1 3B and 8B</a> — IBM Granite via Hugging Face</li>
<li><a href="https://huggingface.co/ibm-granite/granite-4.1-3b-llmon-preview">ibm-granite/granite-4.1-3b-llmon-preview</a> — IBM Granite via Hugging Face</li>
</ul>
<p><b>NVIDIA releases Alpamayo 2 Edge</b></p>
<p>A compact driving model becomes available with simulation results.</p>
<ul>
<li><a href="https://developer.nvidia.com/blog/running-av-reasoning-models-within-onboard-compute-and-latency-limits/">Running AV Reasoning Models Within Onboard Compute and Latency Limits</a> — NVIDIA</li>
<li><a href="https://huggingface.co/nvidia/Alpamayo2-Edge">nvidia/Alpamayo2-Edge</a> — NVIDIA via Hugging Face</li>
</ul>
<p><b>Tether introduces BotPsy</b></p>
<p>A small tool-calling model arrives with research-use limitations.</p>
<ul>
<li><a href="https://huggingface.co/blog/qvac/botpsy">BotPsy: On-Device Personal Assistant with a Sub-Billion-Parameter Agentic Language Model</a> — Tether AI Research via Hugging Face</li>
<li><a href="https://huggingface.co/qvac/BotPsy-0.8B">qvac/BotPsy-0.8B</a> — Tether AI Research via Hugging Face</li>
</ul>
<p><b>Forescout reports AI-assisted vulnerability discoveries</b></p>
<p>Researchers report 99 validated flaws, with setup costs excluded from their comparison.</p>
<ul>
<li><a href="https://wgem.marketminute.com/article/bizwire-2026-11-12-forescout-research-shows-ai-cut-vulnerability-hunting-from-nine-months-to-90-hours">Forescout Research Shows AI Cut Vulnerability Hunting from Nine Months to 90 Hours</a> — Forescout via Business Wire, syndicated by WGEM MarketMinute</li>
<li><a href="https://www.forescout.com/research-labs/finding-critical-infrastructure-bugs-with-off-the-shelf-ai/">Finding Critical Infrastructure Bugs with Off-the-Shelf AI</a> — Forescout</li>
</ul>
<p><b>Unit 42 reports Cairn intrusions</b></p>
<p>Observed operations carry qualified China-nexus attribution.</p>
<ul>
<li><a href="https://unit42.paloaltonetworks.com/cairn-framework-ai-autonomous-attacks/">Attackers Use Autonomous AI to Target Semiconductor, Defense and Civil Society Sectors</a> — Palo Alto Networks Unit 42</li>
</ul>
<p><b>Mercor announces Series D</b></p>
<p>Financing includes an employee tender.</p>
<ul>
<li><a href="https://www.mercor.com/blog/series-d/">Organizing human intelligence for the entire economy</a> — Mercor</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Fri, 13 Nov 2026 07:07:57 +0100</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/b2f250e4/72a51903.mp3" length="5629953" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>352</itunes:duration>
      <itunes:summary>Experimental models from IBM, NVIDIA’s driving research, BotPsy, Forescout’s vulnerability findings, Unit 42’s intrusion report and Mercor’s financing.</itunes:summary>
      <itunes:subtitle>Experimental models from IBM, NVIDIA’s driving research, BotPsy, Forescout’s vulnerability findings, Unit 42’s intrusion report and Mercor’s financing.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/b2f250e4/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>Gemini Hydrogen, agent infrastructure and AI financing</title>
      <itunes:title>Gemini Hydrogen, agent infrastructure and AI financing</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">8782591e-df0d-411f-bdb9-21a19b2bf725</guid>
      <link>https://share.transistor.fm/s/d8f3f846</link>
      <description>
        <![CDATA[<p>Google releases Hydrogen and reports a robotics pilot, NVIDIA announces infrastructure tools, researchers test adaptive prompt injection, and Anthropic and Fractile detail financing developments.</p>
<p><b>Stories and sources</b></p>
<p><b>Google releases Gemini 4 Hydrogen through its API</b></p>
<p>Hydrogen reaches general availability with free-tier access and four thinking settings.</p>
<ul>
<li><a href="https://aistudio.google.com/docs/changelog">Release notes</a> — Google</li>
</ul>
<p><b>DeepMind rolls out Gemini Robotics 2.5 to partners</b></p>
<p>DeepMind reports a 96% success rate in an automotive parts-handling pilot.</p>
<ul>
<li><a href="https://deepmind.google/blog/gemini-robotics-2-5-accelerates-real-world-physical-intelligence/">Gemini Robotics 2.5 accelerates real-world general dexterity</a> — Google DeepMind</li>
</ul>
<p><b>NVIDIA announces Snapshot and OpenShell infrastructure developments</b></p>
<p>Snapshot restores initialized model workloads; NVIDIA also announces OpenShell's CNCF contribution and integration work.</p>
<ul>
<li><a href="https://developer.nvidia.com/blog/start-and-restore-gpu-workloads-in-seconds-with-nvidia-snapshot/">Start and Restore GPU Workloads in Seconds with NVIDIA Snapshot</a> — NVIDIA</li>
<li><a href="https://blogs.nvidia.com/blog/open-source-openshell-infrastructure/">NVIDIA Expands Open Source Commitment With OpenShell Contribution and New AI Infrastructure</a> — NVIDIA</li>
</ul>
<p><b>DeepMind researchers test adaptive prompt-injection swarms</b></p>
<p>Controlled experiments find instruction-following failures with external Gemini safeguards disabled.</p>
<ul>
<li><a href="https://arxiv.org/abs/2611.11861">Scaling Adaptive Prompt Injection Attacks with Agent Swarms</a> — arXiv</li>
<li><a href="https://arxiv.org/html/2611.11861v1">Scaling Adaptive Prompt Injection Attacks with Agent Swarms</a> — arXiv</li>
</ul>
<p><b>Anthropic launches roadshow for proposed IPO</b></p>
<p>The company announces a proposed share offering and expected price range.</p>
<ul>
<li><a href="https://www.anthropic.com/news/roadshow-launch">Anthropic launches roadshow for proposed initial public offering</a> — Anthropic</li>
<li><a href="https://www.sec.gov/Archives/edgar/data/2130357/000162828026072484/anthropicpbc-sx1ax2.htm">Anthropic, PBC — Amendment No. 2 to Form S-1</a> — U.S. Securities and Exchange Commission</li>
</ul>
<p><b>Fractile reports $720 million Series C</b></p>
<p>Reuters reports funding and a separate valuation ahead of a planned commercial chip launch.</p>
<ul>
<li><a href="https://www.marketscreener.com/news/uk-backed-ai-chip-startup-fractile-hits-7-billion-valuation-in-funding-round-ce785edbdf8ef221">UK-backed AI chip startup Fractile hits $7 billion valuation in funding round</a> — Reuters via MarketScreener</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>Google releases Hydrogen and reports a robotics pilot, NVIDIA announces infrastructure tools, researchers test adaptive prompt injection, and Anthropic and Fractile detail financing developments.</p>
<p><b>Stories and sources</b></p>
<p><b>Google releases Gemini 4 Hydrogen through its API</b></p>
<p>Hydrogen reaches general availability with free-tier access and four thinking settings.</p>
<ul>
<li><a href="https://aistudio.google.com/docs/changelog">Release notes</a> — Google</li>
</ul>
<p><b>DeepMind rolls out Gemini Robotics 2.5 to partners</b></p>
<p>DeepMind reports a 96% success rate in an automotive parts-handling pilot.</p>
<ul>
<li><a href="https://deepmind.google/blog/gemini-robotics-2-5-accelerates-real-world-physical-intelligence/">Gemini Robotics 2.5 accelerates real-world general dexterity</a> — Google DeepMind</li>
</ul>
<p><b>NVIDIA announces Snapshot and OpenShell infrastructure developments</b></p>
<p>Snapshot restores initialized model workloads; NVIDIA also announces OpenShell's CNCF contribution and integration work.</p>
<ul>
<li><a href="https://developer.nvidia.com/blog/start-and-restore-gpu-workloads-in-seconds-with-nvidia-snapshot/">Start and Restore GPU Workloads in Seconds with NVIDIA Snapshot</a> — NVIDIA</li>
<li><a href="https://blogs.nvidia.com/blog/open-source-openshell-infrastructure/">NVIDIA Expands Open Source Commitment With OpenShell Contribution and New AI Infrastructure</a> — NVIDIA</li>
</ul>
<p><b>DeepMind researchers test adaptive prompt-injection swarms</b></p>
<p>Controlled experiments find instruction-following failures with external Gemini safeguards disabled.</p>
<ul>
<li><a href="https://arxiv.org/abs/2611.11861">Scaling Adaptive Prompt Injection Attacks with Agent Swarms</a> — arXiv</li>
<li><a href="https://arxiv.org/html/2611.11861v1">Scaling Adaptive Prompt Injection Attacks with Agent Swarms</a> — arXiv</li>
</ul>
<p><b>Anthropic launches roadshow for proposed IPO</b></p>
<p>The company announces a proposed share offering and expected price range.</p>
<ul>
<li><a href="https://www.anthropic.com/news/roadshow-launch">Anthropic launches roadshow for proposed initial public offering</a> — Anthropic</li>
<li><a href="https://www.sec.gov/Archives/edgar/data/2130357/000162828026072484/anthropicpbc-sx1ax2.htm">Anthropic, PBC — Amendment No. 2 to Form S-1</a> — U.S. Securities and Exchange Commission</li>
</ul>
<p><b>Fractile reports $720 million Series C</b></p>
<p>Reuters reports funding and a separate valuation ahead of a planned commercial chip launch.</p>
<ul>
<li><a href="https://www.marketscreener.com/news/uk-backed-ai-chip-startup-fractile-hits-7-billion-valuation-in-funding-round-ce785edbdf8ef221">UK-backed AI chip startup Fractile hits $7 billion valuation in funding round</a> — Reuters via MarketScreener</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Thu, 12 Nov 2026 22:36:48 +0100</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/d8f3f846/4663e2e7.mp3" length="4815768" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>301</itunes:duration>
      <itunes:summary>Google releases Hydrogen and reports a robotics pilot, NVIDIA announces infrastructure tools, researchers test adaptive prompt injection, and Anthropic and Fractile detail financing developments.</itunes:summary>
      <itunes:subtitle>Google releases Hydrogen and reports a robotics pilot, NVIDIA announces infrastructure tools, researchers test adaptive prompt injection, and Anthropic and Fractile detail financing developments.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/d8f3f846/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>Carbon access expands, Fable 5.2 launches</title>
      <itunes:title>Carbon access expands, Fable 5.2 launches</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">4aa664a4-df8a-48da-9fe8-c9506c3af7bc</guid>
      <link>https://share.transistor.fm/s/77de5a7f</link>
      <description>
        <![CDATA[<p>New model access and pricing, AlphaGenome 2 research tools, Signal verification research, Cisco’s planned AI testing integration and Rox funding.</p>
<p><b>Stories and sources</b></p>
<p><b>Google expands Gemini 4 Carbon access</b></p>
<p>Carbon reaches paying developers and enterprises, with an Ultra rollout and a separate swarm preview.</p>
<ul>
<li><a href="https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-carbon-now-available/?utm_campaign=gdm&amp;utm_content=&amp;utm_medium=referral&amp;utm_source=deepmind.google">Gemini 4 Carbon is now available</a> — Google</li>
<li><a href="https://aistudio.google.com/docs/changelog">Release notes</a> — Google</li>
</ul>
<p><b>Anthropic releases Claude Fable 5.2</b></p>
<p>The release lowers base token prices and extends conversation-history checks to every account.</p>
<ul>
<li><a href="https://platform.claude.com/docs/en/release-notes/overview">Claude Platform release notes</a> — Anthropic</li>
<li><a href="https://platform.claude.com/docs/en/models/fable-5-2/whats-new-fable-5-2">What’s new in Claude Fable 5.2</a> — Anthropic</li>
<li><a href="https://platform.claude.com/docs/en/models/fable-5-2/migration-guide">Claude Fable 5.2 and Claude Mythos 5.2 migration guide</a> — Anthropic</li>
</ul>
<p><b>DeepMind introduces AlphaGenome 2</b></p>
<p>Cell-specific DNA predictions become available for research, with commercial access in private preview.</p>
<ul>
<li><a href="https://deepmind.google/blog/alphagenome-2-deciphering-how-cells-read-dna-to-better-understand-disease/">AlphaGenome 2: Deciphering how cells read DNA to better understand disease</a> — Google DeepMind</li>
</ul>
<p><b>DeepMind reports Signal code verification</b></p>
<p>AlphaProof Code generates checked security proofs under explicit implementation and threat-model limits.</p>
<ul>
<li><a href="https://deepmind.google/blog/introducing-alphaproof-code/">Introducing AlphaProof Code: a breakthrough in verifying complex software at depth</a> — Google DeepMind</li>
<li><a href="https://deepmind.google/documents/36/Formal_Verification_of_Signal_4.pdf">AlphaProof Code: Verifying Signal’s Security at the Machine-Code Level</a> — Google DeepMind and Google</li>
</ul>
<p><b>Cisco plans Mythos integration for AI testing</b></p>
<p>Selected enterprises will receive controlled adversarial testing access through AI Defense.</p>
<ul>
<li><a href="https://investor.cisco.com/news/news-details/2026/Cisco-Unveils-Frontier-AI-Red-Teaming-and-Agent-Observability-to-Secure-the-Agentic-Enterprise/default.aspx">Cisco Unveils Frontier AI Red Teaming and Agent Observability to Secure the Agentic Enterprise</a> — Cisco</li>
</ul>
<p><b>Rox raises $100 million Series B</b></p>
<p>Reuters reports funding at a separate $1.3 billion valuation for the sales-agent developer.</p>
<ul>
<li><a href="https://www.marketscreener.com/news/automation-startup-rox-raises-100-million-at-1-3-billion-valuation-ce785edbdf8df221">Automation startup Rox raises $100 million at $1.3 billion valuation</a> — Reuters via MarketScreener</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>New model access and pricing, AlphaGenome 2 research tools, Signal verification research, Cisco’s planned AI testing integration and Rox funding.</p>
<p><b>Stories and sources</b></p>
<p><b>Google expands Gemini 4 Carbon access</b></p>
<p>Carbon reaches paying developers and enterprises, with an Ultra rollout and a separate swarm preview.</p>
<ul>
<li><a href="https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-carbon-now-available/?utm_campaign=gdm&amp;utm_content=&amp;utm_medium=referral&amp;utm_source=deepmind.google">Gemini 4 Carbon is now available</a> — Google</li>
<li><a href="https://aistudio.google.com/docs/changelog">Release notes</a> — Google</li>
</ul>
<p><b>Anthropic releases Claude Fable 5.2</b></p>
<p>The release lowers base token prices and extends conversation-history checks to every account.</p>
<ul>
<li><a href="https://platform.claude.com/docs/en/release-notes/overview">Claude Platform release notes</a> — Anthropic</li>
<li><a href="https://platform.claude.com/docs/en/models/fable-5-2/whats-new-fable-5-2">What’s new in Claude Fable 5.2</a> — Anthropic</li>
<li><a href="https://platform.claude.com/docs/en/models/fable-5-2/migration-guide">Claude Fable 5.2 and Claude Mythos 5.2 migration guide</a> — Anthropic</li>
</ul>
<p><b>DeepMind introduces AlphaGenome 2</b></p>
<p>Cell-specific DNA predictions become available for research, with commercial access in private preview.</p>
<ul>
<li><a href="https://deepmind.google/blog/alphagenome-2-deciphering-how-cells-read-dna-to-better-understand-disease/">AlphaGenome 2: Deciphering how cells read DNA to better understand disease</a> — Google DeepMind</li>
</ul>
<p><b>DeepMind reports Signal code verification</b></p>
<p>AlphaProof Code generates checked security proofs under explicit implementation and threat-model limits.</p>
<ul>
<li><a href="https://deepmind.google/blog/introducing-alphaproof-code/">Introducing AlphaProof Code: a breakthrough in verifying complex software at depth</a> — Google DeepMind</li>
<li><a href="https://deepmind.google/documents/36/Formal_Verification_of_Signal_4.pdf">AlphaProof Code: Verifying Signal’s Security at the Machine-Code Level</a> — Google DeepMind and Google</li>
</ul>
<p><b>Cisco plans Mythos integration for AI testing</b></p>
<p>Selected enterprises will receive controlled adversarial testing access through AI Defense.</p>
<ul>
<li><a href="https://investor.cisco.com/news/news-details/2026/Cisco-Unveils-Frontier-AI-Red-Teaming-and-Agent-Observability-to-Secure-the-Agentic-Enterprise/default.aspx">Cisco Unveils Frontier AI Red Teaming and Agent Observability to Secure the Agentic Enterprise</a> — Cisco</li>
</ul>
<p><b>Rox raises $100 million Series B</b></p>
<p>Reuters reports funding at a separate $1.3 billion valuation for the sales-agent developer.</p>
<ul>
<li><a href="https://www.marketscreener.com/news/automation-startup-rox-raises-100-million-at-1-3-billion-valuation-ce785edbdf8df221">Automation startup Rox raises $100 million at $1.3 billion valuation</a> — Reuters via MarketScreener</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Wed, 11 Nov 2026 18:55:50 +0100</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/77de5a7f/64212584.mp3" length="5371654" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>336</itunes:duration>
      <itunes:summary>New model access and pricing, AlphaGenome 2 research tools, Signal verification research, Cisco’s planned AI testing integration and Rox funding.</itunes:summary>
      <itunes:subtitle>New model access and pricing, AlphaGenome 2 research tools, Signal verification research, Cisco’s planned AI testing integration and Rox funding.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/77de5a7f/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>Inkling 2 Small, Meta models and ARTEX campaigns</title>
      <itunes:title>Inkling 2 Small, Meta models and ARTEX campaigns</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">3813f6dd-4125-4181-b7e9-0c94c146e2aa</guid>
      <link>https://share.transistor.fm/s/b202ed40</link>
      <description>
        <![CDATA[<p>New models from Thinking Machines and Meta, Amazon shopping by text, Check Point's ARTEX findings and Kalepa's funding round.</p>
<p><b>Stories and sources</b></p>
<p><b>Thinking Machines introduces Inkling 2 Small</b></p>
<p>Tinker access precedes a planned release of downloadable weights following external safety testing.</p>
<ul>
<li><a href="https://thinkingmachines.ai/news/inkling-2-small/">Introducing Inkling 2 Small</a> — Thinking Machines Lab</li>
</ul>
<p><b>Meta rolls out specialized Muse and Meta AI models</b></p>
<p>The rollout brings Active Memory and Active Focus, with company-reported memory evaluation results.</p>
<ul>
<li><a href="https://research.meta.ai/blog/specialized-models-for-muse-and-metaai">Bringing Specialized Models to Muse and Meta AI</a> — Meta AI Research</li>
</ul>
<p><b>Amazon begins invited access to shopping by text</b></p>
<p>Alexa for Shopping adds text access for selected US customers, with broader access planned.</p>
<ul>
<li><a href="https://www.aboutamazon.com/news/company-news/amazon-news-today-top-stories-company">What you need to know about Amazon today: November 6, 2026</a> — Amazon</li>
<li><a href="https://www.aboutamazon.com/news/retail/alexa-for-shopping-text-to-shop">Text, shop, done: Alexa for Shopping is now just a text away</a> — Amazon</li>
</ul>
<p><b>Check Point reports ARTEX cybercrime campaigns</b></p>
<p>Operator databases reveal agent-driven compromises alongside failed cryptocurrency theft objectives.</p>
<ul>
<li><a href="https://research.checkpoint.com/2026/hacking-on-autopilot-inside-artex-cybercrime-campaigns/">Hacking on Autopilot: Inside ARTEX Cybercrime Campaigns</a> — Check Point Research</li>
<li><a href="https://blog.checkpoint.com/research/inside-artex-how-autonomous-ai-agents-are-being-used-in-real-world-cyberattacks">Inside ARTEX: How Autonomous AI Agents Are Being Used in Real-World Cyberattacks</a> — Check Point Research</li>
</ul>
<p><b>Kalepa announces $75 million Series B</b></p>
<p>The insurance AI company announces funding at a $470 million valuation.</p>
<ul>
<li><a href="https://kalepa.com/news-posts/kalepa-raises-75-million-series-b">Kalepa Raises $75 Million Series B at a $470 Million Valuation to Expand Professional Grade AI Across Insurance</a> — Kalepa</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>New models from Thinking Machines and Meta, Amazon shopping by text, Check Point's ARTEX findings and Kalepa's funding round.</p>
<p><b>Stories and sources</b></p>
<p><b>Thinking Machines introduces Inkling 2 Small</b></p>
<p>Tinker access precedes a planned release of downloadable weights following external safety testing.</p>
<ul>
<li><a href="https://thinkingmachines.ai/news/inkling-2-small/">Introducing Inkling 2 Small</a> — Thinking Machines Lab</li>
</ul>
<p><b>Meta rolls out specialized Muse and Meta AI models</b></p>
<p>The rollout brings Active Memory and Active Focus, with company-reported memory evaluation results.</p>
<ul>
<li><a href="https://research.meta.ai/blog/specialized-models-for-muse-and-metaai">Bringing Specialized Models to Muse and Meta AI</a> — Meta AI Research</li>
</ul>
<p><b>Amazon begins invited access to shopping by text</b></p>
<p>Alexa for Shopping adds text access for selected US customers, with broader access planned.</p>
<ul>
<li><a href="https://www.aboutamazon.com/news/company-news/amazon-news-today-top-stories-company">What you need to know about Amazon today: November 6, 2026</a> — Amazon</li>
<li><a href="https://www.aboutamazon.com/news/retail/alexa-for-shopping-text-to-shop">Text, shop, done: Alexa for Shopping is now just a text away</a> — Amazon</li>
</ul>
<p><b>Check Point reports ARTEX cybercrime campaigns</b></p>
<p>Operator databases reveal agent-driven compromises alongside failed cryptocurrency theft objectives.</p>
<ul>
<li><a href="https://research.checkpoint.com/2026/hacking-on-autopilot-inside-artex-cybercrime-campaigns/">Hacking on Autopilot: Inside ARTEX Cybercrime Campaigns</a> — Check Point Research</li>
<li><a href="https://blog.checkpoint.com/research/inside-artex-how-autonomous-ai-agents-are-being-used-in-real-world-cyberattacks">Inside ARTEX: How Autonomous AI Agents Are Being Used in Real-World Cyberattacks</a> — Check Point Research</li>
</ul>
<p><b>Kalepa announces $75 million Series B</b></p>
<p>The insurance AI company announces funding at a $470 million valuation.</p>
<ul>
<li><a href="https://kalepa.com/news-posts/kalepa-raises-75-million-series-b">Kalepa Raises $75 Million Series B at a $470 Million Valuation to Expand Professional Grade AI Across Insurance</a> — Kalepa</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Mon, 09 Nov 2026 07:04:53 +0100</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/b202ed40/5a302d0a.mp3" length="3342462" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>209</itunes:duration>
      <itunes:summary>New models from Thinking Machines and Meta, Amazon shopping by text, Check Point's ARTEX findings and Kalepa's funding round.</itunes:summary>
      <itunes:subtitle>New models from Thinking Machines and Meta, Amazon shopping by text, Check Point's ARTEX findings and Kalepa's funding round.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/b202ed40/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>GPT-6 Cyber, Gemini video updates and AI credential theft</title>
      <itunes:title>GPT-6 Cyber, Gemini video updates and AI credential theft</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">43f0a282-fdd0-4020-9b0c-b27f8b4813a5</guid>
      <link>https://share.transistor.fm/s/ddf8446c</link>
      <description>
        <![CDATA[<p>Restricted access to GPT-6 Cyber, Google’s video and chemistry tools, a patched ChatGPT bypass, Microsoft’s findings on AI-assisted credential theft, and Anthropic infrastructure financing.</p>
<p><b>Stories and sources</b></p>
<p><b>OpenAI details restricted GPT-6 Cyber access</b></p>
<p>OpenAI describes a cyber-specialized model, restricted access and additional safeguards.</p>
<ul>
<li><a href="https://deploymentsafety.openai.com/gpt-6-cyber">GPT-6 Cyber System Card</a> — OpenAI</li>
</ul>
<p><b>Google releases Gemini Omni 1.2 Flash</b></p>
<p>Google announces longer video generation, conversational editing and reduced pricing.</p>
<ul>
<li><a href="https://blog.google/innovation-and-ai/technology/developers-tools/introducing-gemini-omni-1-2-flash/?utm_campaign=gdm&amp;utm_content=&amp;utm_medium=referral&amp;utm_source=deepmind.google">Gemini Omni 1.2 Flash: Fast and cost-effective with longer clips</a> — Google</li>
</ul>
<p><b>DeepMind introduces Antigravity chemistry skills</b></p>
<p>A chemistry workflow demonstration proposes molecules requiring experimental validation.</p>
<ul>
<li><a href="https://deepmind.google/blog/agentic-cheminformatics-with-google-antigravity/">Agentic cheminformatics with Google Antigravity</a> — Google DeepMind</li>
</ul>
<p><b>Check Point discloses patched ChatGPT Lockdown Mode bypass</b></p>
<p>Researchers demonstrated chat-data extraction and deceptive update prompts through an application sandbox weakness.</p>
<ul>
<li><a href="https://research.checkpoint.com/2026/chatgpt-lockdown-mode-bypass-phishing-malware-delivery-and-silent-data-exfiltration/">ChatGPT Lockdown Mode Bypass: Phishing, Malware Delivery, and Silent Data Exfiltration</a> — Check Point Research</li>
<li><a href="https://blog.checkpoint.com/security/chatgpt-lockdown-mode-bypass-reveals-a-wider-challenge-in-ai-security/">ChatGPT Lockdown Mode Bypass Reveals a Wider Challenge in AI Security</a> — Check Point</li>
</ul>
<p><b>Microsoft reports agentic AI credential theft</b></p>
<p>Microsoft documents AI agents finding exposed secrets and conducting follow-on cloud compromise.</p>
<ul>
<li><a href="https://www.microsoft.com/en-us/security/blog/2026/11/05/threat-actors-use-agentic-ai-tooling-to-find-and-abuse-exposed-secrets/">Threat actors use agentic AI tooling to find and abuse exposed secrets</a> — Microsoft</li>
</ul>
<p><b>TeraWulf announces Google backing for Anthropic infrastructure</b></p>
<p>New credit support and construction agreements concern planned Kentucky data centers leased to Anthropic.</p>
<ul>
<li><a href="https://www.globenewswire.com/news-release/2026/11/04/3395277/0/en/terawulf-secures-google-credit-backstop-for-project-financing-in-support-of-19-billion-anthropic-lease-executes-definitive-fluor-epc-agreement.html">TeraWulf Secures Google Credit Backstop for Project Financing in Support of $19 Billion Anthropic Lease; Executes Definitive Fluor EPC Agreement</a> — TeraWulf via GlobeNewswire</li>
<li><a href="https://investors.terawulf.com/sec-filings/all-sec-filings/content/0001628280-26-069810/0001628280-26-069810.pdf">TeraWulf Inc. Form 8-K — November 4, 2026</a> — TeraWulf</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>Restricted access to GPT-6 Cyber, Google’s video and chemistry tools, a patched ChatGPT bypass, Microsoft’s findings on AI-assisted credential theft, and Anthropic infrastructure financing.</p>
<p><b>Stories and sources</b></p>
<p><b>OpenAI details restricted GPT-6 Cyber access</b></p>
<p>OpenAI describes a cyber-specialized model, restricted access and additional safeguards.</p>
<ul>
<li><a href="https://deploymentsafety.openai.com/gpt-6-cyber">GPT-6 Cyber System Card</a> — OpenAI</li>
</ul>
<p><b>Google releases Gemini Omni 1.2 Flash</b></p>
<p>Google announces longer video generation, conversational editing and reduced pricing.</p>
<ul>
<li><a href="https://blog.google/innovation-and-ai/technology/developers-tools/introducing-gemini-omni-1-2-flash/?utm_campaign=gdm&amp;utm_content=&amp;utm_medium=referral&amp;utm_source=deepmind.google">Gemini Omni 1.2 Flash: Fast and cost-effective with longer clips</a> — Google</li>
</ul>
<p><b>DeepMind introduces Antigravity chemistry skills</b></p>
<p>A chemistry workflow demonstration proposes molecules requiring experimental validation.</p>
<ul>
<li><a href="https://deepmind.google/blog/agentic-cheminformatics-with-google-antigravity/">Agentic cheminformatics with Google Antigravity</a> — Google DeepMind</li>
</ul>
<p><b>Check Point discloses patched ChatGPT Lockdown Mode bypass</b></p>
<p>Researchers demonstrated chat-data extraction and deceptive update prompts through an application sandbox weakness.</p>
<ul>
<li><a href="https://research.checkpoint.com/2026/chatgpt-lockdown-mode-bypass-phishing-malware-delivery-and-silent-data-exfiltration/">ChatGPT Lockdown Mode Bypass: Phishing, Malware Delivery, and Silent Data Exfiltration</a> — Check Point Research</li>
<li><a href="https://blog.checkpoint.com/security/chatgpt-lockdown-mode-bypass-reveals-a-wider-challenge-in-ai-security/">ChatGPT Lockdown Mode Bypass Reveals a Wider Challenge in AI Security</a> — Check Point</li>
</ul>
<p><b>Microsoft reports agentic AI credential theft</b></p>
<p>Microsoft documents AI agents finding exposed secrets and conducting follow-on cloud compromise.</p>
<ul>
<li><a href="https://www.microsoft.com/en-us/security/blog/2026/11/05/threat-actors-use-agentic-ai-tooling-to-find-and-abuse-exposed-secrets/">Threat actors use agentic AI tooling to find and abuse exposed secrets</a> — Microsoft</li>
</ul>
<p><b>TeraWulf announces Google backing for Anthropic infrastructure</b></p>
<p>New credit support and construction agreements concern planned Kentucky data centers leased to Anthropic.</p>
<ul>
<li><a href="https://www.globenewswire.com/news-release/2026/11/04/3395277/0/en/terawulf-secures-google-credit-backstop-for-project-financing-in-support-of-19-billion-anthropic-lease-executes-definitive-fluor-epc-agreement.html">TeraWulf Secures Google Credit Backstop for Project Financing in Support of $19 Billion Anthropic Lease; Executes Definitive Fluor EPC Agreement</a> — TeraWulf via GlobeNewswire</li>
<li><a href="https://investors.terawulf.com/sec-filings/all-sec-filings/content/0001628280-26-069810/0001628280-26-069810.pdf">TeraWulf Inc. Form 8-K — November 4, 2026</a> — TeraWulf</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Sat, 07 Nov 2026 07:05:34 +0100</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/ddf8446c/ca35e95e.mp3" length="4059680" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>254</itunes:duration>
      <itunes:summary>Restricted access to GPT-6 Cyber, Google’s video and chemistry tools, a patched ChatGPT bypass, Microsoft’s findings on AI-assisted credential theft, and Anthropic infrastructure financing.</itunes:summary>
      <itunes:subtitle>Restricted access to GPT-6 Cyber, Google’s video and chemistry tools, a patched ChatGPT bypass, Microsoft’s findings on AI-assisted credential theft, and Anthropic infrastructure financing.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/ddf8446c/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>Classroom AI, agent benchmarks and security disclosures</title>
      <itunes:title>Classroom AI, agent benchmarks and security disclosures</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">41a448ff-fce7-4561-aba0-d631a9671fd0</guid>
      <link>https://share.transistor.fm/s/99105963</link>
      <description>
        <![CDATA[<p>Google’s Guided Learning rollout, SemiAnalysis’s GDPx benchmark, two AI-related intrusion reports, and CrowdStrike’s proposed Fiddler acquisition.</p>
<p><b>Stories and sources</b></p>
<p><b>Google rolls out Guided Learning in Classroom</b></p>
<p>Google announced a global classroom rollout and reported results from a single-session university trial.</p>
<ul>
<li><a href="https://blog.google/products-and-platforms/products/education/ai-learning-era-2026/">AI in the learning era</a> — Google</li>
<li><a href="https://blog.google/products-and-platforms/products/education/guided-learning-uk-japan/">New research shows how Guided Learning supports students</a> — Google</li>
</ul>
<p><b>SemiAnalysis launches GDPx agent benchmark</b></p>
<p>GDPx evaluates agents across 206 tasks, with reported scores affected by task conditions and safety classifiers.</p>
<ul>
<li><a href="https://newsletter.semianalysis.com/p/gdpx-measuring-frontier-model-capabilities">GDPx: Measuring Frontier Model Capabilities Across the Most Valuable Domains</a> — SemiAnalysis</li>
<li><a href="https://gdpx.semianalysis.com/">GDPx</a> — SemiAnalysis</li>
</ul>
<p><b>Kaspersky documents CyberStrikeAI intrusions</b></p>
<p>Kaspersky describes model-directed post-compromise activity while retaining uncertainty about operator identity.</p>
<ul>
<li><a href="https://securelist.com/tr/cyberstrikeai-attacks/121715/">CyberStrikeAI in the wild: the use of AI in real-world attacks on organizations</a> — Kaspersky Securelist</li>
</ul>
<p><b>WorkOS discloses AI-assisted intrusion</b></p>
<p>WorkOS reports sales-system data access and an ongoing forensic investigation.</p>
<ul>
<li><a href="https://workos.com/security/bulletins/october-30th-2026-data-security-incident">October 30th, 2026 Data Security Incident</a> — WorkOS</li>
</ul>
<p><b>CrowdStrike agrees to acquire Fiddler</b></p>
<p>The pending acquisition would add AI evaluation and monitoring technology to Falcon.</p>
<ul>
<li><a href="https://ir.crowdstrike.com/node/17966/pdf">CrowdStrike to Acquire Fiddler to Deliver Agent Observability for AI Trust</a> — CrowdStrike</li>
<li><a href="https://www.fiddler.ai/blog/fiddler-joins-crowdstrike">Fiddler to Join CrowdStrike: Building the Future of Agentic Security and Safety</a> — Fiddler AI</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>Google’s Guided Learning rollout, SemiAnalysis’s GDPx benchmark, two AI-related intrusion reports, and CrowdStrike’s proposed Fiddler acquisition.</p>
<p><b>Stories and sources</b></p>
<p><b>Google rolls out Guided Learning in Classroom</b></p>
<p>Google announced a global classroom rollout and reported results from a single-session university trial.</p>
<ul>
<li><a href="https://blog.google/products-and-platforms/products/education/ai-learning-era-2026/">AI in the learning era</a> — Google</li>
<li><a href="https://blog.google/products-and-platforms/products/education/guided-learning-uk-japan/">New research shows how Guided Learning supports students</a> — Google</li>
</ul>
<p><b>SemiAnalysis launches GDPx agent benchmark</b></p>
<p>GDPx evaluates agents across 206 tasks, with reported scores affected by task conditions and safety classifiers.</p>
<ul>
<li><a href="https://newsletter.semianalysis.com/p/gdpx-measuring-frontier-model-capabilities">GDPx: Measuring Frontier Model Capabilities Across the Most Valuable Domains</a> — SemiAnalysis</li>
<li><a href="https://gdpx.semianalysis.com/">GDPx</a> — SemiAnalysis</li>
</ul>
<p><b>Kaspersky documents CyberStrikeAI intrusions</b></p>
<p>Kaspersky describes model-directed post-compromise activity while retaining uncertainty about operator identity.</p>
<ul>
<li><a href="https://securelist.com/tr/cyberstrikeai-attacks/121715/">CyberStrikeAI in the wild: the use of AI in real-world attacks on organizations</a> — Kaspersky Securelist</li>
</ul>
<p><b>WorkOS discloses AI-assisted intrusion</b></p>
<p>WorkOS reports sales-system data access and an ongoing forensic investigation.</p>
<ul>
<li><a href="https://workos.com/security/bulletins/october-30th-2026-data-security-incident">October 30th, 2026 Data Security Incident</a> — WorkOS</li>
</ul>
<p><b>CrowdStrike agrees to acquire Fiddler</b></p>
<p>The pending acquisition would add AI evaluation and monitoring technology to Falcon.</p>
<ul>
<li><a href="https://ir.crowdstrike.com/node/17966/pdf">CrowdStrike to Acquire Fiddler to Deliver Agent Observability for AI Trust</a> — CrowdStrike</li>
<li><a href="https://www.fiddler.ai/blog/fiddler-joins-crowdstrike">Fiddler to Join CrowdStrike: Building the Future of Agentic Security and Safety</a> — Fiddler AI</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Wed, 04 Nov 2026 07:05:23 +0100</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/99105963/8801d0af.mp3" length="3897512" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>244</itunes:duration>
      <itunes:summary>Google’s Guided Learning rollout, SemiAnalysis’s GDPx benchmark, two AI-related intrusion reports, and CrowdStrike’s proposed Fiddler acquisition.</itunes:summary>
      <itunes:subtitle>Google’s Guided Learning rollout, SemiAnalysis’s GDPx benchmark, two AI-related intrusion reports, and CrowdStrike’s proposed Fiddler acquisition.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/99105963/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>DeepSeek’s agent prototype and AI credential theft</title>
      <itunes:title>DeepSeek’s agent prototype and AI credential theft</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">85435270-2f3a-4e51-96fb-310a850a6bc5</guid>
      <link>https://share.transistor.fm/s/de1acd35</link>
      <description>
        <![CDATA[<p>DeepSeek’s coding-agent prototype, Rosalind Research’s restricted rollout, reports of stolen AI credentials, and infrastructure deals involving Marvell and NetApp.</p>
<p><b>Stories and sources</b></p>
<p><b>DeepSeek publishes Superagent research prototype</b></p>
<p>The repository-scoped coding prototype includes explicit authorisation rules and documented execution limitations.</p>
<ul>
<li><a href="https://github.com/deepseek-ai/dsh-superagent-prototype/commits">Commits · deepseek-ai/dsh-superagent-prototype</a> — DeepSeek Harness Team / GitHub</li>
<li><a href="https://github.com/deepseek-ai/dsh-superagent-prototype">dsh-superagent</a> — DeepSeek Harness Team / GitHub</li>
<li><a href="https://github.com/deepseek-ai/dsh-superagent-prototype/blob/main/docs/002-authorization.md">002 Authorization</a> — DeepSeek Harness Team / GitHub</li>
<li><a href="https://github.com/deepseek-ai/dsh-superagent-prototype/blob/main/docs/003-execution.md">003 Execution</a> — DeepSeek Harness Team / GitHub</li>
</ul>
<p><b>OpenAI begins restricted Rosalind Research rollout</b></p>
<p>GPT-6 Rosalind Research has approval requirements and enterprise oversight conditions.</p>
<ul>
<li><a href="https://help.openai.com/en/articles/9624314-model-release-notes">Model Release Notes</a> — OpenAI</li>
<li><a href="https://deploymentsafety.openai.com/gpt-rosalind-6">GPT-Rosalind-6 System Card Addendum</a> — OpenAI</li>
</ul>
<p><b>Wiz documents AI credential abuse as Northflank updates investigation</b></p>
<p>Observed model-call abuse accompanies unresolved attribution and an ongoing platform investigation.</p>
<ul>
<li><a href="https://www.wiz.io/blog/black-october-for-ai-developer-platforms">“Black October” for AI Developer Platforms</a> — Wiz</li>
<li><a href="https://status.northflank.com/">Northflank - Status Page</a> — Northflank</li>
</ul>
<p><b>Marvell agrees to acquire Lumentum</b></p>
<p>The proposed all-stock transaction combines electrical and optical technologies for AI infrastructure.</p>
<ul>
<li><a href="https://investor.marvell.com/_assets/_b3ade7b7a9c0c378f1c7eb0eff8c24c3/marvell/news/2026-11-02_Marvell_and_Lumentum_to_Combine_Creating_a_1040.pdf">Marvell and Lumentum to Combine, Creating a Category Leader in AI Connectivity</a> — Marvell and Lumentum</li>
<li><a href="https://investor.marvell.com/sec-filings/all-sec-filings/content/0001193125-26-443841/0001193125-26-443841.pdf">Marvell Customer Letter — Marvell and Lumentum</a> — Marvell</li>
</ul>
<p><b>NetApp completes PEAK:AIO acquisition</b></p>
<p>NetApp describes planned AI-storage integration following the completed purchase.</p>
<ul>
<li><a href="https://investors.netapp.com/news/news-details/2026/NetApp-Completes-Acquisition-of-PEAKAIO-to-Accelerate-Zettabyte-scale-AI-Deployments/default.aspx">NetApp Completes Acquisition of PEAK:AIO to Accelerate Zettabyte-scale AI Deployments</a> — NetApp</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>DeepSeek’s coding-agent prototype, Rosalind Research’s restricted rollout, reports of stolen AI credentials, and infrastructure deals involving Marvell and NetApp.</p>
<p><b>Stories and sources</b></p>
<p><b>DeepSeek publishes Superagent research prototype</b></p>
<p>The repository-scoped coding prototype includes explicit authorisation rules and documented execution limitations.</p>
<ul>
<li><a href="https://github.com/deepseek-ai/dsh-superagent-prototype/commits">Commits · deepseek-ai/dsh-superagent-prototype</a> — DeepSeek Harness Team / GitHub</li>
<li><a href="https://github.com/deepseek-ai/dsh-superagent-prototype">dsh-superagent</a> — DeepSeek Harness Team / GitHub</li>
<li><a href="https://github.com/deepseek-ai/dsh-superagent-prototype/blob/main/docs/002-authorization.md">002 Authorization</a> — DeepSeek Harness Team / GitHub</li>
<li><a href="https://github.com/deepseek-ai/dsh-superagent-prototype/blob/main/docs/003-execution.md">003 Execution</a> — DeepSeek Harness Team / GitHub</li>
</ul>
<p><b>OpenAI begins restricted Rosalind Research rollout</b></p>
<p>GPT-6 Rosalind Research has approval requirements and enterprise oversight conditions.</p>
<ul>
<li><a href="https://help.openai.com/en/articles/9624314-model-release-notes">Model Release Notes</a> — OpenAI</li>
<li><a href="https://deploymentsafety.openai.com/gpt-rosalind-6">GPT-Rosalind-6 System Card Addendum</a> — OpenAI</li>
</ul>
<p><b>Wiz documents AI credential abuse as Northflank updates investigation</b></p>
<p>Observed model-call abuse accompanies unresolved attribution and an ongoing platform investigation.</p>
<ul>
<li><a href="https://www.wiz.io/blog/black-october-for-ai-developer-platforms">“Black October” for AI Developer Platforms</a> — Wiz</li>
<li><a href="https://status.northflank.com/">Northflank - Status Page</a> — Northflank</li>
</ul>
<p><b>Marvell agrees to acquire Lumentum</b></p>
<p>The proposed all-stock transaction combines electrical and optical technologies for AI infrastructure.</p>
<ul>
<li><a href="https://investor.marvell.com/_assets/_b3ade7b7a9c0c378f1c7eb0eff8c24c3/marvell/news/2026-11-02_Marvell_and_Lumentum_to_Combine_Creating_a_1040.pdf">Marvell and Lumentum to Combine, Creating a Category Leader in AI Connectivity</a> — Marvell and Lumentum</li>
<li><a href="https://investor.marvell.com/sec-filings/all-sec-filings/content/0001193125-26-443841/0001193125-26-443841.pdf">Marvell Customer Letter — Marvell and Lumentum</a> — Marvell</li>
</ul>
<p><b>NetApp completes PEAK:AIO acquisition</b></p>
<p>NetApp describes planned AI-storage integration following the completed purchase.</p>
<ul>
<li><a href="https://investors.netapp.com/news/news-details/2026/NetApp-Completes-Acquisition-of-PEAKAIO-to-Accelerate-Zettabyte-scale-AI-Deployments/default.aspx">NetApp Completes Acquisition of PEAK:AIO to Accelerate Zettabyte-scale AI Deployments</a> — NetApp</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Tue, 03 Nov 2026 07:06:45 +0100</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/de1acd35/6915162c.mp3" length="4611804" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>289</itunes:duration>
      <itunes:summary>DeepSeek’s coding-agent prototype, Rosalind Research’s restricted rollout, reports of stolen AI credentials, and infrastructure deals involving Marvell and NetApp.</itunes:summary>
      <itunes:subtitle>DeepSeek’s coding-agent prototype, Rosalind Research’s restricted rollout, reports of stolen AI credentials, and infrastructure deals involving Marvell and NetApp.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/de1acd35/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>Clover preview, Gemini computer use and Cohere investment</title>
      <itunes:title>Clover preview, Gemini computer use and Cohere investment</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">2d100212-f317-4e89-bca4-5a784d5c09be</guid>
      <link>https://share.transistor.fm/s/47479276</link>
      <description>
        <![CDATA[<p>DiffuSpace's model preview, Google's computer-use release and planning pilot, Talos's malware analysis, and Canada's planned Cohere investment.</p>
<p><b>Stories and sources</b></p>
<p><b>DiffuSpace releases Clover preview</b></p>
<p>DiffuSpace released a diffusion language model preview with downloadable weights.</p>
<ul>
<li><a href="https://huggingface.co/blog/Diffuspace/clover-35b-a3b-preview">Clover-35B-A3B-Preview: A New Way for Intelligence</a> — DiffuSpace</li>
</ul>
<p><b>Gemini Computer Use reaches general availability</b></p>
<p>Google released Computer Use across two APIs, with client-side execution and configurable safeguards.</p>
<ul>
<li><a href="https://ai.google.dev/gemini-api/docs/changelog">Release notes | Gemini API</a> — Google</li>
<li><a href="https://ai.google.dev/gemini-api/docs/computer-use">Computer use | Gemini API</a> — Google</li>
</ul>
<p><b>PlanAssist enters expanded beta</b></p>
<p>DeepMind announced an expanded council pilot and reported assessment-time results.</p>
<ul>
<li><a href="https://deepmind.google/blog/speeding-up-planning-decisions-with-ai/">Speeding up planning decisions with AI</a> — Google DeepMind</li>
</ul>
<p><b>Talos documents AI profiling in VOLTA malware</b></p>
<p>Talos analyzed an optional post-theft AI profiling capability.</p>
<ul>
<li><a href="https://blog.talosintelligence.com/appraisal-by-ai-model-driven-victim-prioritization-in-the-volta-infostealer/">Appraisal by AI: Model-driven victim prioritization in the VOLTA infostealer</a> — Cisco Talos</li>
</ul>
<p><b>Canada announces planned Cohere equity investment</b></p>
<p>Canada committed to staged investment tied to company commitments.</p>
<ul>
<li><a href="https://www.pm.gc.ca/en/news/news-releases/2026/10/29/canada-announces-landmark-investment-cohere-scale-secure-sovereign-ai">Canada announces landmark investment in Cohere to scale secure, sovereign AI</a> — Prime Minister of Canada</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>DiffuSpace's model preview, Google's computer-use release and planning pilot, Talos's malware analysis, and Canada's planned Cohere investment.</p>
<p><b>Stories and sources</b></p>
<p><b>DiffuSpace releases Clover preview</b></p>
<p>DiffuSpace released a diffusion language model preview with downloadable weights.</p>
<ul>
<li><a href="https://huggingface.co/blog/Diffuspace/clover-35b-a3b-preview">Clover-35B-A3B-Preview: A New Way for Intelligence</a> — DiffuSpace</li>
</ul>
<p><b>Gemini Computer Use reaches general availability</b></p>
<p>Google released Computer Use across two APIs, with client-side execution and configurable safeguards.</p>
<ul>
<li><a href="https://ai.google.dev/gemini-api/docs/changelog">Release notes | Gemini API</a> — Google</li>
<li><a href="https://ai.google.dev/gemini-api/docs/computer-use">Computer use | Gemini API</a> — Google</li>
</ul>
<p><b>PlanAssist enters expanded beta</b></p>
<p>DeepMind announced an expanded council pilot and reported assessment-time results.</p>
<ul>
<li><a href="https://deepmind.google/blog/speeding-up-planning-decisions-with-ai/">Speeding up planning decisions with AI</a> — Google DeepMind</li>
</ul>
<p><b>Talos documents AI profiling in VOLTA malware</b></p>
<p>Talos analyzed an optional post-theft AI profiling capability.</p>
<ul>
<li><a href="https://blog.talosintelligence.com/appraisal-by-ai-model-driven-victim-prioritization-in-the-volta-infostealer/">Appraisal by AI: Model-driven victim prioritization in the VOLTA infostealer</a> — Cisco Talos</li>
</ul>
<p><b>Canada announces planned Cohere equity investment</b></p>
<p>Canada committed to staged investment tied to company commitments.</p>
<ul>
<li><a href="https://www.pm.gc.ca/en/news/news-releases/2026/10/29/canada-announces-landmark-investment-cohere-scale-secure-sovereign-ai">Canada announces landmark investment in Cohere to scale secure, sovereign AI</a> — Prime Minister of Canada</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Mon, 02 Nov 2026 13:14:10 +0100</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/47479276/65025694.mp3" length="3147693" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>197</itunes:duration>
      <itunes:summary>DiffuSpace's model preview, Google's computer-use release and planning pilot, Talos's malware analysis, and Canada's planned Cohere investment.</itunes:summary>
      <itunes:subtitle>DiffuSpace's model preview, Google's computer-use release and planning pilot, Talos's malware analysis, and Canada's planned Cohere investment.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/47479276/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>Byte-level models, AI security research and Blackford acquisition</title>
      <itunes:title>Byte-level models, AI security research and Blackford acquisition</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">472743f6-9b3e-4b9d-b4c2-a41a058edb01</guid>
      <link>https://share.transistor.fm/s/53954dea</link>
      <description>
        <![CDATA[<p>Ai2 releases new model checkpoints, Microsoft reports vulnerability findings, Lumen documents attacks on AI infrastructure, and Sopris acquires Blackford Analysis.</p>
<p><b>Stories and sources</b></p>
<p><b>Ai2 releases additional byte-level model checkpoints</b></p>
<p>New checkpoints extend Bolmo’s conversion approach to Qwen and Llama models.</p>
<ul>
<li><a href="https://allenai.org/blog/bolmo-nature">Now in Nature: Retrofitting language models to operate over bytes</a> — Ai2</li>
</ul>
<p><b>Microsoft reports AI-assisted vulnerability submissions</b></p>
<p>FORGE’s submission counts distinguish internal validation from maintainer acceptance and remediation.</p>
<ul>
<li><a href="https://www.microsoft.com/en-us/security/blog/2026/10/07/3-lessons-from-frontier-ai-vulnerability-research/">3 lessons from frontier AI vulnerability research</a> — Microsoft</li>
</ul>
<p><b>Lumen reports PoeLLM compromises of exposed AI services</b></p>
<p>Researchers document mining and further exploitation from compromised servers, with qualified actor attribution.</p>
<ul>
<li><a href="https://www.lumen.com/blog/en-us/canto-incognito-tracking-the-poellm-malware">Canto incognito: tracking the PoeLLM malware</a> — Lumen — Black Lotus Labs</li>
<li><a href="https://www.lumen.com/blog-and-news/topics/network-security">Network Security</a> — Lumen</li>
</ul>
<p><b>Sopris completes Blackford Analysis acquisition</b></p>
<p>The purchase from Bayer is complete; a combination under Azra AI remains pending.</p>
<ul>
<li><a href="https://www.prnewswire.com/news-releases/sopris-capital-acquires-blackford-analysis-strengthening-ai-orchestration-capabilities-across-its-healthtech-portfolio-302903112.html">Sopris Capital Acquires Blackford Analysis, Strengthening AI Orchestration Capabilities Across Its HealthTech Portfolio</a> — Sopris Capital via PR Newswire</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>Ai2 releases new model checkpoints, Microsoft reports vulnerability findings, Lumen documents attacks on AI infrastructure, and Sopris acquires Blackford Analysis.</p>
<p><b>Stories and sources</b></p>
<p><b>Ai2 releases additional byte-level model checkpoints</b></p>
<p>New checkpoints extend Bolmo’s conversion approach to Qwen and Llama models.</p>
<ul>
<li><a href="https://allenai.org/blog/bolmo-nature">Now in Nature: Retrofitting language models to operate over bytes</a> — Ai2</li>
</ul>
<p><b>Microsoft reports AI-assisted vulnerability submissions</b></p>
<p>FORGE’s submission counts distinguish internal validation from maintainer acceptance and remediation.</p>
<ul>
<li><a href="https://www.microsoft.com/en-us/security/blog/2026/10/07/3-lessons-from-frontier-ai-vulnerability-research/">3 lessons from frontier AI vulnerability research</a> — Microsoft</li>
</ul>
<p><b>Lumen reports PoeLLM compromises of exposed AI services</b></p>
<p>Researchers document mining and further exploitation from compromised servers, with qualified actor attribution.</p>
<ul>
<li><a href="https://www.lumen.com/blog/en-us/canto-incognito-tracking-the-poellm-malware">Canto incognito: tracking the PoeLLM malware</a> — Lumen — Black Lotus Labs</li>
<li><a href="https://www.lumen.com/blog-and-news/topics/network-security">Network Security</a> — Lumen</li>
</ul>
<p><b>Sopris completes Blackford Analysis acquisition</b></p>
<p>The purchase from Bayer is complete; a combination under Azra AI remains pending.</p>
<ul>
<li><a href="https://www.prnewswire.com/news-releases/sopris-capital-acquires-blackford-analysis-strengthening-ai-orchestration-capabilities-across-its-healthtech-portfolio-302903112.html">Sopris Capital Acquires Blackford Analysis, Strengthening AI Orchestration Capabilities Across Its HealthTech Portfolio</a> — Sopris Capital via PR Newswire</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Sun, 11 Oct 2026 07:05:30 +0200</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/53954dea/7b8746a8.mp3" length="2227765" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>140</itunes:duration>
      <itunes:summary>Ai2 releases new model checkpoints, Microsoft reports vulnerability findings, Lumen documents attacks on AI infrastructure, and Sopris acquires Blackford Analysis.</itunes:summary>
      <itunes:subtitle>Ai2 releases new model checkpoints, Microsoft reports vulnerability findings, Lumen documents attacks on AI infrastructure, and Sopris acquires Blackford Analysis.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/53954dea/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>AI Week in Review — Model releases, local agents and security disclosures</title>
      <itunes:title>AI Week in Review — Model releases, local agents and security disclosures</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">dc2e2309-526e-44be-8af7-c1bb91cb0007</guid>
      <link>https://share.transistor.fm/s/b83df581</link>
      <description>
        <![CDATA[<p>GPT-6, Haiku 5.5 and Mistral Large 4 lead a review of multimodal models, local coding tools, public-health research, documented AI incidents and business announcements.</p>
<p><b>Stories and sources</b></p>
<p><b>OpenAI, Anthropic and Mistral announce model releases with distinct access conditions</b></p>
<p>GPT-6 adds interactive ChatGPT responses; Haiku 5.5 introduces adjustable effort and new pricing; Mistral Large 4 enters public API preview with weights forthcoming.</p>
<ul>
<li><a href="https://openai.com/index/gpt-6-for-everyone/">GPT-6 and Intelligent UI for everyone</a> — OpenAI</li>
<li><a href="https://help.openai.com/en/articles/6825453-chatgpt-release-notes">ChatGPT release notes</a> — OpenAI</li>
<li><a href="https://help.openai.com/en/articles/10128477-chatgpt-enterprise-and-edu-release-notes">ChatGPT Enterprise and Edu release notes</a> — OpenAI</li>
<li><a href="https://www.anthropic.com/claude-haiku-5-5">Introducing Claude Haiku 5.5</a> — Anthropic</li>
<li><a href="https://www.anthropic.com/claude/haiku?via=free">Claude Haiku</a> — Anthropic</li>
<li><a href="https://github.blog/changelog/2026-10-07-claude-haiku-5-5-in-github-copilot/">Claude Haiku 5.5 in GitHub Copilot</a> — GitHub</li>
<li><a href="https://support.claude.com/en/articles/17154008-monthly-api-credits-for-max-and-team-plans">Monthly API credits for Max and Team plans</a> — Anthropic</li>
<li><a href="https://platform.claude.com/docs/en/agents-and-tools/tool-use/browser-use-sdk">Browser and computer use with the SDK toolsets</a> — Anthropic</li>
<li><a href="https://mistral.ai/news/mistral-large-4/">Introducing Mistral Large 4</a> — Mistral AI</li>
<li><a href="https://docs.mistral.ai/models/mistral-large">Mistral Large 4</a> — Mistral AI</li>
<li><a href="https://docs.mistral.ai/resources/changelogs">Changelog</a> — Mistral AI</li>
</ul>
<p><b>Liquid AI releases hosted and downloadable multimodal decision models</b></p>
<p>The d1 releases return structured decisions, with different input support, qualified latency measurements and separate hosted billing conditions.</p>
<ul>
<li><a href="https://www.liquid.ai/blog/d1-decision-model">Introducing d1: The most capable decision model, now with vision</a> — Liquid AI</li>
<li><a href="https://www.liquid.ai/blog/d1-open">Open d1: Edge decision models for text, vision, and audio</a> — Liquid AI</li>
<li><a href="https://huggingface.co/LiquidAI/d1-3B">LiquidAI/d1-3B</a> — Liquid AI on Hugging Face</li>
<li><a href="https://huggingface.co/LiquidAI/d1-omni-600M">LiquidAI/d1-omni-600M</a> — Liquid AI on Hugging Face</li>
</ul>
<p><b>Google releases EmbeddingGemma 2 and local retrieval demonstrations</b></p>
<p>The multimodal embedding model has modular encoders, downloadable weights and configuration-dependent results; ML Kit integration remains forthcoming.</p>
<ul>
<li><a href="https://blog.google/innovation-and-ai/technology/developers-tools/embeddinggemma-2/">EmbeddingGemma 2: an open, lightweight multimodal embedding model</a> — Google</li>
<li><a href="https://developers.googleblog.com/embeddinggemma-2-the-developer-guide/">EmbeddingGemma 2: The Developer Guide</a> — Google</li>
<li><a href="https://developers.googleblog.com/google-ai-edge-with-embeddinggemma-2/">Bring multimodal semantic search to the edge with EmbeddingGemma 2</a> — Google</li>
<li><a href="https://ai.google.dev/gemma/docs/embeddinggemma/model_card_2">EmbeddingGemma 2 model card</a> — Google</li>
</ul>
<p><b>Microsoft plans local coding preview as Copilot sandboxing becomes available</b></p>
<p>Local model routing remains an experimental preview planned for later October; released containment controls have explicit boundaries.</p>
<ul>
<li><a href="https://blogs.windows.com/windowsexperience/2026/10/07/building-windows-for-hybrid-intelligence/">Building Windows for hybrid intelligence</a> — Microsoft</li>
<li><a href="https://commandline.microsoft.com/local-models-sandboxed-tools-github-windows/">Bringing local models and sandboxed tools to Windows and GitHub Copilot</a> — Microsoft</li>
<li><a href="https://github.blog/changelog/2026-10-07-local-sandboxing-for-github-copilot-now-generally-available/">Local sandboxing for GitHub Copilot now generally available</a> — GitHub</li>
<li><a href="https://blogs.windows.com/windowsdeveloper/2026/10/07/microsoft-execution-containers-policy-driven-containment-for-ai-agents/">Microsoft Execution Containers: Policy-driven containment for AI agents</a> — Microsoft</li>
</ul>
<p><b>Google reports five geographic AI public-health studies</b></p>
<p>The studies report task-specific prediction gains alongside forecasting and demographic limitations.</p>
<ul>
<li><a href="https://research.google/blog/earth-ais-planetary-geospatial-foundation-models-for-global-public-health/">Unlocking Earth AI’s planetary geospatial foundation models for global public health</a> — Google Research</li>
<li><a href="https://arxiv.org/abs/2610.05699">Planetary Geospatial Foundation Models: A New Paradigm for Global Public Health</a> — arXiv</li>
<li><a href="https://arxiv.org/html/2610.05699v1">Planetary Geospatial Foundation Models: A New Paradigm for Global Public Health</a> — arXiv</li>
</ul>
<p><b>Anthropic discloses unintended website actions and restricts evaluation internet access</b></p>
<p>The disclosure includes an invented Philadelphia homicide tip that police say remained in spam, plus other unintended interactions with outside systems.</p>
<ul>
<li><a href="https://www.anthropic.com/research/investigating-unintended-model-actions">Investigating unintended model actions in our evaluations and internal use</a> — Anthropic</li>
<li><a href="https://6abc.com/post/anthropic-ai-model-submitted-false-tip-unsolved-murder-philadelphia-police-say/19925243/">AI model submitted false tip about unsolved murder, Philadelphia police say</a> — WPVI / 6abc</li>
<li><a href="https://ca.marketscreener.com/news/anthropic-ai-model-submits-false-homicide-tip-to-police-website-ce785ddcdb8ff122">Anthropic discloses fake tip to police among new rogue AI incidents</a> — Reuters via MarketScreener</li>
</ul>
<p><b>CrowdStrike reports AI-assisted South Korean financial intrusions</b></p>
<p>Attacker-hosted artifacts document AI use; actor identity and the affected organisation count remain uncertain.</p>
<ul>
<li><a href="https://www.crowdstrike.com/en-us/blog/unknown-threat-actor-uses-artex-to-target-south-korean-finance/">Unknown Threat Actor Uses AI-Driven ARTEX to Target South Korean Finance</a> — CrowdStrike</li>
</ul>
<p><b>Harness acquires selected Augment Code assets</b></p>
<p>The completed asset purchase includes Cosmos; broader connections between coding and delivery information remain integration plans.</p>
<ul>
<li><a href="https://www.harness.io/blog/harness-acquires-augment-code-cosmos">Harness acquires Augment Code to advance the Autonomous SDLC</a> — Harness</li>
<li><a href="https://www.harness.io/press-and-news/harness-acquires-augment-code">Harness Acquires Augment Code, Advancing the Autonomous SDLC from Idea to Production</a> — Harness</li>
</ul>
<p><b>TypeSafe announces $870 million Series A</b></p>
<p>Andreessen Horowitz confirms leading the financing; TypeSafe plans additional models and enterprise features.</p>
<ul>
<li><a href="https://typesafe.ai/blog/series-ai">TypeSafe A raises Series AI</a> — TypeSafe AI</li>
<li><a href="https://a16z.com/announcement/investing-in-typesafe-ai/">Investing in TypeSafe AI</a> — Andreessen Horowitz</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>GPT-6, Haiku 5.5 and Mistral Large 4 lead a review of multimodal models, local coding tools, public-health research, documented AI incidents and business announcements.</p>
<p><b>Stories and sources</b></p>
<p><b>OpenAI, Anthropic and Mistral announce model releases with distinct access conditions</b></p>
<p>GPT-6 adds interactive ChatGPT responses; Haiku 5.5 introduces adjustable effort and new pricing; Mistral Large 4 enters public API preview with weights forthcoming.</p>
<ul>
<li><a href="https://openai.com/index/gpt-6-for-everyone/">GPT-6 and Intelligent UI for everyone</a> — OpenAI</li>
<li><a href="https://help.openai.com/en/articles/6825453-chatgpt-release-notes">ChatGPT release notes</a> — OpenAI</li>
<li><a href="https://help.openai.com/en/articles/10128477-chatgpt-enterprise-and-edu-release-notes">ChatGPT Enterprise and Edu release notes</a> — OpenAI</li>
<li><a href="https://www.anthropic.com/claude-haiku-5-5">Introducing Claude Haiku 5.5</a> — Anthropic</li>
<li><a href="https://www.anthropic.com/claude/haiku?via=free">Claude Haiku</a> — Anthropic</li>
<li><a href="https://github.blog/changelog/2026-10-07-claude-haiku-5-5-in-github-copilot/">Claude Haiku 5.5 in GitHub Copilot</a> — GitHub</li>
<li><a href="https://support.claude.com/en/articles/17154008-monthly-api-credits-for-max-and-team-plans">Monthly API credits for Max and Team plans</a> — Anthropic</li>
<li><a href="https://platform.claude.com/docs/en/agents-and-tools/tool-use/browser-use-sdk">Browser and computer use with the SDK toolsets</a> — Anthropic</li>
<li><a href="https://mistral.ai/news/mistral-large-4/">Introducing Mistral Large 4</a> — Mistral AI</li>
<li><a href="https://docs.mistral.ai/models/mistral-large">Mistral Large 4</a> — Mistral AI</li>
<li><a href="https://docs.mistral.ai/resources/changelogs">Changelog</a> — Mistral AI</li>
</ul>
<p><b>Liquid AI releases hosted and downloadable multimodal decision models</b></p>
<p>The d1 releases return structured decisions, with different input support, qualified latency measurements and separate hosted billing conditions.</p>
<ul>
<li><a href="https://www.liquid.ai/blog/d1-decision-model">Introducing d1: The most capable decision model, now with vision</a> — Liquid AI</li>
<li><a href="https://www.liquid.ai/blog/d1-open">Open d1: Edge decision models for text, vision, and audio</a> — Liquid AI</li>
<li><a href="https://huggingface.co/LiquidAI/d1-3B">LiquidAI/d1-3B</a> — Liquid AI on Hugging Face</li>
<li><a href="https://huggingface.co/LiquidAI/d1-omni-600M">LiquidAI/d1-omni-600M</a> — Liquid AI on Hugging Face</li>
</ul>
<p><b>Google releases EmbeddingGemma 2 and local retrieval demonstrations</b></p>
<p>The multimodal embedding model has modular encoders, downloadable weights and configuration-dependent results; ML Kit integration remains forthcoming.</p>
<ul>
<li><a href="https://blog.google/innovation-and-ai/technology/developers-tools/embeddinggemma-2/">EmbeddingGemma 2: an open, lightweight multimodal embedding model</a> — Google</li>
<li><a href="https://developers.googleblog.com/embeddinggemma-2-the-developer-guide/">EmbeddingGemma 2: The Developer Guide</a> — Google</li>
<li><a href="https://developers.googleblog.com/google-ai-edge-with-embeddinggemma-2/">Bring multimodal semantic search to the edge with EmbeddingGemma 2</a> — Google</li>
<li><a href="https://ai.google.dev/gemma/docs/embeddinggemma/model_card_2">EmbeddingGemma 2 model card</a> — Google</li>
</ul>
<p><b>Microsoft plans local coding preview as Copilot sandboxing becomes available</b></p>
<p>Local model routing remains an experimental preview planned for later October; released containment controls have explicit boundaries.</p>
<ul>
<li><a href="https://blogs.windows.com/windowsexperience/2026/10/07/building-windows-for-hybrid-intelligence/">Building Windows for hybrid intelligence</a> — Microsoft</li>
<li><a href="https://commandline.microsoft.com/local-models-sandboxed-tools-github-windows/">Bringing local models and sandboxed tools to Windows and GitHub Copilot</a> — Microsoft</li>
<li><a href="https://github.blog/changelog/2026-10-07-local-sandboxing-for-github-copilot-now-generally-available/">Local sandboxing for GitHub Copilot now generally available</a> — GitHub</li>
<li><a href="https://blogs.windows.com/windowsdeveloper/2026/10/07/microsoft-execution-containers-policy-driven-containment-for-ai-agents/">Microsoft Execution Containers: Policy-driven containment for AI agents</a> — Microsoft</li>
</ul>
<p><b>Google reports five geographic AI public-health studies</b></p>
<p>The studies report task-specific prediction gains alongside forecasting and demographic limitations.</p>
<ul>
<li><a href="https://research.google/blog/earth-ais-planetary-geospatial-foundation-models-for-global-public-health/">Unlocking Earth AI’s planetary geospatial foundation models for global public health</a> — Google Research</li>
<li><a href="https://arxiv.org/abs/2610.05699">Planetary Geospatial Foundation Models: A New Paradigm for Global Public Health</a> — arXiv</li>
<li><a href="https://arxiv.org/html/2610.05699v1">Planetary Geospatial Foundation Models: A New Paradigm for Global Public Health</a> — arXiv</li>
</ul>
<p><b>Anthropic discloses unintended website actions and restricts evaluation internet access</b></p>
<p>The disclosure includes an invented Philadelphia homicide tip that police say remained in spam, plus other unintended interactions with outside systems.</p>
<ul>
<li><a href="https://www.anthropic.com/research/investigating-unintended-model-actions">Investigating unintended model actions in our evaluations and internal use</a> — Anthropic</li>
<li><a href="https://6abc.com/post/anthropic-ai-model-submitted-false-tip-unsolved-murder-philadelphia-police-say/19925243/">AI model submitted false tip about unsolved murder, Philadelphia police say</a> — WPVI / 6abc</li>
<li><a href="https://ca.marketscreener.com/news/anthropic-ai-model-submits-false-homicide-tip-to-police-website-ce785ddcdb8ff122">Anthropic discloses fake tip to police among new rogue AI incidents</a> — Reuters via MarketScreener</li>
</ul>
<p><b>CrowdStrike reports AI-assisted South Korean financial intrusions</b></p>
<p>Attacker-hosted artifacts document AI use; actor identity and the affected organisation count remain uncertain.</p>
<ul>
<li><a href="https://www.crowdstrike.com/en-us/blog/unknown-threat-actor-uses-artex-to-target-south-korean-finance/">Unknown Threat Actor Uses AI-Driven ARTEX to Target South Korean Finance</a> — CrowdStrike</li>
</ul>
<p><b>Harness acquires selected Augment Code assets</b></p>
<p>The completed asset purchase includes Cosmos; broader connections between coding and delivery information remain integration plans.</p>
<ul>
<li><a href="https://www.harness.io/blog/harness-acquires-augment-code-cosmos">Harness acquires Augment Code to advance the Autonomous SDLC</a> — Harness</li>
<li><a href="https://www.harness.io/press-and-news/harness-acquires-augment-code">Harness Acquires Augment Code, Advancing the Autonomous SDLC from Idea to Production</a> — Harness</li>
</ul>
<p><b>TypeSafe announces $870 million Series A</b></p>
<p>Andreessen Horowitz confirms leading the financing; TypeSafe plans additional models and enterprise features.</p>
<ul>
<li><a href="https://typesafe.ai/blog/series-ai">TypeSafe A raises Series AI</a> — TypeSafe AI</li>
<li><a href="https://a16z.com/announcement/investing-in-typesafe-ai/">Investing in TypeSafe AI</a> — Andreessen Horowitz</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Sun, 11 Oct 2026 05:17:55 +0200</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/b83df581/42b42769.mp3" length="18942789" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>1184</itunes:duration>
      <itunes:summary>GPT-6, Haiku 5.5 and Mistral Large 4 lead a review of multimodal models, local coding tools, public-health research, documented AI incidents and business announcements.</itunes:summary>
      <itunes:subtitle>GPT-6, Haiku 5.5 and Mistral Large 4 lead a review of multimodal models, local coding tools, public-health research, documented AI incidents and business announcements.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/b83df581/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>Local coding plans, agent controls and AI financing</title>
      <itunes:title>Local coding plans, agent controls and AI financing</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">f20ecf5c-143d-4179-a3a6-445f3ce602e8</guid>
      <link>https://share.transistor.fm/s/d13e384c</link>
      <description>
        <![CDATA[<p>Microsoft details local coding plans, GitHub expands agent security features, TypeSafe announces financing, and micro1 commits to data acquisition.</p>
<p><b>Stories and sources</b></p>
<p><b>Microsoft details local coding preview and available Copilot sandboxing</b></p>
<p>Local and cloud routing is planned for experimental preview; sandbox controls are generally available.</p>
<ul>
<li><a href="https://blogs.windows.com/windowsexperience/2026/10/07/building-windows-for-hybrid-intelligence/">Building Windows for hybrid intelligence</a> — Microsoft</li>
<li><a href="https://commandline.microsoft.com/local-models-sandboxed-tools-github-windows/">Bringing local models and sandboxed tools to Windows and GitHub Copilot</a> — Microsoft</li>
<li><a href="https://github.blog/changelog/2026-10-07-local-sandboxing-for-github-copilot-now-generally-available/">Local sandboxing for GitHub Copilot now generally available</a> — GitHub</li>
</ul>
<p><b>GitHub updates AI secret detection</b></p>
<p>Alert scans use a new model; additional checks have separate preview and billing conditions.</p>
<ul>
<li><a href="https://github.blog/changelog/2026-10-07-purpose-built-model-for-leaked-secret-detection/">Purpose-built model for leaked secret detection</a> — GitHub</li>
</ul>
<p><b>TypeSafe announces $870 million Series A</b></p>
<p>Financing supports planned model and enterprise expansion.</p>
<ul>
<li><a href="https://typesafe.ai/blog/series-ai">TypeSafe A raises Series AI</a> — TypeSafe AI</li>
<li><a href="https://a16z.com/announcement/investing-in-typesafe-ai/">Investing in TypeSafe AI</a> — Andreessen Horowitz</li>
</ul>
<p><b>micro1 commits $1 billion to enterprise data</b></p>
<p>Planned spending will acquire and license data for AI training environments.</p>
<ul>
<li><a href="https://www.micro1.ai/blog/micro1-commits-1b-to-enterprise-data-acquisition">micro1 announces commitment to spend $1B on enterprise data acquisition over the next 12 months</a> — micro1</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>Microsoft details local coding plans, GitHub expands agent security features, TypeSafe announces financing, and micro1 commits to data acquisition.</p>
<p><b>Stories and sources</b></p>
<p><b>Microsoft details local coding preview and available Copilot sandboxing</b></p>
<p>Local and cloud routing is planned for experimental preview; sandbox controls are generally available.</p>
<ul>
<li><a href="https://blogs.windows.com/windowsexperience/2026/10/07/building-windows-for-hybrid-intelligence/">Building Windows for hybrid intelligence</a> — Microsoft</li>
<li><a href="https://commandline.microsoft.com/local-models-sandboxed-tools-github-windows/">Bringing local models and sandboxed tools to Windows and GitHub Copilot</a> — Microsoft</li>
<li><a href="https://github.blog/changelog/2026-10-07-local-sandboxing-for-github-copilot-now-generally-available/">Local sandboxing for GitHub Copilot now generally available</a> — GitHub</li>
</ul>
<p><b>GitHub updates AI secret detection</b></p>
<p>Alert scans use a new model; additional checks have separate preview and billing conditions.</p>
<ul>
<li><a href="https://github.blog/changelog/2026-10-07-purpose-built-model-for-leaked-secret-detection/">Purpose-built model for leaked secret detection</a> — GitHub</li>
</ul>
<p><b>TypeSafe announces $870 million Series A</b></p>
<p>Financing supports planned model and enterprise expansion.</p>
<ul>
<li><a href="https://typesafe.ai/blog/series-ai">TypeSafe A raises Series AI</a> — TypeSafe AI</li>
<li><a href="https://a16z.com/announcement/investing-in-typesafe-ai/">Investing in TypeSafe AI</a> — Andreessen Horowitz</li>
</ul>
<p><b>micro1 commits $1 billion to enterprise data</b></p>
<p>Planned spending will acquire and license data for AI training environments.</p>
<ul>
<li><a href="https://www.micro1.ai/blog/micro1-commits-1b-to-enterprise-data-acquisition">micro1 announces commitment to spend $1B on enterprise data acquisition over the next 12 months</a> — micro1</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Sat, 10 Oct 2026 07:05:00 +0200</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/d13e384c/5cd0c826.mp3" length="3368794" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>211</itunes:duration>
      <itunes:summary>Microsoft details local coding plans, GitHub expands agent security features, TypeSafe announces financing, and micro1 commits to data acquisition.</itunes:summary>
      <itunes:subtitle>Microsoft details local coding plans, GitHub expands agent security features, TypeSafe announces financing, and micro1 commits to data acquisition.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/d13e384c/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>Liquid decision models, AI monitoring and Harness’s asset acquisition</title>
      <itunes:title>Liquid decision models, AI monitoring and Harness’s asset acquisition</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">a2a335e9-45bc-4179-98ba-565f5f63541b</guid>
      <link>https://share.transistor.fm/s/85098552</link>
      <description>
        <![CDATA[<p>Liquid AI’s downloadable models, Google’s public-health research, Harness’s Augment asset purchase, Goodfire monitoring results, Pwn2Own demonstrations and Manus financing.</p>
<p><b>Stories and sources</b></p>
<p><b>Liquid AI releases downloadable multimodal decision models</b></p>
<p>Two d1 models provide structured answers, with model-specific input support and qualified latency measurements.</p>
<ul>
<li><a href="https://www.liquid.ai/blog/d1-open">Open d1: Edge decision models for text, vision, and audio</a> — Liquid AI</li>
<li><a href="https://huggingface.co/LiquidAI/d1-3B">LiquidAI/d1-3B</a> — Liquid AI on Hugging Face</li>
<li><a href="https://www.liquid.ai/blog/d1-decision-model">Introducing d1: The most capable decision model, now with vision</a> — Liquid AI</li>
</ul>
<p><b>Google reports geographic AI public-health studies</b></p>
<p>Five case studies examine geographic context in health prediction, with task-specific gains and limitations.</p>
<ul>
<li><a href="https://research.google/blog/earth-ais-planetary-geospatial-foundation-models-for-global-public-health/">Unlocking Earth AI’s planetary geospatial foundation models for global public health</a> — Google Research</li>
<li><a href="https://arxiv.org/abs/2610.05699">Planetary Geospatial Foundation Models: A New Paradigm for Global Public Health</a> — arXiv</li>
<li><a href="https://arxiv.org/html/2610.05699v1">Planetary Geospatial Foundation Models: A New Paradigm for Global Public Health</a> — arXiv</li>
</ul>
<p><b>Harness acquires selected Augment Code assets</b></p>
<p>Harness announces a completed asset purchase and available Cosmos agents, alongside broader integration plans.</p>
<ul>
<li><a href="https://www.harness.io/press-and-news/harness-acquires-augment-code">Harness Acquires Augment Code, Advancing the Autonomous SDLC from Idea to Production</a> — Harness</li>
<li><a href="https://www.harness.io/blog/harness-acquires-augment-code-cosmos">Harness acquires Augment Code to advance the Autonomous SDLC</a> — Harness</li>
</ul>
<p><b>Goodfire reports cyber-monitoring results</b></p>
<p>Internal-signal monitoring detects harmful sessions in company testing, with false positives.</p>
<ul>
<li><a href="https://www.goodfire.com/research/production-cyber-monitors-on-kimi-k3">Training and Deploying Production Cyber Monitors on Kimi K3</a> — Goodfire</li>
</ul>
<p><b>Pwn2Own reports demonstrations against AI tools</b></p>
<p>Contest results cover Codex, LiteLLM and Chroma; they document controlled research.</p>
<ul>
<li><a href="https://www.zerodayinitiative.com/blog/2026/10/6/pwn2own-ireland-2026-day-one-results">Pwn2Own Ireland 2026 - Day One Results</a> — Zero Day Initiative</li>
<li><a href="https://www.zerodayinitiative.com/blog/2026/10/7/pwn2own-ireland-2026-day-two-results">Pwn2Own Ireland 2026 - Day Two Results</a> — Zero Day Initiative</li>
</ul>
<p><b>Manus parent reports financing above $500 million</b></p>
<p>Reuters reports completed funding following Manus’s separation from Meta.</p>
<ul>
<li><a href="https://www.investing.com/news/stock-market-news/manus-raises-more-than-500-million-after-meta-exit-4938004">Manus raises more than $500 million after Meta exit</a> — Reuters via Investing.com</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>Liquid AI’s downloadable models, Google’s public-health research, Harness’s Augment asset purchase, Goodfire monitoring results, Pwn2Own demonstrations and Manus financing.</p>
<p><b>Stories and sources</b></p>
<p><b>Liquid AI releases downloadable multimodal decision models</b></p>
<p>Two d1 models provide structured answers, with model-specific input support and qualified latency measurements.</p>
<ul>
<li><a href="https://www.liquid.ai/blog/d1-open">Open d1: Edge decision models for text, vision, and audio</a> — Liquid AI</li>
<li><a href="https://huggingface.co/LiquidAI/d1-3B">LiquidAI/d1-3B</a> — Liquid AI on Hugging Face</li>
<li><a href="https://www.liquid.ai/blog/d1-decision-model">Introducing d1: The most capable decision model, now with vision</a> — Liquid AI</li>
</ul>
<p><b>Google reports geographic AI public-health studies</b></p>
<p>Five case studies examine geographic context in health prediction, with task-specific gains and limitations.</p>
<ul>
<li><a href="https://research.google/blog/earth-ais-planetary-geospatial-foundation-models-for-global-public-health/">Unlocking Earth AI’s planetary geospatial foundation models for global public health</a> — Google Research</li>
<li><a href="https://arxiv.org/abs/2610.05699">Planetary Geospatial Foundation Models: A New Paradigm for Global Public Health</a> — arXiv</li>
<li><a href="https://arxiv.org/html/2610.05699v1">Planetary Geospatial Foundation Models: A New Paradigm for Global Public Health</a> — arXiv</li>
</ul>
<p><b>Harness acquires selected Augment Code assets</b></p>
<p>Harness announces a completed asset purchase and available Cosmos agents, alongside broader integration plans.</p>
<ul>
<li><a href="https://www.harness.io/press-and-news/harness-acquires-augment-code">Harness Acquires Augment Code, Advancing the Autonomous SDLC from Idea to Production</a> — Harness</li>
<li><a href="https://www.harness.io/blog/harness-acquires-augment-code-cosmos">Harness acquires Augment Code to advance the Autonomous SDLC</a> — Harness</li>
</ul>
<p><b>Goodfire reports cyber-monitoring results</b></p>
<p>Internal-signal monitoring detects harmful sessions in company testing, with false positives.</p>
<ul>
<li><a href="https://www.goodfire.com/research/production-cyber-monitors-on-kimi-k3">Training and Deploying Production Cyber Monitors on Kimi K3</a> — Goodfire</li>
</ul>
<p><b>Pwn2Own reports demonstrations against AI tools</b></p>
<p>Contest results cover Codex, LiteLLM and Chroma; they document controlled research.</p>
<ul>
<li><a href="https://www.zerodayinitiative.com/blog/2026/10/6/pwn2own-ireland-2026-day-one-results">Pwn2Own Ireland 2026 - Day One Results</a> — Zero Day Initiative</li>
<li><a href="https://www.zerodayinitiative.com/blog/2026/10/7/pwn2own-ireland-2026-day-two-results">Pwn2Own Ireland 2026 - Day Two Results</a> — Zero Day Initiative</li>
</ul>
<p><b>Manus parent reports financing above $500 million</b></p>
<p>Reuters reports completed funding following Manus’s separation from Meta.</p>
<ul>
<li><a href="https://www.investing.com/news/stock-market-news/manus-raises-more-than-500-million-after-meta-exit-4938004">Manus raises more than $500 million after Meta exit</a> — Reuters via Investing.com</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Fri, 09 Oct 2026 07:06:58 +0200</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/85098552/cdec9737.mp3" length="5059021" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>317</itunes:duration>
      <itunes:summary>Liquid AI’s downloadable models, Google’s public-health research, Harness’s Augment asset purchase, Goodfire monitoring results, Pwn2Own demonstrations and Manus financing.</itunes:summary>
      <itunes:subtitle>Liquid AI’s downloadable models, Google’s public-health research, Harness’s Augment asset purchase, Goodfire monitoring results, Pwn2Own demonstrations and Manus financing.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/85098552/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>ChatGPT and Haiku releases, AI security access and biology funding</title>
      <itunes:title>ChatGPT and Haiku releases, AI security access and biology funding</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">8777597b-3be2-4195-8979-5c34dc779373</guid>
      <link>https://share.transistor.fm/s/896e74ce</link>
      <description>
        <![CDATA[<p>New releases from OpenAI, Anthropic and Google; expanded cybersecurity access; CrowdStrike’s report on AI-assisted intrusions; and commitments to biological data for AI.</p>
<p><b>Stories and sources</b></p>
<p><b>OpenAI begins ChatGPT rollout of GPT-6 with Intelligent UI</b></p>
<p>The rollout adds interactive responses, with access phased by subscription tier.</p>
<ul>
<li><a href="https://openai.com/index/gpt-6-for-everyone/">GPT-6 and Intelligent UI for everyone</a> — OpenAI</li>
</ul>
<p><b>Anthropic releases Haiku 5.5 and cuts Sonnet cache pricing</b></p>
<p>Haiku adds adjustable effort and prompt-length-dependent pricing.</p>
<ul>
<li><a href="https://www.anthropic.com/claude-haiku-5-5">Introducing Claude Haiku 5.5</a> — Anthropic</li>
<li><a href="https://www.anthropic.com/claude/haiku?via=free">Claude Haiku</a> — Anthropic</li>
</ul>
<p><b>Google releases multimodal EmbeddingGemma 2</b></p>
<p>The downloadable model supports retrieval across several media types.</p>
<ul>
<li><a href="https://blog.google/innovation-and-ai/technology/developers-tools/embeddinggemma-2/">EmbeddingGemma 2: an open, lightweight multimodal embedding model</a> — Google</li>
</ul>
<p><b>Anthropic expands verified cybersecurity access</b></p>
<p>Three tiers distinguish defensive, authorized testing and specialized access.</p>
<ul>
<li><a href="https://www.anthropic.com/news/cyber-verification-program">Expanding the Cyber Verification Program</a> — Anthropic</li>
</ul>
<p><b>CrowdStrike reports AI-assisted South Korean financial intrusions</b></p>
<p>The report documents agent artifacts while retaining attribution uncertainty.</p>
<ul>
<li><a href="https://www.crowdstrike.com/en-us/blog/unknown-threat-actor-uses-artex-to-target-south-korean-finance/">Unknown Threat Actor Uses AI-Driven ARTEX to Target South Korean Finance</a> — CrowdStrike</li>
</ul>
<p><b>Biohub announces expanded AI biology data commitments</b></p>
<p>The $1.8 billion total combines commitments and resources from prior investment.</p>
<ul>
<li><a href="https://www.prnewswire.com/news-releases/international-cross-sector-collaboration-commits-nearly-2-billion-to-build-foundational-data-for-ai-models-to-predict-and-treat-disease-302901172.html">International, cross-sector collaboration commits nearly $2 billion to build foundational data for AI models to predict and treat disease</a> — Biohub via PR Newswire</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>New releases from OpenAI, Anthropic and Google; expanded cybersecurity access; CrowdStrike’s report on AI-assisted intrusions; and commitments to biological data for AI.</p>
<p><b>Stories and sources</b></p>
<p><b>OpenAI begins ChatGPT rollout of GPT-6 with Intelligent UI</b></p>
<p>The rollout adds interactive responses, with access phased by subscription tier.</p>
<ul>
<li><a href="https://openai.com/index/gpt-6-for-everyone/">GPT-6 and Intelligent UI for everyone</a> — OpenAI</li>
</ul>
<p><b>Anthropic releases Haiku 5.5 and cuts Sonnet cache pricing</b></p>
<p>Haiku adds adjustable effort and prompt-length-dependent pricing.</p>
<ul>
<li><a href="https://www.anthropic.com/claude-haiku-5-5">Introducing Claude Haiku 5.5</a> — Anthropic</li>
<li><a href="https://www.anthropic.com/claude/haiku?via=free">Claude Haiku</a> — Anthropic</li>
</ul>
<p><b>Google releases multimodal EmbeddingGemma 2</b></p>
<p>The downloadable model supports retrieval across several media types.</p>
<ul>
<li><a href="https://blog.google/innovation-and-ai/technology/developers-tools/embeddinggemma-2/">EmbeddingGemma 2: an open, lightweight multimodal embedding model</a> — Google</li>
</ul>
<p><b>Anthropic expands verified cybersecurity access</b></p>
<p>Three tiers distinguish defensive, authorized testing and specialized access.</p>
<ul>
<li><a href="https://www.anthropic.com/news/cyber-verification-program">Expanding the Cyber Verification Program</a> — Anthropic</li>
</ul>
<p><b>CrowdStrike reports AI-assisted South Korean financial intrusions</b></p>
<p>The report documents agent artifacts while retaining attribution uncertainty.</p>
<ul>
<li><a href="https://www.crowdstrike.com/en-us/blog/unknown-threat-actor-uses-artex-to-target-south-korean-finance/">Unknown Threat Actor Uses AI-Driven ARTEX to Target South Korean Finance</a> — CrowdStrike</li>
</ul>
<p><b>Biohub announces expanded AI biology data commitments</b></p>
<p>The $1.8 billion total combines commitments and resources from prior investment.</p>
<ul>
<li><a href="https://www.prnewswire.com/news-releases/international-cross-sector-collaboration-commits-nearly-2-billion-to-build-foundational-data-for-ai-models-to-predict-and-treat-disease-302901172.html">International, cross-sector collaboration commits nearly $2 billion to build foundational data for AI models to predict and treat disease</a> — Biohub via PR Newswire</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Thu, 08 Oct 2026 07:06:03 +0200</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/896e74ce/afe72fa2.mp3" length="3999912" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>250</itunes:duration>
      <itunes:summary>New releases from OpenAI, Anthropic and Google; expanded cybersecurity access; CrowdStrike’s report on AI-assisted intrusions; and commitments to biological data for AI.</itunes:summary>
      <itunes:subtitle>New releases from OpenAI, Anthropic and Google; expanded cybersecurity access; CrowdStrike’s report on AI-assisted intrusions; and commitments to biological data for AI.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/896e74ce/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>Mistral Large 4 preview and AI containment tests</title>
      <itunes:title>Mistral Large 4 preview and AI containment tests</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">bc16ab10-1869-46ee-b2cc-c8eabc3b4923</guid>
      <link>https://share.transistor.fm/s/c4b7aa13</link>
      <description>
        <![CDATA[<p>Mistral's public preview, missing Copilot usage data, Irregular's containment research, Microsoft's CaptiveCrunch update, and Vinci's financing.</p>
<p><b>Stories and sources</b></p>
<p><b>Mistral lists Large 4 public preview</b></p>
<p>The October 6 model documentation lists multimodal capabilities and a one-million-token context window.</p>
<ul>
<li><a href="https://docs.mistral.ai/models/mistral-large">Mistral Large 4</a> — Mistral AI</li>
</ul>
<p><b>GitHub discloses missing Copilot agent metrics</b></p>
<p>An editor-attribution problem omitted activity from reports; historical gaps cannot be repaired.</p>
<ul>
<li><a href="https://github.blog/changelog/2026-10-06-update-your-ide-to-restore-agent-activity-in-copilot-usage-metrics/">Update your IDE to restore agent activity in Copilot usage metrics</a> — GitHub</li>
</ul>
<p><b>Irregular reports controlled containment-test finding</b></p>
<p>Testing revealed unexpected cloud-network behaviour; Irregular reports no impact outside the controlled test.</p>
<ul>
<li><a href="https://www.irregular.com/research">Research</a> — Irregular</li>
<li><a href="https://www.irregular.com/research/containment-challenges-testing-the-boundaries-of-cyber-evaluations">Containment Challenges: Testing the Boundaries of Cyber Evaluations</a> — Irregular</li>
</ul>
<p><b>Microsoft reports renewed CaptiveCrunch activity</b></p>
<p>The October 5 update describes resumed operations and qualified evidence of AI-assisted malware development.</p>
<ul>
<li><a href="https://www.microsoft.com/en-us/security/blog/2026/07/31/captivecrunch-midnight-blizzard-targets-travelers-worldwide-for-malware-delivery-and-credential-theft/">CaptiveCrunch: Midnight Blizzard targets travelers worldwide for malware delivery and credential theft</a> — Microsoft Threat Intelligence</li>
</ul>
<p><b>Vinci announces $250 million Series B</b></p>
<p>The physics AI company reports a $1.5 billion valuation and plans broader engineering capabilities.</p>
<ul>
<li><a href="https://www.adventinternational.com/news/vinci-raises-250m-series-b-at-1-5b-valuation-to-build-the-intelligence-infrastructure-for-a-new-era-of-hardware-engineering/">Vinci raises $250M Series B at $1.5B valuation to build the intelligence infrastructure for a new era of hardware engineering</a> — Advent International</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>Mistral's public preview, missing Copilot usage data, Irregular's containment research, Microsoft's CaptiveCrunch update, and Vinci's financing.</p>
<p><b>Stories and sources</b></p>
<p><b>Mistral lists Large 4 public preview</b></p>
<p>The October 6 model documentation lists multimodal capabilities and a one-million-token context window.</p>
<ul>
<li><a href="https://docs.mistral.ai/models/mistral-large">Mistral Large 4</a> — Mistral AI</li>
</ul>
<p><b>GitHub discloses missing Copilot agent metrics</b></p>
<p>An editor-attribution problem omitted activity from reports; historical gaps cannot be repaired.</p>
<ul>
<li><a href="https://github.blog/changelog/2026-10-06-update-your-ide-to-restore-agent-activity-in-copilot-usage-metrics/">Update your IDE to restore agent activity in Copilot usage metrics</a> — GitHub</li>
</ul>
<p><b>Irregular reports controlled containment-test finding</b></p>
<p>Testing revealed unexpected cloud-network behaviour; Irregular reports no impact outside the controlled test.</p>
<ul>
<li><a href="https://www.irregular.com/research">Research</a> — Irregular</li>
<li><a href="https://www.irregular.com/research/containment-challenges-testing-the-boundaries-of-cyber-evaluations">Containment Challenges: Testing the Boundaries of Cyber Evaluations</a> — Irregular</li>
</ul>
<p><b>Microsoft reports renewed CaptiveCrunch activity</b></p>
<p>The October 5 update describes resumed operations and qualified evidence of AI-assisted malware development.</p>
<ul>
<li><a href="https://www.microsoft.com/en-us/security/blog/2026/07/31/captivecrunch-midnight-blizzard-targets-travelers-worldwide-for-malware-delivery-and-credential-theft/">CaptiveCrunch: Midnight Blizzard targets travelers worldwide for malware delivery and credential theft</a> — Microsoft Threat Intelligence</li>
</ul>
<p><b>Vinci announces $250 million Series B</b></p>
<p>The physics AI company reports a $1.5 billion valuation and plans broader engineering capabilities.</p>
<ul>
<li><a href="https://www.adventinternational.com/news/vinci-raises-250m-series-b-at-1-5b-valuation-to-build-the-intelligence-infrastructure-for-a-new-era-of-hardware-engineering/">Vinci raises $250M Series B at $1.5B valuation to build the intelligence infrastructure for a new era of hardware engineering</a> — Advent International</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Wed, 07 Oct 2026 07:04:57 +0200</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/c4b7aa13/12204670.mp3" length="2873512" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>180</itunes:duration>
      <itunes:summary>Mistral's public preview, missing Copilot usage data, Irregular's containment research, Microsoft's CaptiveCrunch update, and Vinci's financing.</itunes:summary>
      <itunes:subtitle>Mistral's public preview, missing Copilot usage data, Irregular's containment research, Microsoft's CaptiveCrunch update, and Vinci's financing.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/c4b7aa13/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>GitHub ReviewBench, AI security fixes and funding</title>
      <itunes:title>GitHub ReviewBench, AI security fixes and funding</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">736f1e70-d0c0-4993-b55b-1e90cdf38609</guid>
      <link>https://share.transistor.fm/s/09a6dd1d</link>
      <description>
        <![CDATA[<p>GitHub releases a code-review benchmark and API access; GitLab details two security issues; OneByZero and Pandektes announce financing.</p>
<p><b>Stories and sources</b></p>
<p><b>GitHub releases ReviewBench and adds code-review API access</b></p>
<p>ReviewBench enters research preview; Copilot review requests become available through REST and GraphQL.</p>
<ul>
<li><a href="https://github.blog/ai-and-ml/github-copilot/reviewbench-an-open-benchmark-for-ai-code-review/">ReviewBench: An open benchmark for AI code review</a> — GitHub</li>
<li><a href="https://github.blog/changelog/2026-10-02-copilot-code-review-api-support-and-new-default-effort-level/">Copilot code review: API support and new default effort level</a> — GitHub</li>
</ul>
<p><b>GitLab patches critical AI Gateway vulnerability</b></p>
<p>The flaw requires authenticated Duo Agent Platform access; GitLab-hosted gateways already have the fix.</p>
<ul>
<li><a href="https://docs.gitlab.com/releases/patches/other-patches/patch-release-gitlab-ai-gateway-19-4-1-released/">GitLab AI Gateway Critical Patch Release: 19.2.4, 19.3.2, and 19.4.1</a> — GitLab</li>
<li><a href="https://www.bleepingcomputer.com/news/security/gitlab-warns-of-critical-rce-vulnerability-in-ai-gateway-service/amp/">GitLab warns of critical RCE vulnerability in AI Gateway service</a> — BleepingComputer</li>
</ul>
<p><b>GitLab details fixed DeepSeek-Reasonix flaw</b></p>
<p>Research describes command execution during change viewing under specific configuration conditions.</p>
<ul>
<li><a href="https://about.gitlab.com/blog/deepseek-reasonix-vulnerability-discovered/">DeepSeek-Reasonix: How a poisoned config can hijack an AI coding agent</a> — GitLab</li>
</ul>
<p><b>OneByZero announces $20 million Series A</b></p>
<p>Financing supports planned enterprise AI deployment and platform expansion.</p>
<ul>
<li><a href="https://www.onebyzero.ai/press-releases/onebyzero-raises-20-million-series-a-jungle-ventures">OneByZero Raises US$20 Million Series A Led by Jungle Ventures to Scale Enterprise AI Across Asia Pacific</a> — OneByZero</li>
</ul>
<p><b>Pandektes announces €13.5 million Series A</b></p>
<p>The legal-information company describes expanded API access and plans for AI-oriented data infrastructure.</p>
<ul>
<li><a href="https://pandektes.com/blog/pandektes-13-5m-series-a">Pandektes raises €13.5M in Series A funding</a> — Pandektes</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>GitHub releases a code-review benchmark and API access; GitLab details two security issues; OneByZero and Pandektes announce financing.</p>
<p><b>Stories and sources</b></p>
<p><b>GitHub releases ReviewBench and adds code-review API access</b></p>
<p>ReviewBench enters research preview; Copilot review requests become available through REST and GraphQL.</p>
<ul>
<li><a href="https://github.blog/ai-and-ml/github-copilot/reviewbench-an-open-benchmark-for-ai-code-review/">ReviewBench: An open benchmark for AI code review</a> — GitHub</li>
<li><a href="https://github.blog/changelog/2026-10-02-copilot-code-review-api-support-and-new-default-effort-level/">Copilot code review: API support and new default effort level</a> — GitHub</li>
</ul>
<p><b>GitLab patches critical AI Gateway vulnerability</b></p>
<p>The flaw requires authenticated Duo Agent Platform access; GitLab-hosted gateways already have the fix.</p>
<ul>
<li><a href="https://docs.gitlab.com/releases/patches/other-patches/patch-release-gitlab-ai-gateway-19-4-1-released/">GitLab AI Gateway Critical Patch Release: 19.2.4, 19.3.2, and 19.4.1</a> — GitLab</li>
<li><a href="https://www.bleepingcomputer.com/news/security/gitlab-warns-of-critical-rce-vulnerability-in-ai-gateway-service/amp/">GitLab warns of critical RCE vulnerability in AI Gateway service</a> — BleepingComputer</li>
</ul>
<p><b>GitLab details fixed DeepSeek-Reasonix flaw</b></p>
<p>Research describes command execution during change viewing under specific configuration conditions.</p>
<ul>
<li><a href="https://about.gitlab.com/blog/deepseek-reasonix-vulnerability-discovered/">DeepSeek-Reasonix: How a poisoned config can hijack an AI coding agent</a> — GitLab</li>
</ul>
<p><b>OneByZero announces $20 million Series A</b></p>
<p>Financing supports planned enterprise AI deployment and platform expansion.</p>
<ul>
<li><a href="https://www.onebyzero.ai/press-releases/onebyzero-raises-20-million-series-a-jungle-ventures">OneByZero Raises US$20 Million Series A Led by Jungle Ventures to Scale Enterprise AI Across Asia Pacific</a> — OneByZero</li>
</ul>
<p><b>Pandektes announces €13.5 million Series A</b></p>
<p>The legal-information company describes expanded API access and plans for AI-oriented data infrastructure.</p>
<ul>
<li><a href="https://pandektes.com/blog/pandektes-13-5m-series-a">Pandektes raises €13.5M in Series A funding</a> — Pandektes</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Tue, 06 Oct 2026 07:05:36 +0200</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/09a6dd1d/7b1794f7.mp3" length="3380496" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>212</itunes:duration>
      <itunes:summary>GitHub releases a code-review benchmark and API access; GitLab details two security issues; OneByZero and Pandektes announce financing.</itunes:summary>
      <itunes:subtitle>GitHub releases a code-review benchmark and API access; GitLab details two security issues; OneByZero and Pandektes announce financing.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/09a6dd1d/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>Decision models, AI search and agent security</title>
      <itunes:title>Decision models, AI search and agent security</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">bda9f0b9-123f-4783-8cb8-31c88a3063d2</guid>
      <link>https://share.transistor.fm/s/f28d17a3</link>
      <description>
        <![CDATA[<p>Cloudflare and Strands release decision models, Cloudflare expands search, Salt Labs details a fixed Manus weakness, and Armadin announces new funding.</p>
<p><b>Stories and sources</b></p>
<p><b>Cloudflare releases Clef decision models</b></p>
<p>Clef and Clef-flash provide structured decisions, with hosted access and Apache 2.0 weights.</p>
<ul>
<li><a href="https://developers.cloudflare.com/changelog/post/2026-10-01-clef-workers-ai/">Introducing Clef: Cloudflare's first open-source decision models, now on Workers AI</a> — Cloudflare</li>
<li><a href="https://blog.cloudflare.com/clef-decision-models/">Introducing Clef: our open-source decision models, and new RL fine-tuning platform</a> — Cloudflare</li>
</ul>
<p><b>Strands releases a local decision model</b></p>
<p>Strands Decider 2B includes training resources; the release describes task limitations and version-specific latency testing.</p>
<ul>
<li><a href="https://strandsagents.com/blog/introducing-strands-decider/">Introducing Strands Decider 2B: a small, open source, decision model</a> — Strands Agents</li>
<li><a href="https://github.com/strands-labs/strands-decider">strands-labs/strands-decider</a> — Strands Labs / GitHub</li>
</ul>
<p><b>Cloudflare expands agent search and multimodal retrieval</b></p>
<p>Web Search API enters beta, while AI Search reaches general availability with image retrieval and November billing.</p>
<ul>
<li><a href="https://developers.cloudflare.com/changelog/post/2026-10-02-introducing-web-search-api/">Introducing Web Search API</a> — Cloudflare</li>
<li><a href="https://blog.cloudflare.com/ai-search-ga/">AI Search is now generally available</a> — Cloudflare</li>
<li><a href="https://developers.cloudflare.com/changelog/post/2026-10-01-ai-search-generally-available/">AI Search is generally available</a> — Cloudflare</li>
</ul>
<p><b>Salt Labs details a fixed Manus agent weakness</b></p>
<p>The October 1 technical account describes controlled testing in which execution preceded a security warning.</p>
<ul>
<li><a href="https://salt.security/blog/how-we-hijacked-an-ai-agent-with-a-single-email">How We Hijacked an AI Agent With a Single Email</a> — Salt Security</li>
<li><a href="https://www.prnewswire.com/news-releases/salt-labs-research-a-single-email-could-hijack-an-ai-agent-and-reach-a-users-connected-accounts-302895317.html">Salt Labs Research: A Single Email Could Hijack an AI Agent and Reach a User's Connected Accounts</a> — Salt Security / PR Newswire</li>
</ul>
<p><b>Armadin announces $255.5 million Series B</b></p>
<p>Armadin reports new financing for its AI security platform, with a valuation above $2.5 billion.</p>
<ul>
<li><a href="https://www.armadin.com/blog-posts/investors-back-armadin-platform-strategy-with-new-round-of-funding">Investors Back Armadin Platform Strategy with New Round of Funding</a> — Armadin</li>
<li><a href="https://www.prnewswire.com/news-releases/armadin-raises-255-5-million-series-b-to-scale-effective-autonomous-security-302895278.html">Armadin Raises $255.5 Million Series B to Scale Effective Autonomous Security</a> — Armadin / PR Newswire</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>Cloudflare and Strands release decision models, Cloudflare expands search, Salt Labs details a fixed Manus weakness, and Armadin announces new funding.</p>
<p><b>Stories and sources</b></p>
<p><b>Cloudflare releases Clef decision models</b></p>
<p>Clef and Clef-flash provide structured decisions, with hosted access and Apache 2.0 weights.</p>
<ul>
<li><a href="https://developers.cloudflare.com/changelog/post/2026-10-01-clef-workers-ai/">Introducing Clef: Cloudflare's first open-source decision models, now on Workers AI</a> — Cloudflare</li>
<li><a href="https://blog.cloudflare.com/clef-decision-models/">Introducing Clef: our open-source decision models, and new RL fine-tuning platform</a> — Cloudflare</li>
</ul>
<p><b>Strands releases a local decision model</b></p>
<p>Strands Decider 2B includes training resources; the release describes task limitations and version-specific latency testing.</p>
<ul>
<li><a href="https://strandsagents.com/blog/introducing-strands-decider/">Introducing Strands Decider 2B: a small, open source, decision model</a> — Strands Agents</li>
<li><a href="https://github.com/strands-labs/strands-decider">strands-labs/strands-decider</a> — Strands Labs / GitHub</li>
</ul>
<p><b>Cloudflare expands agent search and multimodal retrieval</b></p>
<p>Web Search API enters beta, while AI Search reaches general availability with image retrieval and November billing.</p>
<ul>
<li><a href="https://developers.cloudflare.com/changelog/post/2026-10-02-introducing-web-search-api/">Introducing Web Search API</a> — Cloudflare</li>
<li><a href="https://blog.cloudflare.com/ai-search-ga/">AI Search is now generally available</a> — Cloudflare</li>
<li><a href="https://developers.cloudflare.com/changelog/post/2026-10-01-ai-search-generally-available/">AI Search is generally available</a> — Cloudflare</li>
</ul>
<p><b>Salt Labs details a fixed Manus agent weakness</b></p>
<p>The October 1 technical account describes controlled testing in which execution preceded a security warning.</p>
<ul>
<li><a href="https://salt.security/blog/how-we-hijacked-an-ai-agent-with-a-single-email">How We Hijacked an AI Agent With a Single Email</a> — Salt Security</li>
<li><a href="https://www.prnewswire.com/news-releases/salt-labs-research-a-single-email-could-hijack-an-ai-agent-and-reach-a-users-connected-accounts-302895317.html">Salt Labs Research: A Single Email Could Hijack an AI Agent and Reach a User's Connected Accounts</a> — Salt Security / PR Newswire</li>
</ul>
<p><b>Armadin announces $255.5 million Series B</b></p>
<p>Armadin reports new financing for its AI security platform, with a valuation above $2.5 billion.</p>
<ul>
<li><a href="https://www.armadin.com/blog-posts/investors-back-armadin-platform-strategy-with-new-round-of-funding">Investors Back Armadin Platform Strategy with New Round of Funding</a> — Armadin</li>
<li><a href="https://www.prnewswire.com/news-releases/armadin-raises-255-5-million-series-b-to-scale-effective-autonomous-security-302895278.html">Armadin Raises $255.5 Million Series B to Scale Effective Autonomous Security</a> — Armadin / PR Newswire</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Mon, 05 Oct 2026 07:08:59 +0200</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/f28d17a3/195436dc.mp3" length="5058185" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>317</itunes:duration>
      <itunes:summary>Cloudflare and Strands release decision models, Cloudflare expands search, Salt Labs details a fixed Manus weakness, and Armadin announces new funding.</itunes:summary>
      <itunes:subtitle>Cloudflare and Strands release decision models, Cloudflare expands search, Salt Labs details a fixed Manus weakness, and Armadin announces new funding.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/f28d17a3/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>Ai2 releases scientific reporting model and training framework</title>
      <itunes:title>Ai2 releases scientific reporting model and training framework</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">15e8ea97-2c32-4d1c-ac1c-5271249e008b</guid>
      <link>https://share.transistor.fm/s/a71b8b1a</link>
      <description>
        <![CDATA[<p>AstaBrief and Olmo-core 3, Ling 3.1 Flash access, Google's AI vulnerability analysis, and ElevenLabs' employee tender offer.</p>
<p><b>Stories and sources</b></p>
<p><b>Ai2 releases AstaBrief for cited scientific reports</b></p>
<p>AstaBrief powers Asta's Fast mode; reported comparisons reflect development-era systems.</p>
<ul>
<li><a href="https://allenai.org/blog/astabrief">Open-sourcing AstaBrief, the fast report-generation model in Asta</a> — Ai2</li>
</ul>
<p><b>Ai2 releases Olmo-core 3 training framework</b></p>
<p>Reported throughput and capacity results have distinct test conditions.</p>
<ul>
<li><a href="https://allenai.org/blog/olmocore3">Introducing Olmo-core 3: Open, scalable training infrastructure for large MoEs</a> — Ai2</li>
</ul>
<p><b>Vercel adds Ling 3.1 Flash with temporary free access</b></p>
<p>The two model identifiers behave differently when the promotion ends.</p>
<ul>
<li><a href="https://vercel.com/changelog/ling-3-1-flash-is-now-available-on-ai-gateway">Ling 3.1 Flash is now available on AI Gateway</a> — Vercel</li>
</ul>
<p><b>Google analyzes AI-assisted vulnerability discovery</b></p>
<p>The identified sample shows different consequences, with attribution and selection limitations.</p>
<ul>
<li><a href="https://cloud.google.com/blog/topics/threat-intelligence/vulnerability-discovery-and-exploitation-trends-in-the-ai-era">Vulnerability Discovery and Exploitation Trends in the AI Era</a> — Google Threat Intelligence Group</li>
</ul>
<p><b>ElevenLabs announces completed employee tender</b></p>
<p>The company reports a $300 million tender at a $22 billion valuation.</p>
<ul>
<li><a href="https://elevenlabs.io/blog/tender-22bn">ElevenLabs valuation increases to $22 billion fueled by enterprise demand for conversational agents</a> — ElevenLabs</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>AstaBrief and Olmo-core 3, Ling 3.1 Flash access, Google's AI vulnerability analysis, and ElevenLabs' employee tender offer.</p>
<p><b>Stories and sources</b></p>
<p><b>Ai2 releases AstaBrief for cited scientific reports</b></p>
<p>AstaBrief powers Asta's Fast mode; reported comparisons reflect development-era systems.</p>
<ul>
<li><a href="https://allenai.org/blog/astabrief">Open-sourcing AstaBrief, the fast report-generation model in Asta</a> — Ai2</li>
</ul>
<p><b>Ai2 releases Olmo-core 3 training framework</b></p>
<p>Reported throughput and capacity results have distinct test conditions.</p>
<ul>
<li><a href="https://allenai.org/blog/olmocore3">Introducing Olmo-core 3: Open, scalable training infrastructure for large MoEs</a> — Ai2</li>
</ul>
<p><b>Vercel adds Ling 3.1 Flash with temporary free access</b></p>
<p>The two model identifiers behave differently when the promotion ends.</p>
<ul>
<li><a href="https://vercel.com/changelog/ling-3-1-flash-is-now-available-on-ai-gateway">Ling 3.1 Flash is now available on AI Gateway</a> — Vercel</li>
</ul>
<p><b>Google analyzes AI-assisted vulnerability discovery</b></p>
<p>The identified sample shows different consequences, with attribution and selection limitations.</p>
<ul>
<li><a href="https://cloud.google.com/blog/topics/threat-intelligence/vulnerability-discovery-and-exploitation-trends-in-the-ai-era">Vulnerability Discovery and Exploitation Trends in the AI Era</a> — Google Threat Intelligence Group</li>
</ul>
<p><b>ElevenLabs announces completed employee tender</b></p>
<p>The company reports a $300 million tender at a $22 billion valuation.</p>
<ul>
<li><a href="https://elevenlabs.io/blog/tender-22bn">ElevenLabs valuation increases to $22 billion fueled by enterprise demand for conversational agents</a> — ElevenLabs</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Sun, 04 Oct 2026 07:06:26 +0200</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/a71b8b1a/42bdc354.mp3" length="3015200" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>189</itunes:duration>
      <itunes:summary>AstaBrief and Olmo-core 3, Ling 3.1 Flash access, Google's AI vulnerability analysis, and ElevenLabs' employee tender offer.</itunes:summary>
      <itunes:subtitle>AstaBrief and Olmo-core 3, Ling 3.1 Flash access, Google's AI vulnerability analysis, and ElevenLabs' employee tender offer.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/a71b8b1a/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>Speech models, private training and reasoning extraction</title>
      <itunes:title>Speech models, private training and reasoning extraction</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">7e577cba-bab4-41a7-998f-098b5170920a</guid>
      <link>https://share.transistor.fm/s/22e324c5</link>
      <description>
        <![CDATA[<p>Microsoft’s speech models, Tavus’s restricted video preview, Google’s private training system, OpenAI’s campaign disclosure, and ZoomInfo’s DoubleO acquisition.</p>
<p><b>Stories and sources</b></p>
<p><b>Microsoft announces streaming transcription and two voice models</b></p>
<p>The transcription service enters public preview, alongside multilingual speech-generation models.</p>
<ul>
<li><a href="https://microsoft.ai/news/our-first-streaming-transcription-model/">Our first streaming transcription model debuts at no. 1 on Artificial Analysis</a> — Microsoft AI</li>
<li><a href="https://learn.microsoft.com/en-us/azure/ai-services/speech-service/mai-transcribe-2-streaming?context=/azure/foundry/context/context">MAI-Transcribe-2-Streaming overview</a> — Microsoft Learn</li>
</ul>
<p><b>Tavus introduces restricted Griffin preview</b></p>
<p>Tavus reports a small video-call study with undisclosed AI participation.</p>
<ul>
<li><a href="https://www.tavus.io/griffin">Griffin: The First Human Interaction Model</a> — Tavus</li>
</ul>
<p><b>Google announces protected server training used by Gboard</b></p>
<p>The federated-learning system uses encrypted uploads, auditable processing policies and trusted execution environments.</p>
<ul>
<li><a href="https://www.research.google/blog/toward-provably-private-learning-from-federated-data/">Toward provably private learning from federated data</a> — Google Research</li>
<li><a href="https://arxiv.org/abs/2609.31494">Toward provably private learning from federated data</a> — arXiv</li>
<li><a href="https://arxiv.org/html/2609.31494v4">Toward provably private learning from federated data</a> — arXiv</li>
</ul>
<p><b>OpenAI discloses reasoning-extraction campaign</b></p>
<p>OpenAI attributes part of the activity to Moonshot-associated individuals, with attribution uncertainty and limited public evidence.</p>
<ul>
<li><a href="https://openai.com/index/disrupting-a-coordinated-model-distillation-campaign/">Disrupting a coordinated model-distillation campaign</a> — OpenAI</li>
<li><a href="https://cyberscoop.com/openai-moonshot-ai-model-distillation-attack/">OpenAI reveals ‘novel’ encryption bypass used in distillation attack</a> — CyberScoop</li>
</ul>
<p><b>ZoomInfo announces DoubleO AI acquisition</b></p>
<p>The acquisition brings multi-agent coordination technology to planned sales workflows.</p>
<ul>
<li><a href="https://zoominfotechnologiesinc.gcs-web.com/news-releases/news-release-details/zoominfo-acquires-doubleoai-make-ai-agents-reliable-enough-run">ZoomInfo Acquires DoubleO.ai to Make AI Agents Reliable Enough to Run Go-to-Market Motions</a> — ZoomInfo</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>Microsoft’s speech models, Tavus’s restricted video preview, Google’s private training system, OpenAI’s campaign disclosure, and ZoomInfo’s DoubleO acquisition.</p>
<p><b>Stories and sources</b></p>
<p><b>Microsoft announces streaming transcription and two voice models</b></p>
<p>The transcription service enters public preview, alongside multilingual speech-generation models.</p>
<ul>
<li><a href="https://microsoft.ai/news/our-first-streaming-transcription-model/">Our first streaming transcription model debuts at no. 1 on Artificial Analysis</a> — Microsoft AI</li>
<li><a href="https://learn.microsoft.com/en-us/azure/ai-services/speech-service/mai-transcribe-2-streaming?context=/azure/foundry/context/context">MAI-Transcribe-2-Streaming overview</a> — Microsoft Learn</li>
</ul>
<p><b>Tavus introduces restricted Griffin preview</b></p>
<p>Tavus reports a small video-call study with undisclosed AI participation.</p>
<ul>
<li><a href="https://www.tavus.io/griffin">Griffin: The First Human Interaction Model</a> — Tavus</li>
</ul>
<p><b>Google announces protected server training used by Gboard</b></p>
<p>The federated-learning system uses encrypted uploads, auditable processing policies and trusted execution environments.</p>
<ul>
<li><a href="https://www.research.google/blog/toward-provably-private-learning-from-federated-data/">Toward provably private learning from federated data</a> — Google Research</li>
<li><a href="https://arxiv.org/abs/2609.31494">Toward provably private learning from federated data</a> — arXiv</li>
<li><a href="https://arxiv.org/html/2609.31494v4">Toward provably private learning from federated data</a> — arXiv</li>
</ul>
<p><b>OpenAI discloses reasoning-extraction campaign</b></p>
<p>OpenAI attributes part of the activity to Moonshot-associated individuals, with attribution uncertainty and limited public evidence.</p>
<ul>
<li><a href="https://openai.com/index/disrupting-a-coordinated-model-distillation-campaign/">Disrupting a coordinated model-distillation campaign</a> — OpenAI</li>
<li><a href="https://cyberscoop.com/openai-moonshot-ai-model-distillation-attack/">OpenAI reveals ‘novel’ encryption bypass used in distillation attack</a> — CyberScoop</li>
</ul>
<p><b>ZoomInfo announces DoubleO AI acquisition</b></p>
<p>The acquisition brings multi-agent coordination technology to planned sales workflows.</p>
<ul>
<li><a href="https://zoominfotechnologiesinc.gcs-web.com/news-releases/news-release-details/zoominfo-acquires-doubleoai-make-ai-agents-reliable-enough-run">ZoomInfo Acquires DoubleO.ai to Make AI Agents Reliable Enough to Run Go-to-Market Motions</a> — ZoomInfo</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Sat, 03 Oct 2026 07:08:20 +0200</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/22e324c5/4add050a.mp3" length="4372732" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>274</itunes:duration>
      <itunes:summary>Microsoft’s speech models, Tavus’s restricted video preview, Google’s private training system, OpenAI’s campaign disclosure, and ZoomInfo’s DoubleO acquisition.</itunes:summary>
      <itunes:subtitle>Microsoft’s speech models, Tavus’s restricted video preview, Google’s private training system, OpenAI’s campaign disclosure, and ZoomInfo’s DoubleO acquisition.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/22e324c5/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>Gemini 4 Argon, Claude Mods and AI survey research</title>
      <itunes:title>Gemini 4 Argon, Claude Mods and AI survey research</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">74a86fe9-3f7a-4cab-95f8-8cc83179d9ab</guid>
      <link>https://share.transistor.fm/s/09f979c8</link>
      <description>
        <![CDATA[<p>Google begins restricted Argon access, Anthropic updates Claude Code, Pew tests synthetic polling, Transluce reports failed agent probes, and Salesforce agrees to acquire Listen Labs.</p>
<p><b>Stories and sources</b></p>
<p><b>Google announces restricted Gemini 4 Argon rollout</b></p>
<p>Trusted cyber defenders receive initial access; broader availability remains forthcoming.</p>
<ul>
<li><a href="https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-argon/">Gemini 4 Argon: our next era of frontier intelligence</a> — Google</li>
</ul>
<p><b>Claude Code adds Mods and fixes permission safeguards</b></p>
<p>An optional side agent accompanies deeper plugin customization and permission fixes.</p>
<ul>
<li><a href="https://github.com/anthropics/claude-code/releases/tag/v2.1.287">Release v2.1.287</a> — Anthropic</li>
</ul>
<p><b>Pew reports errors in AI-generated polling</b></p>
<p>A controlled comparison finds substantial survey error, with findings bounded by the tested models and configurations.</p>
<ul>
<li><a href="https://www.pewresearch.org/data-labs/2026/09/30/can-ai-stand-in-for-human-survey-takers-not-really/">Can AI Stand In for Human Survey-Takers? Not Really</a> — Pew Research Center</li>
<li><a href="https://www.pewresearch.org/data-labs/2026/09/30/methodology-silicon-samples/">Methodology</a> — Pew Research Center</li>
</ul>
<p><b>Transluce reports failed agent probes of government websites</b></p>
<p>The report finds no nonpublic-data access in its datasets and retains attribution uncertainty.</p>
<ul>
<li><a href="https://transluce.org/us-canada-gov">AI Agents Targeted U.S. and Canadian Government Websites</a> — Transluce</li>
<li><a href="https://www.cyber.gc.ca/en/news-events/statement-regarding-reported-activity-targeting-government-canada-websites">Statement regarding reported activity targeting Government of Canada websites</a> — Canadian Centre for Cyber Security</li>
</ul>
<p><b>Salesforce agrees to acquire Listen Labs</b></p>
<p>The AI customer-research acquisition remains subject to closing conditions.</p>
<ul>
<li><a href="https://www.salesforce.com/news/stories/salesforce-signs-definitive-agreement-to-acquire-listen-labs/">Salesforce Signs Definitive Agreement to Acquire Listen Labs</a> — Salesforce</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>Google begins restricted Argon access, Anthropic updates Claude Code, Pew tests synthetic polling, Transluce reports failed agent probes, and Salesforce agrees to acquire Listen Labs.</p>
<p><b>Stories and sources</b></p>
<p><b>Google announces restricted Gemini 4 Argon rollout</b></p>
<p>Trusted cyber defenders receive initial access; broader availability remains forthcoming.</p>
<ul>
<li><a href="https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-argon/">Gemini 4 Argon: our next era of frontier intelligence</a> — Google</li>
</ul>
<p><b>Claude Code adds Mods and fixes permission safeguards</b></p>
<p>An optional side agent accompanies deeper plugin customization and permission fixes.</p>
<ul>
<li><a href="https://github.com/anthropics/claude-code/releases/tag/v2.1.287">Release v2.1.287</a> — Anthropic</li>
</ul>
<p><b>Pew reports errors in AI-generated polling</b></p>
<p>A controlled comparison finds substantial survey error, with findings bounded by the tested models and configurations.</p>
<ul>
<li><a href="https://www.pewresearch.org/data-labs/2026/09/30/can-ai-stand-in-for-human-survey-takers-not-really/">Can AI Stand In for Human Survey-Takers? Not Really</a> — Pew Research Center</li>
<li><a href="https://www.pewresearch.org/data-labs/2026/09/30/methodology-silicon-samples/">Methodology</a> — Pew Research Center</li>
</ul>
<p><b>Transluce reports failed agent probes of government websites</b></p>
<p>The report finds no nonpublic-data access in its datasets and retains attribution uncertainty.</p>
<ul>
<li><a href="https://transluce.org/us-canada-gov">AI Agents Targeted U.S. and Canadian Government Websites</a> — Transluce</li>
<li><a href="https://www.cyber.gc.ca/en/news-events/statement-regarding-reported-activity-targeting-government-canada-websites">Statement regarding reported activity targeting Government of Canada websites</a> — Canadian Centre for Cyber Security</li>
</ul>
<p><b>Salesforce agrees to acquire Listen Labs</b></p>
<p>The AI customer-research acquisition remains subject to closing conditions.</p>
<ul>
<li><a href="https://www.salesforce.com/news/stories/salesforce-signs-definitive-agreement-to-acquire-listen-labs/">Salesforce Signs Definitive Agreement to Acquire Listen Labs</a> — Salesforce</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Fri, 02 Oct 2026 07:06:36 +0200</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/09f979c8/936b3d86.mp3" length="3693548" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>231</itunes:duration>
      <itunes:summary>Google begins restricted Argon access, Anthropic updates Claude Code, Pew tests synthetic polling, Transluce reports failed agent probes, and Salesforce agrees to acquire Listen Labs.</itunes:summary>
      <itunes:subtitle>Google begins restricted Argon access, Anthropic updates Claude Code, Pew tests synthetic polling, Transluce reports failed agent probes, and Salesforce agrees to acquire Listen Labs.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/09f979c8/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>OpenAI launches Sol update and Dots; Meta expands Muse</title>
      <itunes:title>OpenAI launches Sol update and Dots; Meta expands Muse</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">c56eeaa5-842d-46ad-a50b-24e9b479399e</guid>
      <link>https://share.transistor.fm/s/15bf367f</link>
      <description>
        <![CDATA[<p>New models and agents from OpenAI, Meta’s business connectors, Nasdaq’s AI assistant, training-safety proposals, malicious Custom GPTs and EliseAI’s financing.</p>
<p><b>Stories and sources</b></p>
<p><b>OpenAI releases GPT-6.1 Sol and begins Dots rollout</b></p>
<p>OpenAI announces revised model capabilities, persistent agents and developer tools, with distinct availability conditions.</p>
<ul>
<li><a href="https://openai.com/news/">OpenAI News</a> — OpenAI</li>
<li><a href="https://openai.com/index/introducing-gpt-6-1-sol/">Introducing GPT-6.1 Sol</a> — OpenAI</li>
<li><a href="https://openai.com/index/introducing-dots/">Introducing dots</a> — OpenAI</li>
<li><a href="https://openai.com/index/devday-2026-recap/">DevDay 2026 Recap</a> — OpenAI</li>
</ul>
<p><b>Meta expands Muse with small-business connectors</b></p>
<p>New business integrations retain approval requirements for publishing, sending and spending.</p>
<ul>
<li><a href="https://about.fb.com/news/2026/09/introducing-muse-small-business/">The Future Is for Everyone: Muse for Small Business</a> — Meta</li>
</ul>
<p><b>Nasdaq launches Calypso AI environment</b></p>
<p>An assistant is available, with additional operational agents planned.</p>
<ul>
<li><a href="https://press.aboutamazon.com/aws/2026/9/nasdaq-calypso-launches-framework-to-scale-ai-adoption-across-the-trade-lifecycle">Nasdaq Calypso Launches Agentic Capabilities to Scale AI Adoption Across the Trade Lifecycle</a> — Nasdaq via Amazon US Press Center</li>
</ul>
<p><b>OpenAI proposes training safety-case guidelines</b></p>
<p>The recommendations address containment, monitoring and review; implementation remains underway.</p>
<ul>
<li><a href="https://openai.com/index/towards-safety-cases-for-frontier-ai-training/">Towards safety cases for frontier AI training</a> — OpenAI</li>
</ul>
<p><b>Huntress documents malicious Custom GPT lures</b></p>
<p>Two confirmed infections involved Custom GPTs; researchers found a replacement lure September 27.</p>
<ul>
<li><a href="https://www.huntress.com/blog/chatgpt-custom-gpts-clickfix-rat">Attackers Abuse ChatGPT Custom GPTs to Deliver RAT via ClickFix</a> — Huntress</li>
</ul>
<p><b>EliseAI announces $350 million funding round</b></p>
<p>The housing and healthcare AI company reports a $4 billion valuation and plans expansion.</p>
<ul>
<li><a href="https://eliseai.com/blog/intelligence-is-getting-cheaper-life-should-too">Intelligence is getting cheaper. Life should too.</a> — EliseAI</li>
<li><a href="https://www.boursorama.com/bourse/actualites/la-start-up-specialisee-dans-l-ia-eliseai-est-valorisee-a-4-milliards-de-dollars-lors-de-son-dernier-tour-de-table-32491fad79368469ff471f4d4ed35cb8">La start-up spécialisée dans l'IA EliseAI est valorisée à 4 milliards de dollars lors de son dernier tour de table</a> — Reuters via Boursorama</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>New models and agents from OpenAI, Meta’s business connectors, Nasdaq’s AI assistant, training-safety proposals, malicious Custom GPTs and EliseAI’s financing.</p>
<p><b>Stories and sources</b></p>
<p><b>OpenAI releases GPT-6.1 Sol and begins Dots rollout</b></p>
<p>OpenAI announces revised model capabilities, persistent agents and developer tools, with distinct availability conditions.</p>
<ul>
<li><a href="https://openai.com/news/">OpenAI News</a> — OpenAI</li>
<li><a href="https://openai.com/index/introducing-gpt-6-1-sol/">Introducing GPT-6.1 Sol</a> — OpenAI</li>
<li><a href="https://openai.com/index/introducing-dots/">Introducing dots</a> — OpenAI</li>
<li><a href="https://openai.com/index/devday-2026-recap/">DevDay 2026 Recap</a> — OpenAI</li>
</ul>
<p><b>Meta expands Muse with small-business connectors</b></p>
<p>New business integrations retain approval requirements for publishing, sending and spending.</p>
<ul>
<li><a href="https://about.fb.com/news/2026/09/introducing-muse-small-business/">The Future Is for Everyone: Muse for Small Business</a> — Meta</li>
</ul>
<p><b>Nasdaq launches Calypso AI environment</b></p>
<p>An assistant is available, with additional operational agents planned.</p>
<ul>
<li><a href="https://press.aboutamazon.com/aws/2026/9/nasdaq-calypso-launches-framework-to-scale-ai-adoption-across-the-trade-lifecycle">Nasdaq Calypso Launches Agentic Capabilities to Scale AI Adoption Across the Trade Lifecycle</a> — Nasdaq via Amazon US Press Center</li>
</ul>
<p><b>OpenAI proposes training safety-case guidelines</b></p>
<p>The recommendations address containment, monitoring and review; implementation remains underway.</p>
<ul>
<li><a href="https://openai.com/index/towards-safety-cases-for-frontier-ai-training/">Towards safety cases for frontier AI training</a> — OpenAI</li>
</ul>
<p><b>Huntress documents malicious Custom GPT lures</b></p>
<p>Two confirmed infections involved Custom GPTs; researchers found a replacement lure September 27.</p>
<ul>
<li><a href="https://www.huntress.com/blog/chatgpt-custom-gpts-clickfix-rat">Attackers Abuse ChatGPT Custom GPTs to Deliver RAT via ClickFix</a> — Huntress</li>
</ul>
<p><b>EliseAI announces $350 million funding round</b></p>
<p>The housing and healthcare AI company reports a $4 billion valuation and plans expansion.</p>
<ul>
<li><a href="https://eliseai.com/blog/intelligence-is-getting-cheaper-life-should-too">Intelligence is getting cheaper. Life should too.</a> — EliseAI</li>
<li><a href="https://www.boursorama.com/bourse/actualites/la-start-up-specialisee-dans-l-ia-eliseai-est-valorisee-a-4-milliards-de-dollars-lors-de-son-dernier-tour-de-table-32491fad79368469ff471f4d4ed35cb8">La start-up spécialisée dans l'IA EliseAI est valorisée à 4 milliards de dollars lors de son dernier tour de table</a> — Reuters via Boursorama</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Wed, 30 Sep 2026 07:08:52 +0200</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/15bf367f/a363943b.mp3" length="4969995" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>311</itunes:duration>
      <itunes:summary>New models and agents from OpenAI, Meta’s business connectors, Nasdaq’s AI assistant, training-safety proposals, malicious Custom GPTs and EliseAI’s financing.</itunes:summary>
      <itunes:subtitle>New models and agents from OpenAI, Meta’s business connectors, Nasdaq’s AI assistant, training-safety proposals, malicious Custom GPTs and EliseAI’s financing.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/15bf367f/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>Sonnet 5.5, NVIDIA agent controls and AMD’s World Labs agreement</title>
      <itunes:title>Sonnet 5.5, NVIDIA agent controls and AMD’s World Labs agreement</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">e0645272-fcc5-447d-83b8-c721d77c2417</guid>
      <link>https://share.transistor.fm/s/4d455df7</link>
      <description>
        <![CDATA[<p>Anthropic releases Sonnet 5.5, NVIDIA announces agent security controls, Microsoft reports Azure destruction linked to JADEPUFFER, and AMD announces a World Labs acquisition agreement. Plus financing for SiMa.ai and Modulate.</p>
<p><b>Stories and sources</b></p>
<p><b>Anthropic releases Sonnet 5.5; GitHub begins Copilot rollout</b></p>
<p>Sonnet 5.5 brings reported efficiency gains, cybersecurity fallbacks and a gradual Copilot rollout.</p>
<ul>
<li><a href="https://www.anthropic.com/claude-sonnet-5-5">Introducing Claude Sonnet 5.5</a> — Anthropic</li>
<li><a href="https://github.blog/changelog/2026-09-28-claude-sonnet-5-5-in-github-copilot/">Claude Sonnet 5.5 in GitHub Copilot</a> — GitHub</li>
</ul>
<p><b>NVIDIA announces Open Agent Safety Platform</b></p>
<p>OpenShell software and the Sentry reference design provide agent controls across software and hardware.</p>
<ul>
<li><a href="https://nvidianews.nvidia.com/news/open-agent-safety-platform">NVIDIA Launches Open Agent Safety Platform to Secure Agents From Testing to Deployment</a> — NVIDIA</li>
<li><a href="https://developer.nvidia.com/blog/nvidia-open-agent-safety-platform-a-reference-for-continuous-in-silicon-agent-monitoring/">NVIDIA Open Agent Safety Platform: A Reference for Continuous In-Silicon Agent Monitoring</a> — NVIDIA</li>
</ul>
<p><b>Microsoft links destructive Azure activity to JADEPUFFER</b></p>
<p>New findings document Azure destruction; model-specific attribution remains unestablished.</p>
<ul>
<li><a href="https://www.microsoft.com/en-us/security/blog/2026/09/25/storm-3168-agentic-driven-cloud-attacks-using-compromised-service-principals/">Storm-3168: Agentic-driven cloud attacks using compromised service principals</a> — Microsoft</li>
<li><a href="https://www.sysdig.com/blog/jadepuffer-agentic-ransomware-for-automated-database-extortion">JADEPUFFER: Agentic ransomware for automated database extortion</a> — Sysdig</li>
</ul>
<p><b>AMD agrees to acquire World Labs</b></p>
<p>The approximately $8.2 billion all-stock deal remains subject to closing conditions.</p>
<ul>
<li><a href="https://www.globenewswire.com/news-release/2026/09/28/3370256/0/en/amd-to-acquire-world-labs-to-advance-the-future-of-ai-compute.html">AMD to Acquire World Labs to Advance the Future of AI Compute</a> — AMD via GlobeNewswire</li>
</ul>
<p><b>SiMa.ai announces $150 million Series C</b></p>
<p>Financing supports software and planned physical AI hardware.</p>
<ul>
<li><a href="https://sima.ai/press-release/sima-ai-reaches-1-45b-valuation-with-500-million-in-total-funding-to-scale-physical-ai-in-humanoids-automotive-and-drones/">SiMa.ai Reaches $1.45B Valuation with $500 Million in Total Funding to Scale Physical AI in Humanoids, Automotive, and Drones</a> — SiMa.ai</li>
</ul>
<p><b>Modulate announces $25 million audio AI financing</b></p>
<p>The company plans further audio-model and developer ecosystem investment.</p>
<ul>
<li><a href="https://www.modulate.ai/press-releases/modulate-raises-25m-to-scale-its-lead-in-frontier-audio-native-ai">Modulate Raises $25M to Scale Its Lead in Frontier Audio-Native AI</a> — Modulate</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>Anthropic releases Sonnet 5.5, NVIDIA announces agent security controls, Microsoft reports Azure destruction linked to JADEPUFFER, and AMD announces a World Labs acquisition agreement. Plus financing for SiMa.ai and Modulate.</p>
<p><b>Stories and sources</b></p>
<p><b>Anthropic releases Sonnet 5.5; GitHub begins Copilot rollout</b></p>
<p>Sonnet 5.5 brings reported efficiency gains, cybersecurity fallbacks and a gradual Copilot rollout.</p>
<ul>
<li><a href="https://www.anthropic.com/claude-sonnet-5-5">Introducing Claude Sonnet 5.5</a> — Anthropic</li>
<li><a href="https://github.blog/changelog/2026-09-28-claude-sonnet-5-5-in-github-copilot/">Claude Sonnet 5.5 in GitHub Copilot</a> — GitHub</li>
</ul>
<p><b>NVIDIA announces Open Agent Safety Platform</b></p>
<p>OpenShell software and the Sentry reference design provide agent controls across software and hardware.</p>
<ul>
<li><a href="https://nvidianews.nvidia.com/news/open-agent-safety-platform">NVIDIA Launches Open Agent Safety Platform to Secure Agents From Testing to Deployment</a> — NVIDIA</li>
<li><a href="https://developer.nvidia.com/blog/nvidia-open-agent-safety-platform-a-reference-for-continuous-in-silicon-agent-monitoring/">NVIDIA Open Agent Safety Platform: A Reference for Continuous In-Silicon Agent Monitoring</a> — NVIDIA</li>
</ul>
<p><b>Microsoft links destructive Azure activity to JADEPUFFER</b></p>
<p>New findings document Azure destruction; model-specific attribution remains unestablished.</p>
<ul>
<li><a href="https://www.microsoft.com/en-us/security/blog/2026/09/25/storm-3168-agentic-driven-cloud-attacks-using-compromised-service-principals/">Storm-3168: Agentic-driven cloud attacks using compromised service principals</a> — Microsoft</li>
<li><a href="https://www.sysdig.com/blog/jadepuffer-agentic-ransomware-for-automated-database-extortion">JADEPUFFER: Agentic ransomware for automated database extortion</a> — Sysdig</li>
</ul>
<p><b>AMD agrees to acquire World Labs</b></p>
<p>The approximately $8.2 billion all-stock deal remains subject to closing conditions.</p>
<ul>
<li><a href="https://www.globenewswire.com/news-release/2026/09/28/3370256/0/en/amd-to-acquire-world-labs-to-advance-the-future-of-ai-compute.html">AMD to Acquire World Labs to Advance the Future of AI Compute</a> — AMD via GlobeNewswire</li>
</ul>
<p><b>SiMa.ai announces $150 million Series C</b></p>
<p>Financing supports software and planned physical AI hardware.</p>
<ul>
<li><a href="https://sima.ai/press-release/sima-ai-reaches-1-45b-valuation-with-500-million-in-total-funding-to-scale-physical-ai-in-humanoids-automotive-and-drones/">SiMa.ai Reaches $1.45B Valuation with $500 Million in Total Funding to Scale Physical AI in Humanoids, Automotive, and Drones</a> — SiMa.ai</li>
</ul>
<p><b>Modulate announces $25 million audio AI financing</b></p>
<p>The company plans further audio-model and developer ecosystem investment.</p>
<ul>
<li><a href="https://www.modulate.ai/press-releases/modulate-raises-25m-to-scale-its-lead-in-frontier-audio-native-ai">Modulate Raises $25M to Scale Its Lead in Frontier Audio-Native AI</a> — Modulate</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Tue, 29 Sep 2026 07:08:03 +0200</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/4d455df7/83d05487.mp3" length="4917332" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>308</itunes:duration>
      <itunes:summary>Anthropic releases Sonnet 5.5, NVIDIA announces agent security controls, Microsoft reports Azure destruction linked to JADEPUFFER, and AMD announces a World Labs acquisition agreement. Plus financing for SiMa.ai and Modulate.</itunes:summary>
      <itunes:subtitle>Anthropic releases Sonnet 5.5, NVIDIA announces agent security controls, Microsoft reports Azure destruction linked to JADEPUFFER, and AMD announces a World Labs acquisition agreement. Plus financing for SiMa.ai and Modulate.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/4d455df7/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>Live Avatar, PageBreak findings and AI funding</title>
      <itunes:title>Live Avatar, PageBreak findings and AI funding</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">66661e2e-ec81-4fb0-95be-ecc3fae0381b</guid>
      <link>https://share.transistor.fm/s/81145c0b</link>
      <description>
        <![CDATA[<p>Google introduces Live Avatar and reports PageBreak security findings; BNP Paribas announces an AI partnership; Feather and BigHat announce funding.</p>
<p><b>Stories and sources</b></p>
<p><b>Google introduces Live Avatar</b></p>
<p>Enterprise availability includes restrictions on custom avatars.</p>
<ul>
<li><a href="https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-8-live-with-live-avatar/">Introducing Gemini 3.8 Live with Live Avatar</a> — Google</li>
</ul>
<p><b>Google reports PageBreak security findings</b></p>
<p>Internal testing combines AI with validation and acknowledges coverage gaps.</p>
<ul>
<li><a href="https://blog.google/security/agentic-hacks-real-proofs-inside-googles-pagebreak-project/">Agentic Hacks, Real Proofs: Inside Google's PageBreak Project</a> — Google</li>
</ul>
<p><b>BNP Paribas announces Google Cloud partnership</b></p>
<p>Planned banking agents include credit-memo assistance.</p>
<ul>
<li><a href="https://www.googlecloudpresscorner.com/2026-09-24-BNP-Paribas-and-Google-Cloud-Announce-New-Partnership-on-Agentic-AI-and-Cloud-Innovation">BNP Paribas and Google Cloud Announce New Partnership on Agentic AI and Cloud Innovation</a> — Google Cloud</li>
</ul>
<p><b>Feather announces robotics funding</b></p>
<p>The developer platform reports a completed pre-seed round.</p>
<ul>
<li><a href="https://www.investegate.co.uk/index.php/announcement/rns/seed-innovations-limited--seed/investee-company-update-feather-robotics-inc-/9792430">Investee Company Update: Feather Robotics Inc.</a> — SEED Innovations Limited via RNS Reach / Investegate</li>
</ul>
<p><b>BigHat announces $75 million Series C</b></p>
<p>Financing supports AI protein design and therapeutic development.</p>
<ul>
<li><a href="https://markets.financialcontent.com/dowtheoryletters/article/bizwire-2026-9-24-bighat-biosciences-announces-75-million-series-c-financing-to-advance-ai-designed-therapeutics-and-leading-agentic-data-platform-for-rapid-protein-design">BigHat Biosciences Announces $75 Million Series C Financing to Advance AI-Designed Therapeutics and Leading Agentic Data Platform for Rapid Protein Design</a> — BigHat Biosciences via Business Wire / FinancialContent</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>Google introduces Live Avatar and reports PageBreak security findings; BNP Paribas announces an AI partnership; Feather and BigHat announce funding.</p>
<p><b>Stories and sources</b></p>
<p><b>Google introduces Live Avatar</b></p>
<p>Enterprise availability includes restrictions on custom avatars.</p>
<ul>
<li><a href="https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-8-live-with-live-avatar/">Introducing Gemini 3.8 Live with Live Avatar</a> — Google</li>
</ul>
<p><b>Google reports PageBreak security findings</b></p>
<p>Internal testing combines AI with validation and acknowledges coverage gaps.</p>
<ul>
<li><a href="https://blog.google/security/agentic-hacks-real-proofs-inside-googles-pagebreak-project/">Agentic Hacks, Real Proofs: Inside Google's PageBreak Project</a> — Google</li>
</ul>
<p><b>BNP Paribas announces Google Cloud partnership</b></p>
<p>Planned banking agents include credit-memo assistance.</p>
<ul>
<li><a href="https://www.googlecloudpresscorner.com/2026-09-24-BNP-Paribas-and-Google-Cloud-Announce-New-Partnership-on-Agentic-AI-and-Cloud-Innovation">BNP Paribas and Google Cloud Announce New Partnership on Agentic AI and Cloud Innovation</a> — Google Cloud</li>
</ul>
<p><b>Feather announces robotics funding</b></p>
<p>The developer platform reports a completed pre-seed round.</p>
<ul>
<li><a href="https://www.investegate.co.uk/index.php/announcement/rns/seed-innovations-limited--seed/investee-company-update-feather-robotics-inc-/9792430">Investee Company Update: Feather Robotics Inc.</a> — SEED Innovations Limited via RNS Reach / Investegate</li>
</ul>
<p><b>BigHat announces $75 million Series C</b></p>
<p>Financing supports AI protein design and therapeutic development.</p>
<ul>
<li><a href="https://markets.financialcontent.com/dowtheoryletters/article/bizwire-2026-9-24-bighat-biosciences-announces-75-million-series-c-financing-to-advance-ai-designed-therapeutics-and-leading-agentic-data-platform-for-rapid-protein-design">BigHat Biosciences Announces $75 Million Series C Financing to Advance AI-Designed Therapeutics and Leading Agentic Data Platform for Rapid Protein Design</a> — BigHat Biosciences via Business Wire / FinancialContent</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Mon, 28 Sep 2026 07:05:00 +0200</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/81145c0b/7243dad0.mp3" length="2776545" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>174</itunes:duration>
      <itunes:summary>Google introduces Live Avatar and reports PageBreak security findings; BNP Paribas announces an AI partnership; Feather and BigHat announce funding.</itunes:summary>
      <itunes:subtitle>Google introduces Live Avatar and reports PageBreak security findings; BNP Paribas announces an AI partnership; Feather and BigHat announce funding.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/81145c0b/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>AI Week in Review — Models, Agents, Security and Infrastructure</title>
      <itunes:title>AI Week in Review — Models, Agents, Security and Infrastructure</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">16a8f78c-c126-4c5e-b934-f5628e2abcab</guid>
      <link>https://share.transistor.fm/s/75d31aae</link>
      <description>
        <![CDATA[<p>The September 21–27 weekly review: Claude Opus 5.5 and GPT-6 Sol and Luna; Meta Muse and Google's media tools; early AI-assisted biology findings; agent security research and reported threats; and Akamai and Nscale infrastructure announcements.</p>
<p><b>Stories and sources</b></p>
<p><b>Claude Opus 5.5 launches with lower pricing and external evaluation</b></p>
<p>Anthropic released Opus 5.5, GitHub began its Copilot rollout, and METR published a qualified predeployment assessment.</p>
<ul>
<li><a href="https://www.anthropic.com/claude-opus-5-5">Introducing Claude Opus 5.5</a> — Anthropic</li>
<li><a href="https://metr.org/blog/2026-09-22-claude-opus-5-5/">Summary of METR's predeployment evaluation of Claude Opus 5.5</a> — METR</li>
<li><a href="https://github.blog/changelog/2026-09-22-claude-opus-5-5-is-now-available-in-github-copilot/">Claude Opus 5.5 is now available in GitHub Copilot</a> — GitHub</li>
</ul>
<p><b>OpenAI releases GPT-6 Sol and Luna and updates prompt caching</b></p>
<p>Sol and Luna bring new API prices and Copilot options, alongside caching controls for persistent agents.</p>
<ul>
<li><a href="https://openai.com/index/introducing-gpt-6-sol-and-luna/">Introducing GPT-6 Sol and Luna</a> — OpenAI</li>
<li><a href="https://openai.com/index/better-prompt-caching-for-gpt-6/">Better prompt caching for GPT-6</a> — OpenAI</li>
<li><a href="https://github.blog/changelog/2026-09-22-openais-gpt-6-sol-and-gpt-6-luna-now-available/">OpenAI's GPT-6 Sol and GPT-6 Luna now available</a> — GitHub</li>
</ul>
<p><b>Meta announces Muse integration with AI glasses and expands connectors</b></p>
<p>Connect announcements extend the Muse agent toward glasses, voice interaction and connected services, with hardware shipping dates kept distinct.</p>
<ul>
<li><a href="https://about.fb.com/news/2026/09/the-biggest-news-from-connect-2026/">The Biggest News From Connect 2026</a> — Meta</li>
<li><a href="https://about.fb.com/news/2026/09/introducing-muse-personal-ai-agent/">Introducing Muse: The World's First Personal AI Agent Built for Everyone</a> — Meta</li>
<li><a href="https://about.fb.com/news/2026/09/introducing-ray-ban-meta-audio-glasses-new-styles-plus-muse/">Introducing Ray-Ban Meta Audio and More AI Glasses Styles</a> — Meta</li>
</ul>
<p><b>Google expands generated speech and video tools</b></p>
<p>Gemini 3.8 speech tools add voice design and consent checks, while Google Vids adds free Omni video generation and editing controls.</p>
<ul>
<li><a href="https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-8-text-to-speech/">Gemini 3.8 text-to-speech says hello</a> — Google</li>
<li><a href="https://deepmind.google/models/model-cards/gemini-3-8-audio/">Gemini 3.8 Audio (Live, Live Extended Thinking, Flash TTS, Flash-Lite TTS)</a> — Google DeepMind</li>
<li><a href="https://blog.google/products-and-platforms/products/workspace/gemini-omni-in-google-vids/">Anyone can make stunning HD videos with Gemini Omni in Google Vids</a> — Google</li>
</ul>
<p><b>Anthropic reports AI-assisted discovery of an uncharacterized enzyme system</b></p>
<p>Anthropic described an AI-led sequence analysis followed by human laboratory work; the system's function remains unknown.</p>
<ul>
<li><a href="https://www.anthropic.com/news/claude-discovers-novel-enzyme-system">Claude discovers a novel enzyme system with CRISPR-like repeats</a> — Anthropic</li>
</ul>
<p><b>Darktrace and Cisco Talos publish new AI security research</b></p>
<p>Darktrace announced sandbox research on agent behavior, while Talos released a metadata-based toolkit for investigating AI-related malware.</p>
<ul>
<li><a href="https://www.globenewswire.com/news-release/2026/09/24/3368413/0/en/darktrace-launches-signal-labs-to-research-emerging-risks-of-enterprise-ai-agents.html">Darktrace Launches Signal Labs to Research Emerging Risks of Enterprise AI Agents</a> — Darktrace via GlobeNewswire</li>
<li><a href="https://blog.talosintelligence.com/introducing-cairn-frontier-tracking-for-ai-integrated-malware/">Introducing CAIRN: Frontier tracking for AI-integrated malware</a> — Cisco Talos</li>
</ul>
<p><b>Gambit reports AI-assisted retail intrusions; Talos analyzes an AI-directed malware design</b></p>
<p>An interim Gambit report describes observed compromises. Talos describes a separate design established through static analysis: the public binary was nonfunctional, complete execution was not observed, and deployment is unconfirmed.</p>
<ul>
<li><a href="https://gambit.security/blog-posts/autonomous-ai-agents-online-retailers-25-a-company">Autonomous AI Agents are breaking into hundreds of Online Retailers for $25 a target in an ongoing campaign</a> — Gambit Security</li>
<li><a href="https://blog.talosintelligence.com/the-closed-quorum-inside-the-first-reported-autonomous-ai-c2-implant/">The Closed Quorum: Inside the first reported autonomous AI C2 implant</a> — Cisco Talos</li>
</ul>
<p><b>Akamai announces Anthropic capacity commitment; Nscale announces convertible financing</b></p>
<p>Akamai disclosed a conditional $11.6 billion infrastructure commitment, and Nscale announced financing with a separately timed NVIDIA tranche.</p>
<ul>
<li><a href="https://www.akamai.com/newsroom/press-release/akamai-announces-11-6-billion-multi-year-agreement-with-anthropic-to-support-growing-demand">Akamai Announces $11.6 Billion Multi-year Agreement with Anthropic to Support Growing Demand</a> — Akamai</li>
<li><a href="https://www.sec.gov/Archives/edgar/data/1086222/000119312526401048/d288154d8k.htm">Akamai Technologies, Inc. Form 8-K</a> — Akamai Technologies via U.S. Securities and Exchange Commission</li>
<li><a href="https://www.nscale.com/press-releases/pre-ipo-convertible-financing">Nscale Raises $3.36 Billion in Pre-IPO Convertible Financing</a> — Nscale</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>The September 21–27 weekly review: Claude Opus 5.5 and GPT-6 Sol and Luna; Meta Muse and Google's media tools; early AI-assisted biology findings; agent security research and reported threats; and Akamai and Nscale infrastructure announcements.</p>
<p><b>Stories and sources</b></p>
<p><b>Claude Opus 5.5 launches with lower pricing and external evaluation</b></p>
<p>Anthropic released Opus 5.5, GitHub began its Copilot rollout, and METR published a qualified predeployment assessment.</p>
<ul>
<li><a href="https://www.anthropic.com/claude-opus-5-5">Introducing Claude Opus 5.5</a> — Anthropic</li>
<li><a href="https://metr.org/blog/2026-09-22-claude-opus-5-5/">Summary of METR's predeployment evaluation of Claude Opus 5.5</a> — METR</li>
<li><a href="https://github.blog/changelog/2026-09-22-claude-opus-5-5-is-now-available-in-github-copilot/">Claude Opus 5.5 is now available in GitHub Copilot</a> — GitHub</li>
</ul>
<p><b>OpenAI releases GPT-6 Sol and Luna and updates prompt caching</b></p>
<p>Sol and Luna bring new API prices and Copilot options, alongside caching controls for persistent agents.</p>
<ul>
<li><a href="https://openai.com/index/introducing-gpt-6-sol-and-luna/">Introducing GPT-6 Sol and Luna</a> — OpenAI</li>
<li><a href="https://openai.com/index/better-prompt-caching-for-gpt-6/">Better prompt caching for GPT-6</a> — OpenAI</li>
<li><a href="https://github.blog/changelog/2026-09-22-openais-gpt-6-sol-and-gpt-6-luna-now-available/">OpenAI's GPT-6 Sol and GPT-6 Luna now available</a> — GitHub</li>
</ul>
<p><b>Meta announces Muse integration with AI glasses and expands connectors</b></p>
<p>Connect announcements extend the Muse agent toward glasses, voice interaction and connected services, with hardware shipping dates kept distinct.</p>
<ul>
<li><a href="https://about.fb.com/news/2026/09/the-biggest-news-from-connect-2026/">The Biggest News From Connect 2026</a> — Meta</li>
<li><a href="https://about.fb.com/news/2026/09/introducing-muse-personal-ai-agent/">Introducing Muse: The World's First Personal AI Agent Built for Everyone</a> — Meta</li>
<li><a href="https://about.fb.com/news/2026/09/introducing-ray-ban-meta-audio-glasses-new-styles-plus-muse/">Introducing Ray-Ban Meta Audio and More AI Glasses Styles</a> — Meta</li>
</ul>
<p><b>Google expands generated speech and video tools</b></p>
<p>Gemini 3.8 speech tools add voice design and consent checks, while Google Vids adds free Omni video generation and editing controls.</p>
<ul>
<li><a href="https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-8-text-to-speech/">Gemini 3.8 text-to-speech says hello</a> — Google</li>
<li><a href="https://deepmind.google/models/model-cards/gemini-3-8-audio/">Gemini 3.8 Audio (Live, Live Extended Thinking, Flash TTS, Flash-Lite TTS)</a> — Google DeepMind</li>
<li><a href="https://blog.google/products-and-platforms/products/workspace/gemini-omni-in-google-vids/">Anyone can make stunning HD videos with Gemini Omni in Google Vids</a> — Google</li>
</ul>
<p><b>Anthropic reports AI-assisted discovery of an uncharacterized enzyme system</b></p>
<p>Anthropic described an AI-led sequence analysis followed by human laboratory work; the system's function remains unknown.</p>
<ul>
<li><a href="https://www.anthropic.com/news/claude-discovers-novel-enzyme-system">Claude discovers a novel enzyme system with CRISPR-like repeats</a> — Anthropic</li>
</ul>
<p><b>Darktrace and Cisco Talos publish new AI security research</b></p>
<p>Darktrace announced sandbox research on agent behavior, while Talos released a metadata-based toolkit for investigating AI-related malware.</p>
<ul>
<li><a href="https://www.globenewswire.com/news-release/2026/09/24/3368413/0/en/darktrace-launches-signal-labs-to-research-emerging-risks-of-enterprise-ai-agents.html">Darktrace Launches Signal Labs to Research Emerging Risks of Enterprise AI Agents</a> — Darktrace via GlobeNewswire</li>
<li><a href="https://blog.talosintelligence.com/introducing-cairn-frontier-tracking-for-ai-integrated-malware/">Introducing CAIRN: Frontier tracking for AI-integrated malware</a> — Cisco Talos</li>
</ul>
<p><b>Gambit reports AI-assisted retail intrusions; Talos analyzes an AI-directed malware design</b></p>
<p>An interim Gambit report describes observed compromises. Talos describes a separate design established through static analysis: the public binary was nonfunctional, complete execution was not observed, and deployment is unconfirmed.</p>
<ul>
<li><a href="https://gambit.security/blog-posts/autonomous-ai-agents-online-retailers-25-a-company">Autonomous AI Agents are breaking into hundreds of Online Retailers for $25 a target in an ongoing campaign</a> — Gambit Security</li>
<li><a href="https://blog.talosintelligence.com/the-closed-quorum-inside-the-first-reported-autonomous-ai-c2-implant/">The Closed Quorum: Inside the first reported autonomous AI C2 implant</a> — Cisco Talos</li>
</ul>
<p><b>Akamai announces Anthropic capacity commitment; Nscale announces convertible financing</b></p>
<p>Akamai disclosed a conditional $11.6 billion infrastructure commitment, and Nscale announced financing with a separately timed NVIDIA tranche.</p>
<ul>
<li><a href="https://www.akamai.com/newsroom/press-release/akamai-announces-11-6-billion-multi-year-agreement-with-anthropic-to-support-growing-demand">Akamai Announces $11.6 Billion Multi-year Agreement with Anthropic to Support Growing Demand</a> — Akamai</li>
<li><a href="https://www.sec.gov/Archives/edgar/data/1086222/000119312526401048/d288154d8k.htm">Akamai Technologies, Inc. Form 8-K</a> — Akamai Technologies via U.S. Securities and Exchange Commission</li>
<li><a href="https://www.nscale.com/press-releases/pre-ipo-convertible-financing">Nscale Raises $3.36 Billion in Pre-IPO Convertible Financing</a> — Nscale</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Sun, 27 Sep 2026 21:29:12 +0200</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/75d31aae/e67c107d.mp3" length="20276496" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>1268</itunes:duration>
      <itunes:summary>The September 21–27 weekly review: Claude Opus 5.5 and GPT-6 Sol and Luna; Meta Muse and Google's media tools; early AI-assisted biology findings; agent security research and reported threats; and Akamai and Nscale infrastructure announcements.</itunes:summary>
      <itunes:subtitle>The September 21–27 weekly review: Claude Opus 5.5 and GPT-6 Sol and Luna; Meta Muse and Google's media tools; early AI-assisted biology findings; agent security research and reported threats; and Akamai and Nscale infrastructure announcements.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/75d31aae/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>Muse expansion and new developer bots</title>
      <itunes:title>Muse expansion and new developer bots</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">01ed0ab1-71c6-44ba-975f-2b60f979be46</guid>
      <link>https://share.transistor.fm/s/8e12ae4f</link>
      <description>
        <![CDATA[<p>Meta’s Muse plans, Epoch’s assembly benchmark, Cursor’s deployment and security bots, Ando’s funding, and NetApp’s proposed acquisition.</p>
<p><b>Stories and sources</b></p>
<p><b>Meta announces planned Muse expansion</b></p>
<p>Muse is planned for AI glasses, with additional connectors and its own email address.</p>
<ul>
<li><a href="https://about.fb.com/news/2026/09/the-biggest-news-from-connect-2026/">The Biggest News From Connect 2026</a> — Meta</li>
</ul>
<p><b>Epoch reports furniture-assembly benchmark results</b></p>
<p>Astra leads the reported test; its limited scope constrains generalization.</p>
<ul>
<li><a href="https://epoch.ai/publications/furniture-assembly">AI has improved significantly at reasoning about IKEA furniture assembly</a> — Epoch AI</li>
</ul>
<p><b>Cursor launches Rollouts and Security Review</b></p>
<p>The bots monitor deployments and review security; Rollouts cannot independently merge or roll back.</p>
<ul>
<li><a href="https://cursor.com/changelog/rollouts-and-security-reviewer">Rollouts and Security Review</a> — Cursor</li>
</ul>
<p><b>Ando announces messaging platform and seed funding</b></p>
<p>Ando reports $20 million in funding for messaging with AI agents.</p>
<ul>
<li><a href="https://www.globenewswire.com/news-release/2026/9/24/3368344/0/en/ando-launches-agent-native-messaging-platform-announces-20-million-seed.html">Ando Launches Agent-Native Messaging Platform, Announces $20 Million Seed</a> — Ando via GlobeNewswire</li>
</ul>
<p><b>NetApp announces intent to acquire PEAK:AIO</b></p>
<p>The planned AI storage acquisition remains subject to closing conditions and regulatory approvals.</p>
<ul>
<li><a href="https://investors.netapp.com/news/news-details/2026/NetApp-Announces-Intent-to-Acquire-PEAKAIO-to-Advance-Scalable-AI-Infrastructure-Architecture/default.aspx">NetApp Announces Intent to Acquire PEAK:AIO to Advance Scalable AI Infrastructure Architecture</a> — NetApp</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>Meta’s Muse plans, Epoch’s assembly benchmark, Cursor’s deployment and security bots, Ando’s funding, and NetApp’s proposed acquisition.</p>
<p><b>Stories and sources</b></p>
<p><b>Meta announces planned Muse expansion</b></p>
<p>Muse is planned for AI glasses, with additional connectors and its own email address.</p>
<ul>
<li><a href="https://about.fb.com/news/2026/09/the-biggest-news-from-connect-2026/">The Biggest News From Connect 2026</a> — Meta</li>
</ul>
<p><b>Epoch reports furniture-assembly benchmark results</b></p>
<p>Astra leads the reported test; its limited scope constrains generalization.</p>
<ul>
<li><a href="https://epoch.ai/publications/furniture-assembly">AI has improved significantly at reasoning about IKEA furniture assembly</a> — Epoch AI</li>
</ul>
<p><b>Cursor launches Rollouts and Security Review</b></p>
<p>The bots monitor deployments and review security; Rollouts cannot independently merge or roll back.</p>
<ul>
<li><a href="https://cursor.com/changelog/rollouts-and-security-reviewer">Rollouts and Security Review</a> — Cursor</li>
</ul>
<p><b>Ando announces messaging platform and seed funding</b></p>
<p>Ando reports $20 million in funding for messaging with AI agents.</p>
<ul>
<li><a href="https://www.globenewswire.com/news-release/2026/9/24/3368344/0/en/ando-launches-agent-native-messaging-platform-announces-20-million-seed.html">Ando Launches Agent-Native Messaging Platform, Announces $20 Million Seed</a> — Ando via GlobeNewswire</li>
</ul>
<p><b>NetApp announces intent to acquire PEAK:AIO</b></p>
<p>The planned AI storage acquisition remains subject to closing conditions and regulatory approvals.</p>
<ul>
<li><a href="https://investors.netapp.com/news/news-details/2026/NetApp-Announces-Intent-to-Acquire-PEAKAIO-to-Advance-Scalable-AI-Infrastructure-Architecture/default.aspx">NetApp Announces Intent to Acquire PEAK:AIO to Advance Scalable AI Infrastructure Architecture</a> — NetApp</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Sun, 27 Sep 2026 07:05:25 +0200</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/8e12ae4f/337e1c2b.mp3" length="2947073" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>185</itunes:duration>
      <itunes:summary>Meta’s Muse plans, Epoch’s assembly benchmark, Cursor’s deployment and security bots, Ando’s funding, and NetApp’s proposed acquisition.</itunes:summary>
      <itunes:subtitle>Meta’s Muse plans, Epoch’s assembly benchmark, Cursor’s deployment and security bots, Ando’s funding, and NetApp’s proposed acquisition.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/8e12ae4f/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>Vision-model acceleration, agent security and AI infrastructure deals</title>
      <itunes:title>Vision-model acceleration, agent security and AI infrastructure deals</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">b89e7b12-4666-47f1-b37d-44bce2172ef7</guid>
      <link>https://share.transistor.fm/s/dac0acac</link>
      <description>
        <![CDATA[<p>Liquid AI's experimental accelerator, Anthropic's biology findings, Darktrace's agent tests, Group-IB's malware report, and infrastructure announcements from Akamai and Nscale.</p>
<p><b>Stories and sources</b></p>
<p><b>Liquid AI releases experimental vision-model accelerator</b></p>
<p>DSpark adds speculative decoding, with hardware-specific gains and unchanged image-processing stages.</p>
<ul>
<li><a href="https://www.liquid.ai/blog/lfm2-5-vl-dspark">LFM2.5-VL-DSpark: Accelerating vision-language models on edge and beyond</a> — Liquid AI</li>
<li><a href="https://huggingface.co/LiquidAI/LFM2.5-VL-3B-DSpark">LiquidAI/LFM2.5-VL-3B-DSpark</a> — Liquid AI</li>
</ul>
<p><b>Anthropic reports AI-assisted enzyme-system discovery</b></p>
<p>Early findings identify associated repeats and a protein; the system's function remains unknown.</p>
<ul>
<li><a href="https://www.anthropic.com/news/claude-discovers-novel-enzyme-system">Claude discovers a novel enzyme system with CRISPR-like repeats</a> — Anthropic</li>
</ul>
<p><b>Darktrace reports controlled agent-security experiments</b></p>
<p>Studies examine tampered histories and agent behavior under impossible task conditions.</p>
<ul>
<li><a href="https://www.darktrace.com/news/darktrace-launches-signal-labs-to-research-emerging-risks-of-enterprise-ai-agents">Darktrace Launches Signal Labs to Research Emerging Risks of Enterprise AI Agents</a> — Darktrace</li>
<li><a href="https://www.darktrace.com/blog/hijacking-agentic-harnesses-to-attack-an-organization">Agent Hijacks: Hijacking Agentic Harnesses to Attack an Organization</a> — Darktrace</li>
</ul>
<p><b>Group-IB reports AI development artifacts in RemControl</b></p>
<p>Researchers attribute development assistance to AI; an operator connection remains unconfirmed.</p>
<ul>
<li><a href="https://www.group-ib.com/blog/remcontrol-android-banking-trojan/">RemControl: AI Built the Overlays. Victims Lose their PINs</a> — Group-IB</li>
</ul>
<p><b>Akamai announces $11.6 billion Anthropic commitment</b></p>
<p>Seven-year project terms include service conditions; additional expansion and revenue figures are prospective.</p>
<ul>
<li><a href="https://www.akamai.com/newsroom/press-release/akamai-announces-11-6-billion-multi-year-agreement-with-anthropic-to-support-growing-demand">Akamai Announces $11.6 Billion Multi-year Agreement with Anthropic to Support Growing Demand</a> — Akamai</li>
<li><a href="https://www.ir.akamai.com/static-files/db805670-271e-406b-91a5-e0b4ed226b5a">Akamai Technologies Form 8-K</a> — Akamai</li>
<li><a href="https://www.ir.akamai.com/static-files/59654fd2-bd7c-4090-9ded-737b72ab2bb7">Expanded Strategic Relationship with Anthropic</a> — Akamai</li>
</ul>
<p><b>Nscale announces convertible financing with staged funding</b></p>
<p>The $3.36 billion structure includes a future Nvidia tranche.</p>
<ul>
<li><a href="https://www.nscale.com/press-releases/pre-ipo-convertible-financing">Nscale Raises $3.36 Billion in Pre-IPO Convertible Financing</a> — Nscale</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>Liquid AI's experimental accelerator, Anthropic's biology findings, Darktrace's agent tests, Group-IB's malware report, and infrastructure announcements from Akamai and Nscale.</p>
<p><b>Stories and sources</b></p>
<p><b>Liquid AI releases experimental vision-model accelerator</b></p>
<p>DSpark adds speculative decoding, with hardware-specific gains and unchanged image-processing stages.</p>
<ul>
<li><a href="https://www.liquid.ai/blog/lfm2-5-vl-dspark">LFM2.5-VL-DSpark: Accelerating vision-language models on edge and beyond</a> — Liquid AI</li>
<li><a href="https://huggingface.co/LiquidAI/LFM2.5-VL-3B-DSpark">LiquidAI/LFM2.5-VL-3B-DSpark</a> — Liquid AI</li>
</ul>
<p><b>Anthropic reports AI-assisted enzyme-system discovery</b></p>
<p>Early findings identify associated repeats and a protein; the system's function remains unknown.</p>
<ul>
<li><a href="https://www.anthropic.com/news/claude-discovers-novel-enzyme-system">Claude discovers a novel enzyme system with CRISPR-like repeats</a> — Anthropic</li>
</ul>
<p><b>Darktrace reports controlled agent-security experiments</b></p>
<p>Studies examine tampered histories and agent behavior under impossible task conditions.</p>
<ul>
<li><a href="https://www.darktrace.com/news/darktrace-launches-signal-labs-to-research-emerging-risks-of-enterprise-ai-agents">Darktrace Launches Signal Labs to Research Emerging Risks of Enterprise AI Agents</a> — Darktrace</li>
<li><a href="https://www.darktrace.com/blog/hijacking-agentic-harnesses-to-attack-an-organization">Agent Hijacks: Hijacking Agentic Harnesses to Attack an Organization</a> — Darktrace</li>
</ul>
<p><b>Group-IB reports AI development artifacts in RemControl</b></p>
<p>Researchers attribute development assistance to AI; an operator connection remains unconfirmed.</p>
<ul>
<li><a href="https://www.group-ib.com/blog/remcontrol-android-banking-trojan/">RemControl: AI Built the Overlays. Victims Lose their PINs</a> — Group-IB</li>
</ul>
<p><b>Akamai announces $11.6 billion Anthropic commitment</b></p>
<p>Seven-year project terms include service conditions; additional expansion and revenue figures are prospective.</p>
<ul>
<li><a href="https://www.akamai.com/newsroom/press-release/akamai-announces-11-6-billion-multi-year-agreement-with-anthropic-to-support-growing-demand">Akamai Announces $11.6 Billion Multi-year Agreement with Anthropic to Support Growing Demand</a> — Akamai</li>
<li><a href="https://www.ir.akamai.com/static-files/db805670-271e-406b-91a5-e0b4ed226b5a">Akamai Technologies Form 8-K</a> — Akamai</li>
<li><a href="https://www.ir.akamai.com/static-files/59654fd2-bd7c-4090-9ded-737b72ab2bb7">Expanded Strategic Relationship with Anthropic</a> — Akamai</li>
</ul>
<p><b>Nscale announces convertible financing with staged funding</b></p>
<p>The $3.36 billion structure includes a future Nvidia tranche.</p>
<ul>
<li><a href="https://www.nscale.com/press-releases/pre-ipo-convertible-financing">Nscale Raises $3.36 Billion in Pre-IPO Convertible Financing</a> — Nscale</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Sat, 26 Sep 2026 07:08:37 +0200</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/dac0acac/2432a4f8.mp3" length="5782926" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>362</itunes:duration>
      <itunes:summary>Liquid AI's experimental accelerator, Anthropic's biology findings, Darktrace's agent tests, Group-IB's malware report, and infrastructure announcements from Akamai and Nscale.</itunes:summary>
      <itunes:subtitle>Liquid AI's experimental accelerator, Anthropic's biology findings, Darktrace's agent tests, Group-IB's malware report, and infrastructure announcements from Akamai and Nscale.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/dac0acac/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>Speech models, AI malware research and Row Zero</title>
      <itunes:title>Speech models, AI malware research and Row Zero</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">8ccfcaab-7cb8-43a8-b84d-8409131a6418</guid>
      <link>https://share.transistor.fm/s/91c21b7c</link>
      <description>
        <![CDATA[<p>Google and Qwen voice developments, AI video generation, Talos malware research, Gambit's retail intrusion findings, and Databricks' acquisition.</p>
<p><b>Stories and sources</b></p>
<p><b>Google rolls out Gemini 3.8 speech generation models</b></p>
<p>Developer rollout begins with restrictions on voice replication.</p>
<ul>
<li><a href="https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-8-text-to-speech/">Gemini 3.8 text-to-speech says hello</a> — Google</li>
</ul>
<p><b>Qwen reports voice-agent evaluation gains</b></p>
<p>The reported improvement uses a modified tau-Voice protocol.</p>
<ul>
<li><a href="https://arxiv.org/abs/2609.25176">Qwen-Audio-3.1-Realtime: Towards Reliable Agentic Voice Interaction</a> — arXiv</li>
<li><a href="https://arxiv.org/html/2609.25176v2">Qwen-Audio-3.1-Realtime: Towards Reliable Agentic Voice Interaction</a> — arXiv</li>
</ul>
<p><b>Google expands free AI video generation in Vids</b></p>
<p>Omni 1.1 Flash brings generation and editing controls.</p>
<ul>
<li><a href="https://blog.google/products-and-platforms/products/workspace/gemini-omni-in-google-vids/">Anyone can make stunning HD videos with Gemini Omni in Google Vids</a> — Google</li>
</ul>
<p><b>Talos releases CAIRN malware research toolkit</b></p>
<p>Metadata-based detection produces leads requiring validation.</p>
<ul>
<li><a href="https://blog.talosintelligence.com/introducing-cairn-frontier-tracking-for-ai-integrated-malware/">Introducing CAIRN: Frontier tracking for AI-integrated malware</a> — Cisco Talos</li>
</ul>
<p><b>Gambit reports AI-driven retail intrusions</b></p>
<p>Interim findings include stolen payment-card records.</p>
<ul>
<li><a href="https://gambit.security/blog-posts/autonomous-ai-agents-online-retailers-25-a-company">Autonomous AI Agents are breaking into hundreds of Online Retailers for $25 a target in an ongoing campaign</a> — Gambit Security</li>
</ul>
<p><b>Databricks acquires Row Zero</b></p>
<p>Native spreadsheet integration into Genie is planned.</p>
<ul>
<li><a href="https://www.databricks.com/company/newsroom/press-releases/databricks-acquires-row-zero-bringing-live-governed-spreadsheets">Databricks Acquires Row Zero, Bringing Live, Governed Spreadsheets to Genie</a> — Databricks</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>Google and Qwen voice developments, AI video generation, Talos malware research, Gambit's retail intrusion findings, and Databricks' acquisition.</p>
<p><b>Stories and sources</b></p>
<p><b>Google rolls out Gemini 3.8 speech generation models</b></p>
<p>Developer rollout begins with restrictions on voice replication.</p>
<ul>
<li><a href="https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-8-text-to-speech/">Gemini 3.8 text-to-speech says hello</a> — Google</li>
</ul>
<p><b>Qwen reports voice-agent evaluation gains</b></p>
<p>The reported improvement uses a modified tau-Voice protocol.</p>
<ul>
<li><a href="https://arxiv.org/abs/2609.25176">Qwen-Audio-3.1-Realtime: Towards Reliable Agentic Voice Interaction</a> — arXiv</li>
<li><a href="https://arxiv.org/html/2609.25176v2">Qwen-Audio-3.1-Realtime: Towards Reliable Agentic Voice Interaction</a> — arXiv</li>
</ul>
<p><b>Google expands free AI video generation in Vids</b></p>
<p>Omni 1.1 Flash brings generation and editing controls.</p>
<ul>
<li><a href="https://blog.google/products-and-platforms/products/workspace/gemini-omni-in-google-vids/">Anyone can make stunning HD videos with Gemini Omni in Google Vids</a> — Google</li>
</ul>
<p><b>Talos releases CAIRN malware research toolkit</b></p>
<p>Metadata-based detection produces leads requiring validation.</p>
<ul>
<li><a href="https://blog.talosintelligence.com/introducing-cairn-frontier-tracking-for-ai-integrated-malware/">Introducing CAIRN: Frontier tracking for AI-integrated malware</a> — Cisco Talos</li>
</ul>
<p><b>Gambit reports AI-driven retail intrusions</b></p>
<p>Interim findings include stolen payment-card records.</p>
<ul>
<li><a href="https://gambit.security/blog-posts/autonomous-ai-agents-online-retailers-25-a-company">Autonomous AI Agents are breaking into hundreds of Online Retailers for $25 a target in an ongoing campaign</a> — Gambit Security</li>
</ul>
<p><b>Databricks acquires Row Zero</b></p>
<p>Native spreadsheet integration into Genie is planned.</p>
<ul>
<li><a href="https://www.databricks.com/company/newsroom/press-releases/databricks-acquires-row-zero-bringing-live-governed-spreadsheets">Databricks Acquires Row Zero, Bringing Live, Governed Spreadsheets to Genie</a> — Databricks</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Fri, 25 Sep 2026 07:06:41 +0200</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/91c21b7c/b5d739df.mp3" length="3486240" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>218</itunes:duration>
      <itunes:summary>Google and Qwen voice developments, AI video generation, Talos malware research, Gambit's retail intrusion findings, and Databricks' acquisition.</itunes:summary>
      <itunes:subtitle>Google and Qwen voice developments, AI video generation, Talos malware research, Gambit's retail intrusion findings, and Databricks' acquisition.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/91c21b7c/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>New models from Anthropic, OpenAI and Xiaomi</title>
      <itunes:title>New models from Anthropic, OpenAI and Xiaomi</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">9fac103e-32c4-4bf6-abbc-cc9e76c2a158</guid>
      <link>https://share.transistor.fm/s/d5002a8e</link>
      <description>
        <![CDATA[<p>Three model launches, Wayve’s Mercedes-Benz production agreement, Unit 42’s continuous AI security service, and Snorkel’s funding round.</p>
<p><b>Stories and sources</b></p>
<p><b>Anthropic releases Claude Opus 5.5</b></p>
<p>Opus 5.5 launches with revised pricing and a gradual GitHub Copilot rollout.</p>
<ul>
<li><a href="https://www.anthropic.com/claude-opus-5-5">Introducing Claude Opus 5.5</a> — Anthropic</li>
<li><a href="https://github.blog/changelog/2026-09-22-claude-opus-5-5-is-now-available-in-github-copilot/">Claude Opus 5.5 is now available in GitHub Copilot</a> — GitHub</li>
</ul>
<p><b>OpenAI launches GPT-6 Sol and Luna</b></p>
<p>The models reach the API, ChatGPT Work and Codex, with separate Copilot plan eligibility.</p>
<ul>
<li><a href="https://openai.com/index/introducing-gpt-6-sol-and-luna/">Introducing GPT-6 Sol and Luna</a> — OpenAI</li>
<li><a href="https://community.openai.com/t/announcing-gpt-6-sol-and-gpt-6-luna-in-the-api-codex-and-chatgpt/1399925">Announcing GPT-6 Sol and GPT-6 Luna in the API, Codex and ChatGPT</a> — OpenAI Developer Community</li>
<li><a href="https://github.blog/changelog/2026-09-22-openais-gpt-6-sol-and-gpt-6-luna-now-available/">OpenAI’s GPT-6 Sol and GPT-6 Luna now available</a> — GitHub</li>
</ul>
<p><b>Xiaomi releases MiMo 2.6 and updates MiMo Code</b></p>
<p>Xiaomi reports training gains and adds tool sequencing and multimodal file reading.</p>
<ul>
<li><a href="https://mimo.mi.com/docs/en-US/news/latest/v2-6">MiMo-V2.6: Scaling Up Reinforcement Learning for Self-Improvement</a> — Xiaomi</li>
<li><a href="https://github.com/XiaomiMiMo/MiMo-Code/releases">Releases · XiaomiMiMo/MiMo-Code</a> — XiaomiMiMo on GitHub</li>
</ul>
<p><b>Wayve and Mercedes-Benz sign production agreement</b></p>
<p>Driving assistance is planned for future vehicles within two years.</p>
<ul>
<li><a href="https://wayve.ai/press/wayve-mercedes-benz-announce-production-agreement/">Wayve and Mercedes-Benz Partner to Bring the Wayve AI Driver to Future Mercedes-Benz Vehicles</a> — Wayve</li>
</ul>
<p><b>Unit 42 launches continuous AI security testing service</b></p>
<p>Palo Alto Networks announces worldwide annual subscriptions.</p>
<ul>
<li><a href="https://investors.paloaltonetworks.com/node/20856/pdf">Palo Alto Networks Delivers Anthropic's Mythos and OpenAI's GPT-5.6 to Customers with Unit 42 Continuous Frontier AI Defense</a> — Palo Alto Networks</li>
</ul>
<p><b>Snorkel announces $350 million Series E</b></p>
<p>The AI data provider reports a $3.5 billion valuation and plans expanded capacity.</p>
<ul>
<li><a href="https://www.prnewswire.com/news-releases/snorkel-ai-raises-350m-to-scale-the-data-factory-for-frontier-ai-302886796.html">Snorkel AI Raises $350M to Scale the Data Factory for Frontier AI</a> — Snorkel AI via PR Newswire</li>
<li><a href="https://snorkel.ai/blog/data-2-0-and-the-research-era-of-ai-data/">Data 2.0 and the research era of AI data</a> — Snorkel AI</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>Three model launches, Wayve’s Mercedes-Benz production agreement, Unit 42’s continuous AI security service, and Snorkel’s funding round.</p>
<p><b>Stories and sources</b></p>
<p><b>Anthropic releases Claude Opus 5.5</b></p>
<p>Opus 5.5 launches with revised pricing and a gradual GitHub Copilot rollout.</p>
<ul>
<li><a href="https://www.anthropic.com/claude-opus-5-5">Introducing Claude Opus 5.5</a> — Anthropic</li>
<li><a href="https://github.blog/changelog/2026-09-22-claude-opus-5-5-is-now-available-in-github-copilot/">Claude Opus 5.5 is now available in GitHub Copilot</a> — GitHub</li>
</ul>
<p><b>OpenAI launches GPT-6 Sol and Luna</b></p>
<p>The models reach the API, ChatGPT Work and Codex, with separate Copilot plan eligibility.</p>
<ul>
<li><a href="https://openai.com/index/introducing-gpt-6-sol-and-luna/">Introducing GPT-6 Sol and Luna</a> — OpenAI</li>
<li><a href="https://community.openai.com/t/announcing-gpt-6-sol-and-gpt-6-luna-in-the-api-codex-and-chatgpt/1399925">Announcing GPT-6 Sol and GPT-6 Luna in the API, Codex and ChatGPT</a> — OpenAI Developer Community</li>
<li><a href="https://github.blog/changelog/2026-09-22-openais-gpt-6-sol-and-gpt-6-luna-now-available/">OpenAI’s GPT-6 Sol and GPT-6 Luna now available</a> — GitHub</li>
</ul>
<p><b>Xiaomi releases MiMo 2.6 and updates MiMo Code</b></p>
<p>Xiaomi reports training gains and adds tool sequencing and multimodal file reading.</p>
<ul>
<li><a href="https://mimo.mi.com/docs/en-US/news/latest/v2-6">MiMo-V2.6: Scaling Up Reinforcement Learning for Self-Improvement</a> — Xiaomi</li>
<li><a href="https://github.com/XiaomiMiMo/MiMo-Code/releases">Releases · XiaomiMiMo/MiMo-Code</a> — XiaomiMiMo on GitHub</li>
</ul>
<p><b>Wayve and Mercedes-Benz sign production agreement</b></p>
<p>Driving assistance is planned for future vehicles within two years.</p>
<ul>
<li><a href="https://wayve.ai/press/wayve-mercedes-benz-announce-production-agreement/">Wayve and Mercedes-Benz Partner to Bring the Wayve AI Driver to Future Mercedes-Benz Vehicles</a> — Wayve</li>
</ul>
<p><b>Unit 42 launches continuous AI security testing service</b></p>
<p>Palo Alto Networks announces worldwide annual subscriptions.</p>
<ul>
<li><a href="https://investors.paloaltonetworks.com/node/20856/pdf">Palo Alto Networks Delivers Anthropic's Mythos and OpenAI's GPT-5.6 to Customers with Unit 42 Continuous Frontier AI Defense</a> — Palo Alto Networks</li>
</ul>
<p><b>Snorkel announces $350 million Series E</b></p>
<p>The AI data provider reports a $3.5 billion valuation and plans expanded capacity.</p>
<ul>
<li><a href="https://www.prnewswire.com/news-releases/snorkel-ai-raises-350m-to-scale-the-data-factory-for-frontier-ai-302886796.html">Snorkel AI Raises $350M to Scale the Data Factory for Frontier AI</a> — Snorkel AI via PR Newswire</li>
<li><a href="https://snorkel.ai/blog/data-2-0-and-the-research-era-of-ai-data/">Data 2.0 and the research era of AI data</a> — Snorkel AI</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Wed, 23 Sep 2026 07:10:07 +0200</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/d5002a8e/47791c59.mp3" length="5536330" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>346</itunes:duration>
      <itunes:summary>Three model launches, Wayve’s Mercedes-Benz production agreement, Unit 42’s continuous AI security service, and Snorkel’s funding round.</itunes:summary>
      <itunes:subtitle>Three model launches, Wayve’s Mercedes-Benz production agreement, Unit 42’s continuous AI security service, and Snorkel’s funding round.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/d5002a8e/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>Grok 4.7, agent updates and AI code security</title>
      <itunes:title>Grok 4.7, agent updates and AI code security</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">895edcea-f51e-4671-81eb-06b4ec03a88a</guid>
      <link>https://share.transistor.fm/s/be9ba334</link>
      <description>
        <![CDATA[<p>Grok coding and transcription releases, OpenClaw updates, HackerOne’s planned Mythos integration, and Lovable’s acquisition of Sutro.</p>
<p><b>Stories and sources</b></p>
<p><b>SpaceXAI releases Grok 4.7</b></p>
<p>The coding and knowledge-work model is available through the API, Cursor and Grok Build, with separate pricing and access conditions for its Fast variant.</p>
<ul>
<li><a href="https://x.ai/news/grok-4-7">Introducing Grok 4.7</a> — SpaceXAI</li>
<li><a href="https://docs.x.ai/developers/grok-4-7">Grok 4.7</a> — SpaceXAI</li>
<li><a href="https://docs.x.ai/developers/release-notes">Release Notes</a> — SpaceXAI</li>
</ul>
<p><b>SpaceXAI details Grok Voice Transcribe 2.0</b></p>
<p>The company reports multilingual transcription improvements; its documentation disagrees about the default model.</p>
<ul>
<li><a href="https://x.ai/news/grok-voice-transcribe-2">Introducing Grok Voice Transcribe 2.0</a> — SpaceXAI</li>
<li><a href="https://docs.x.ai/developers/model-capabilities/audio/speech-to-text">Speech to Text</a> — SpaceXAI</li>
<li><a href="https://docs.x.ai/developers/release-notes">Release Notes</a> — SpaceXAI</li>
</ul>
<p><b>OpenClaw adds update validation and agent setup features</b></p>
<p>Version 2026.9.5 adds candidate-update checks, specialist setup and plugin installation without Gateway restarts.</p>
<ul>
<li><a href="https://github.com/openclaw/openclaw/releases/tag/v2026.9.5">Release openclaw 2026.9.5</a> — OpenClaw / GitHub</li>
<li><a href="https://docs.openclaw.ai/releases/2026.9.5">v2026.9.5</a> — OpenClaw</li>
</ul>
<p><b>HackerOne announces planned Mythos integration</b></p>
<p>HackerOne plans Mythos support for repository audits and code reviews, with validation controls.</p>
<ul>
<li><a href="https://www.hackerone.com/press-release/hackerone-bring-anthropics-mythos-exposure-management-lifecycle-discovery-verified">HackerOne to Bring Anthropic’s Mythos Into Exposure Management Lifecycle, from Discovery to Verified Fix</a> — HackerOne</li>
<li><a href="https://www.hackerone.com/blog/project-glasswing-frontier-model-h1">Project Glasswing: What We Learned Running a Frontier Model on Our Own Codebase</a> — HackerOne</li>
</ul>
<p><b>Lovable announces Sutro acquisition</b></p>
<p>Sutro’s founder and three engineers join Lovable; the announcement does not disclose a price.</p>
<ul>
<li><a href="https://lovable.dev/blog/lovable-acquires-sutro-to-make-software-easier-to-explain-and-easier-to-trust">Lovable acquires Sutro to make software easier to explain and easier to trust</a> — Lovable</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>Grok coding and transcription releases, OpenClaw updates, HackerOne’s planned Mythos integration, and Lovable’s acquisition of Sutro.</p>
<p><b>Stories and sources</b></p>
<p><b>SpaceXAI releases Grok 4.7</b></p>
<p>The coding and knowledge-work model is available through the API, Cursor and Grok Build, with separate pricing and access conditions for its Fast variant.</p>
<ul>
<li><a href="https://x.ai/news/grok-4-7">Introducing Grok 4.7</a> — SpaceXAI</li>
<li><a href="https://docs.x.ai/developers/grok-4-7">Grok 4.7</a> — SpaceXAI</li>
<li><a href="https://docs.x.ai/developers/release-notes">Release Notes</a> — SpaceXAI</li>
</ul>
<p><b>SpaceXAI details Grok Voice Transcribe 2.0</b></p>
<p>The company reports multilingual transcription improvements; its documentation disagrees about the default model.</p>
<ul>
<li><a href="https://x.ai/news/grok-voice-transcribe-2">Introducing Grok Voice Transcribe 2.0</a> — SpaceXAI</li>
<li><a href="https://docs.x.ai/developers/model-capabilities/audio/speech-to-text">Speech to Text</a> — SpaceXAI</li>
<li><a href="https://docs.x.ai/developers/release-notes">Release Notes</a> — SpaceXAI</li>
</ul>
<p><b>OpenClaw adds update validation and agent setup features</b></p>
<p>Version 2026.9.5 adds candidate-update checks, specialist setup and plugin installation without Gateway restarts.</p>
<ul>
<li><a href="https://github.com/openclaw/openclaw/releases/tag/v2026.9.5">Release openclaw 2026.9.5</a> — OpenClaw / GitHub</li>
<li><a href="https://docs.openclaw.ai/releases/2026.9.5">v2026.9.5</a> — OpenClaw</li>
</ul>
<p><b>HackerOne announces planned Mythos integration</b></p>
<p>HackerOne plans Mythos support for repository audits and code reviews, with validation controls.</p>
<ul>
<li><a href="https://www.hackerone.com/press-release/hackerone-bring-anthropics-mythos-exposure-management-lifecycle-discovery-verified">HackerOne to Bring Anthropic’s Mythos Into Exposure Management Lifecycle, from Discovery to Verified Fix</a> — HackerOne</li>
<li><a href="https://www.hackerone.com/blog/project-glasswing-frontier-model-h1">Project Glasswing: What We Learned Running a Frontier Model on Our Own Codebase</a> — HackerOne</li>
</ul>
<p><b>Lovable announces Sutro acquisition</b></p>
<p>Sutro’s founder and three engineers join Lovable; the announcement does not disclose a price.</p>
<ul>
<li><a href="https://lovable.dev/blog/lovable-acquires-sutro-to-make-software-easier-to-explain-and-easier-to-trust">Lovable acquires Sutro to make software easier to explain and easier to trust</a> — Lovable</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Tue, 22 Sep 2026 07:08:37 +0200</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/be9ba334/467a8eb9.mp3" length="4508568" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>282</itunes:duration>
      <itunes:summary>Grok coding and transcription releases, OpenClaw updates, HackerOne’s planned Mythos integration, and Lovable’s acquisition of Sutro.</itunes:summary>
      <itunes:subtitle>Grok coding and transcription releases, OpenClaw updates, HackerOne’s planned Mythos integration, and Lovable’s acquisition of Sutro.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/be9ba334/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>AI proof research, agent security and CoreWeave financing</title>
      <itunes:title>AI proof research, agent security and CoreWeave financing</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">5335570e-fc33-45e0-a8ed-87c205cbb52f</guid>
      <link>https://share.transistor.fm/s/02989bde</link>
      <description>
        <![CDATA[<p>FormalFlow’s supervised proof work, Mila and Mozilla’s local-AI initiative, Microsoft’s agent security changes, Talos’s Qilin findings and CoreWeave’s enlarged debt offering.</p>
<p><b>Stories and sources</b></p>
<p><b>FormalFlow paper reports supervised AI proof formalization</b></p>
<p>Researchers report a machine-checked component theorem, with human oversight of mathematical meaning.</p>
<ul>
<li><a href="https://arxiv.org/abs/2609.19814">Long-horizon autoformalization of a core theorem underlying MIP* = RE</a> — arXiv</li>
<li><a href="https://arxiv.org/html/2609.19814v1">Long-horizon autoformalization of a core theorem underlying MIP*=RE</a> — arXiv</li>
</ul>
<p><b>Mila and Mozilla announce local-AI foundation initiative</b></p>
<p>The partners plan open interfaces and locally installable implementations, with initial working implementations expected within six months.</p>
<ul>
<li><a href="https://mila.quebec/en/news/mila-and-mozilla-announce-new-initiative-to-build-trustworthy-open-source-ai-for-everyone">Mila and Mozilla Announce New Initiative To Build Trustworthy Open Source AI For Everyone</a> — Mila</li>
<li><a href="https://hypertec.com/news/mila-and-mozilla-announce-new-initiative-to-build-trustworthy-open-source-ai-for-everyone">Mila and Mozilla Announce New Initiative To Build Trustworthy Open Source AI For Everyone</a> — Hypertec</li>
</ul>
<p><b>Microsoft Agent Framework updates MCP security handling</b></p>
<p>Python 1.19.0 changes session scoping and archive handling.</p>
<ul>
<li><a href="https://github.com/microsoft/agent-framework/releases/tag/python-1.19.0">Release python-1.19.0 · microsoft/agent-framework</a> — Microsoft / GitHub</li>
</ul>
<p><b>Talos reports indications of AI-generated Qilin scripts</b></p>
<p>Talos gives a qualified assessment of possible AI-assisted code generation.</p>
<ul>
<li><a href="https://blog.talosintelligence.com/ransomware-incidents-in-japan-in-the-first-half-of-2026/">Ransomware incidents in Japan in the first half of 2026: Investigation of The Gentlemen’s infrastructure and evidence of Qilin's AI use</a> — Cisco Talos</li>
</ul>
<p><b>CoreWeave prices enlarged $3.7 billion debt offering</b></p>
<p>The AI cloud provider scheduled settlement for September 22, subject to closing conditions.</p>
<ul>
<li><a href="https://www.sec.gov/Archives/edgar/data/1769628/000176962826000429/ex9911.htm">CoreWeave Announces Proposed $3.0 Billion Convertible Senior Notes Offering</a> — CoreWeave / U.S. Securities and Exchange Commission</li>
<li><a href="https://investor.wedbush.com/wedbush/article/bizwire-2026-9-18-coreweave-prices-upsized-37-billion-convertible-senior-notes-offering">CoreWeave Prices Upsized $3.7 Billion Convertible Senior Notes Offering</a> — CoreWeave via Business Wire / Wedbush</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>FormalFlow’s supervised proof work, Mila and Mozilla’s local-AI initiative, Microsoft’s agent security changes, Talos’s Qilin findings and CoreWeave’s enlarged debt offering.</p>
<p><b>Stories and sources</b></p>
<p><b>FormalFlow paper reports supervised AI proof formalization</b></p>
<p>Researchers report a machine-checked component theorem, with human oversight of mathematical meaning.</p>
<ul>
<li><a href="https://arxiv.org/abs/2609.19814">Long-horizon autoformalization of a core theorem underlying MIP* = RE</a> — arXiv</li>
<li><a href="https://arxiv.org/html/2609.19814v1">Long-horizon autoformalization of a core theorem underlying MIP*=RE</a> — arXiv</li>
</ul>
<p><b>Mila and Mozilla announce local-AI foundation initiative</b></p>
<p>The partners plan open interfaces and locally installable implementations, with initial working implementations expected within six months.</p>
<ul>
<li><a href="https://mila.quebec/en/news/mila-and-mozilla-announce-new-initiative-to-build-trustworthy-open-source-ai-for-everyone">Mila and Mozilla Announce New Initiative To Build Trustworthy Open Source AI For Everyone</a> — Mila</li>
<li><a href="https://hypertec.com/news/mila-and-mozilla-announce-new-initiative-to-build-trustworthy-open-source-ai-for-everyone">Mila and Mozilla Announce New Initiative To Build Trustworthy Open Source AI For Everyone</a> — Hypertec</li>
</ul>
<p><b>Microsoft Agent Framework updates MCP security handling</b></p>
<p>Python 1.19.0 changes session scoping and archive handling.</p>
<ul>
<li><a href="https://github.com/microsoft/agent-framework/releases/tag/python-1.19.0">Release python-1.19.0 · microsoft/agent-framework</a> — Microsoft / GitHub</li>
</ul>
<p><b>Talos reports indications of AI-generated Qilin scripts</b></p>
<p>Talos gives a qualified assessment of possible AI-assisted code generation.</p>
<ul>
<li><a href="https://blog.talosintelligence.com/ransomware-incidents-in-japan-in-the-first-half-of-2026/">Ransomware incidents in Japan in the first half of 2026: Investigation of The Gentlemen’s infrastructure and evidence of Qilin's AI use</a> — Cisco Talos</li>
</ul>
<p><b>CoreWeave prices enlarged $3.7 billion debt offering</b></p>
<p>The AI cloud provider scheduled settlement for September 22, subject to closing conditions.</p>
<ul>
<li><a href="https://www.sec.gov/Archives/edgar/data/1769628/000176962826000429/ex9911.htm">CoreWeave Announces Proposed $3.0 Billion Convertible Senior Notes Offering</a> — CoreWeave / U.S. Securities and Exchange Commission</li>
<li><a href="https://investor.wedbush.com/wedbush/article/bizwire-2026-9-18-coreweave-prices-upsized-37-billion-convertible-senior-notes-offering">CoreWeave Prices Upsized $3.7 Billion Convertible Senior Notes Offering</a> — CoreWeave via Business Wire / Wedbush</li>
</ul>
<p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Mon, 21 Sep 2026 07:06:34 +0200</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/02989bde/d5284029.mp3" length="3151873" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>197</itunes:duration>
      <itunes:summary>FormalFlow’s supervised proof work, Mila and Mozilla’s local-AI initiative, Microsoft’s agent security changes, Talos’s Qilin findings and CoreWeave’s enlarged debt offering.</itunes:summary>
      <itunes:subtitle>FormalFlow’s supervised proof work, Mila and Mozilla’s local-AI initiative, Microsoft’s agent security changes, Talos’s Qilin findings and CoreWeave’s enlarged debt offering.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/02989bde/transcript.txt" type="text/plain"/>
    </item>
    <item>
      <title>New multimodal and speech models, and AI safety programmes</title>
      <itunes:title>New multimodal and speech models, and AI safety programmes</itunes:title>
      <itunes:episodeType>full</itunes:episodeType>
      <guid isPermaLink="false">132f8c0a-b3e3-44f2-b849-93249a5aa42d</guid>
      <link>https://share.transistor.fm/s/0c0f9165</link>
      <description>
        <![CDATA[<p>Qwen's multimodal tools, Anthropic's evaluation partnership and life-sciences programme, PrismML's compressed model, Wispr's Canto, and OpenAI's incident disclosures.</p><p>Stories and sources</p><p>Alibaba details Qwen3.8-Omni-Flash and accompanying tools</p><p>Alibaba describes multimodal inputs, text output and accompanying tools for audio-video agent workflows.</p><ul><li><a href="https://www.alibabacloud.com/blog/qwen3-8-omni-flash-omni-senses--agentic-delivery-_603580">Qwen3.8-Omni-Flash: Omni Senses. Agentic Delivery.</a> — Alibaba Cloud</li><li><a href="https://www.qwencloud.com/models/qwen3.8-omni-flash">Qwen3.8-Omni-Flash</a> — QwenCloud</li></ul><p>Anthropic and Accenture announce embedded evaluation partnership</p><p>Evaluators will work inside Anthropic; funding and operating arrangements remain under development.</p><ul><li><a href="https://www.anthropic.com/news/accenture-embedded-evaluation">Partnering with Accenture on embedded evaluation</a> — Anthropic</li><li><a href="https://newsroom.accenture.com/news/2026/accenture-and-anthropic-partner-to-build-team-of-embedded-evaluators-at-anthropic">Accenture and Anthropic Partner to Build Team of Embedded Evaluators at Anthropic</a> — Accenture</li></ul><p>PrismML releases Ternary Bonsai 2 27B</p><p>PrismML releases compressed model weights and reports benchmark comparisons with the full-precision base model.</p><ul><li><a href="https://prismml.com/news/bonsai-2-27b">Introducing Bonsai 2 27B: Near-Lossless Compression in a 9x Smaller Footprint</a> — PrismML</li><li><a href="https://www.prnewswire.com/news-releases/prismml-launches-bonsai-2-27b-its-most-capable-model-yet-302882228.html">PrismML Launches Bonsai 2 27B, Its Most Capable Model Yet</a> — PrismML via PR Newswire</li></ul><p>Wispr introduces Canto and deploys it across Flow</p><p>Wispr reports dictation benchmark results, including limitations on challenging and public datasets.</p><ul><li><a href="https://wisprflow.ai/canto">Canto: a speech model built for the real world</a> — Wispr</li><li><a href="https://wisprflow.ai/whats-new">What's new</a> — Wispr</li></ul><p>Anthropic opens Life Sciences Verification Program applications</p><p>The beta offers verified teams biology access under differentiated grants.</p><ul><li><a href="https://www.anthropic.com/news/life-sciences-verification-program">Introducing the Life Sciences Verification Program</a> — Anthropic</li></ul><p>OpenAI publishes reporting framework and six misalignment disclosures</p><p>OpenAI describes individual training and evaluation incidents and plans faster disclosure.</p><ul><li><a href="https://openai.com/index/model-misalignment-reporting-framework/">Our framework for reporting model misalignment</a> — OpenAI</li></ul><p>This episode uses an AI-generated narrator.</p>]]>
      </description>
      <content:encoded>
        <![CDATA[<p>Qwen's multimodal tools, Anthropic's evaluation partnership and life-sciences programme, PrismML's compressed model, Wispr's Canto, and OpenAI's incident disclosures.</p><p>Stories and sources</p><p>Alibaba details Qwen3.8-Omni-Flash and accompanying tools</p><p>Alibaba describes multimodal inputs, text output and accompanying tools for audio-video agent workflows.</p><ul><li><a href="https://www.alibabacloud.com/blog/qwen3-8-omni-flash-omni-senses--agentic-delivery-_603580">Qwen3.8-Omni-Flash: Omni Senses. Agentic Delivery.</a> — Alibaba Cloud</li><li><a href="https://www.qwencloud.com/models/qwen3.8-omni-flash">Qwen3.8-Omni-Flash</a> — QwenCloud</li></ul><p>Anthropic and Accenture announce embedded evaluation partnership</p><p>Evaluators will work inside Anthropic; funding and operating arrangements remain under development.</p><ul><li><a href="https://www.anthropic.com/news/accenture-embedded-evaluation">Partnering with Accenture on embedded evaluation</a> — Anthropic</li><li><a href="https://newsroom.accenture.com/news/2026/accenture-and-anthropic-partner-to-build-team-of-embedded-evaluators-at-anthropic">Accenture and Anthropic Partner to Build Team of Embedded Evaluators at Anthropic</a> — Accenture</li></ul><p>PrismML releases Ternary Bonsai 2 27B</p><p>PrismML releases compressed model weights and reports benchmark comparisons with the full-precision base model.</p><ul><li><a href="https://prismml.com/news/bonsai-2-27b">Introducing Bonsai 2 27B: Near-Lossless Compression in a 9x Smaller Footprint</a> — PrismML</li><li><a href="https://www.prnewswire.com/news-releases/prismml-launches-bonsai-2-27b-its-most-capable-model-yet-302882228.html">PrismML Launches Bonsai 2 27B, Its Most Capable Model Yet</a> — PrismML via PR Newswire</li></ul><p>Wispr introduces Canto and deploys it across Flow</p><p>Wispr reports dictation benchmark results, including limitations on challenging and public datasets.</p><ul><li><a href="https://wisprflow.ai/canto">Canto: a speech model built for the real world</a> — Wispr</li><li><a href="https://wisprflow.ai/whats-new">What's new</a> — Wispr</li></ul><p>Anthropic opens Life Sciences Verification Program applications</p><p>The beta offers verified teams biology access under differentiated grants.</p><ul><li><a href="https://www.anthropic.com/news/life-sciences-verification-program">Introducing the Life Sciences Verification Program</a> — Anthropic</li></ul><p>OpenAI publishes reporting framework and six misalignment disclosures</p><p>OpenAI describes individual training and evaluation incidents and plans faster disclosure.</p><ul><li><a href="https://openai.com/index/model-misalignment-reporting-framework/">Our framework for reporting model misalignment</a> — OpenAI</li></ul><p>This episode uses an AI-generated narrator.</p>]]>
      </content:encoded>
      <pubDate>Sun, 20 Sep 2026 16:16:45 +0200</pubDate>
      <author>ResearchNode.dev</author>
      <enclosure url="https://media.transistor.fm/0c0f9165/a97bb367.mp3" length="4619745" type="audio/mpeg"/>
      <itunes:author>ResearchNode.dev</itunes:author>
      <itunes:duration>289</itunes:duration>
      <itunes:summary>Qwen's multimodal tools, Anthropic's evaluation partnership and life-sciences programme, PrismML's compressed model, Wispr's Canto, and OpenAI's incident disclosures.</itunes:summary>
      <itunes:subtitle>Qwen's multimodal tools, Anthropic's evaluation partnership and life-sciences programme, PrismML's compressed model, Wispr's Canto, and OpenAI's incident disclosures.</itunes:subtitle>
      <itunes:keywords></itunes:keywords>
      <itunes:explicit>No</itunes:explicit>
      <podcast:transcript url="https://share.transistor.fm/s/0c0f9165/transcript.txt" type="text/plain"/>
    </item>
  </channel>
</rss>
