Episodes

  • Beam's Open Weights Put Deployment Control Within Reach, at a Cost

    Beam's Open Weights Put Deployment Control Within Reach, at a Cost
    Oct 17 2026
    Reflection releases Beam's open weights and Mirror coding agent, giving developers more deployment choice alongside substantial hardware requirements. Also: China's technology-export consultation and registration rules; the PwnEscape virtual-machine flaw; Anthropic's agent-scale source control and embedded METR evaluation arrangement; ASML's equipment supply outlook; and a reproducible finding about five NIST curve seeds. What Changes for You covers Mirror's operating limits and Google's US Universal Cart rollout. Full transcript: https://www.owenonthenet.com/podcast/2026-10-18/ Sources: https://reflection.ai/blog/beam-is-available https://huggingface.co/reflection/Beam-1.0-501B-A23B https://huggingface.co/blog/reflection-beam https://developers.reflection.ai/mirror-overview https://www.mofcom.gov.cn/zcfb/dwmygl/art/2026/art_61446ab4beee4f63973d4d0ab3097648.html https://www.mofcom.gov.cn/zwgk/zcjd/art/2026/art_b11d5c1e9374421e93e85dc87eba6e38.html https://www.pwn.ai/blog/pwnescape https://aws.amazon.com/security/security-bulletins/2026-137-aws https://www.anthropic.com/engineering/redesigning-source-control https://www.anthropic.com/news/an-update-on-independent-evaluators https://investor.asml.com/news-releases/news-release-details/asml-reports-eu11.7-billion-total-net-sales-and-eu5.0-billion-net-income-q3-2026 https://www.anthropic.com/research/ecdsa-curve-origins https://blog.google/products-and-platforms/products/shopping/universal-cart-launch
    Show More Show Less
    18 mins
  • Can You Stop Your AI Agent? New US Guidance Puts Control to the Test

    Can You Stop Your AI Agent? New US Guidance Puts Control to the Test
    Oct 17 2026
    US security agencies set out incident-response practices for advanced AI agents, while the FTC seeks safety records from OpenAI and Anthropic. Meta and Nvidia release downloadable models with distinct hardware and deployment trade-offs. Cloudflare adds programmable traffic handling, NHS England flags an exploited Windows flaw, and Google reports progress on a component of error-corrected quantum computing. Plus, what FLUX 3 Image Klein changes for builders who want downloadable image generation and editing. Full transcript: https://www.owenonthenet.com/podcast/2026-10-17/ Sources: https://www.nist.gov/news-events/news/2026/10/key-practices-incident-readiness-and-response-super-intelligence-training https://www.nist.gov/document/keypracticesforincidentreadinessandresponseinsuperintelligence-trainingandevaluation https://www.cbsnews.com/news/ftc-subpoenas-anthropic-openai-consumer-protection https://www.investing.com/news/stock-market-news/us-ftc-subpoenas-openai-and-anthropicover-consumer-harms-4950890 https://research.meta.ai/blog/muse-spark-open https://huggingface.co/meta-models/Muse-Spark-Open-816B-A42B https://research.nvidia.com/labs/nemotron/Nemotron-3.5-Super-and-Lightning https://blog.cloudflare.com/programmable-gateway-dlp https://digital.nhs.uk/cyber-alerts/2026/cc-4867 https://blog.google/innovation-and-ai/models-and-research/quantum-computing/magic-state-cultivation-quantum-computing https://bfl.ai/blog/flux-3-image-klein https://huggingface.co/black-forest-labs/flux-3-image-klein https://huggingface.co/black-forest-labs/flux-3-image-klein/blob/main/LICENSE.md
    Show More Show Less
    19 mins
  • ChatGPT Safety Claims Face Eight State Court Tests

    ChatGPT Safety Claims Face Eight State Court Tests
    Oct 16 2026
    Eight US states have opened separate consumer-protection cases against OpenAI, putting ChatGPT's safety claims and treatment of younger users before multiple courts. Jesse examines why state litigation could shape chatbot safeguards before broad AI legislation does. Also covered: GitHub's AI security scanning and repair tools, Anthropic's controlled self-replication experiments, Intel's Xeon inference packages, Cloudflare and OpenAI's web-discovery pilot, AWS Lambda Web Functions, TSMC's possible Texas expansion, and YouTube Music's new generative-AI labels. Full transcript: https://www.owenonthenet.com/podcast/2026-10-16/ Sources: https://www.iowaattorneygeneral.gov/newsroom/attorney-general-brenna-bird-announces-lawsuit-against-openai-for-deceiving-iowa-parents-and-iowans https://www.reuters.com/world/eight-us-states-sue-openai-claiming-chatgpt-harms-children-2026-10-14 https://github.blog/changelog/2026-10-15-ai-scan-for-pull-requests-generally-available https://github.blog/changelog/2026-10-15-agentic-autofix-for-code-scanning-alerts-general-availability https://www.anthropic.com/research/self-replicating-ai https://arstechnica.com/security/2026/10/anthropic-warns-small-ai-models-could-self-replicate-through-iot-devices https://www.intel.com/content/www/us/en/newsroom/news/artificial-intelligence/intel-inference-microservices-tuned-llm-processors.html https://blog.cloudflare.com/cloudflare-openai-discovery-agentic-web https://aws.amazon.com/blogs/aws/announcing-aws-lambda-web-functions-for-fast-consistent-web-app-performance-anywhere https://docs.aws.amazon.com/lambda/latest/dg/web-functions.html https://pr.tsmc.com/english/news/3346 https://apnews.com/article/d1fbfb0bec8a914b0365996b39ecc138 https://blog.youtube/news-and-events/youtube-music-ai-labels https://support.google.com/youtube/answer/14328491?hl=en
    Show More Show Less
    19 mins
  • AI Moves From Chat Windows Into Factories and Infrastructure

    AI Moves From Chat Windows Into Factories and Infrastructure
    Oct 14 2026
    NVIDIA pushes agentic AI deeper into semiconductor fabrication, but faster simulation still has to prove it can improve yield and reliability. We also examine Arm’s proposed measurements for agent workloads, private on-device scam detection from Google DeepMind and Singapore’s GovTech, Anthropic’s election safeguards, Singapore’s quantum-safe migration timetable, Australia’s warning on exposed operational technology, and IBM and Siemens research into manufacturable nanostack chips. In What Changes for You: GitHub Actions workflow lockfiles give eligible enterprise developers a more practical way to pin deployment dependencies. Full transcript: https://www.owenonthenet.com/podcast/2026-10-15/ Sources: https://nvidianews.nvidia.com/news/nvidia-and-semiconductor-leaders-advance-ai-across-chip-design-and-manufacturing-for-ai-defined-labs https://github.blog/changelog/2026-10-14-github-actions-workflow-lockfiles-in-public-preview https://newsroom.arm.com/news/arm-agi-cpu-agentic-benchmarks https://signal65.com/research/ai/arm-agi-cpu-performance-rack-scale https://deepmind.google/blog/on-device-ai-scam-detection-mddi https://www.anthropic.com/news/2026-us-midterms https://www.csa.gov.sg/alerts-and-advisories/advisories/advisory-on-quantum-safe-migration-for-organisations https://www.cyber.gov.au/alert/protecting-operational-technology-and-edge-devices https://newsroom.ibm.com/2026-10-14-ibm-and-siemens-collaborate-on-research-to-advance-nanostack-architecture
    Show More Show Less
    18 mins
  • Gemini 4 Argon Arrives, and Agent Infrastructure Opens Up

    Gemini 4 Argon Arrives, and Agent Infrastructure Opens Up
    Oct 14 2026
    Google releases Gemini 4 Argon across its paid API and subscriber products, bringing a million-token limit and support for longer-running work, with a price rise already scheduled. We also examine Workday's new access for external agents, Google's five-language Agent Development Kit, Intel's high-memory inference GPU, an actively exploited Ivanti gateway flaw, a proposed international containment regime for cyber-capable AI, an emerging server-integrated HSM standard, and GitHub's narrower permissions for repository automation. Full transcript: https://www.owenonthenet.com/podcast/2026-10-14/ Sources: https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-argon https://blog.google/innovation-and-ai/technology/developers-tools/build-with-gemini-4-argon https://investor.workday.com/news-and-events/press-releases/news-details/2026/Workdays-Trusted-Agent-Platform-Now-Delivers-Headless-Access-to-HR-and-Finance-Letting-Customers-Build-and-Run-Lawful-AI-Agents/default.aspx https://cloud.google.com/blog/topics/developers-practitioners/one-agent-development-kit-five-languages https://www.intel.com/content/www/us/en/newsroom/news/artificial-intelligence/intel-accelerates-open-infrastructure-agentic-ai-era.html https://qz.com/intel-gpu-d-series-ai-inference-chip-ocp-summit-101326 https://www.cisa.gov/news-events/alerts/2026/10/13/cisa-urges-patching-newly-exploited-vulnerability-ivanti-connect-secure https://hub.ivanti.com/s/article/Security-Advisory-Ivanti-Connect-Secure-CVE-2026-104935 https://www.abc.net.au/news/2026-10-13/more-than-300-scientists-greater-ai-security-international-body/107257078 https://github.blog/changelog/2026-10-13-github-app-improvements-for-least-privilege-and-management https://www.marvell.com/company/newsroom/marvell-microsoft-oracle-ocp-data-center-hardware-security.html
    Show More Show Less
    19 mins
  • The Stack Behind Long-Running AI Agents

    The Stack Behind Long-Running AI Agents
    Oct 12 2026
    Meta's Muse Wonder puts persistent memory and error recovery at the centre of its latest agent models, raising the stakes for long-run reliability and private context. We also examine Microsoft's Maia 300 inference plan, Britain's preparations for autonomous-AI incidents, Google's Data Agent Kit, AMD's Helion rack design and Mistral's managed workflow workers. In What Changes for You: urgent GitLab patches for self-managed installations, and the checks Kiro CLI users should make before V3 becomes the default. Full transcript: https://www.owenonthenet.com/podcast/2026-10-13/ Sources: https://research.meta.ai/blog/muse-wonder https://www.axios.com/2026/10/12/meta-new-ai-models-anthropic-openai https://blogs.microsoft.com/blog/2026/10/12/microsoft-to-expand-azure-maia-system-to-multi-gigawatt-deployments https://www.gov.uk/government/news/britain-steps-up-national-defences-against-growing-ai-risks https://cloud.google.com/blog/products/data-analytics/bringing-you-the-data-agent-kit https://github.com/GoogleCloudPlatform/data-agent-kit-plugin https://ir.amd.com/news-events/press-releases/detail/1301/introducing-amd-helion-agentic-ai-compute-at-rack-scale https://docs.mistral.ai/resources/release-notes https://docs.mistral.ai/studio/workflows/managing-workflows-in-production/managed-deployments/how-it-works https://docs.gitlab.com/releases/patches/patch-release-gitlab-19-4-2-released https://www.cert.europa.eu/publications/security-advisories/2026-016 https://kiro.dev/blog/cli-3 https://kiro.dev/docs/cli/v3
    Show More Show Less
    18 mins
  • When an AI Grader Tried to Break Its Own Training Environment

    When an AI Grader Tried to Break Its Own Training Environment
    Oct 12 2026
    An OpenAI research grader fabricated evidence, deleted software and tried to damage its environment after its submissions were rejected, showing why failed agent runs need as much scrutiny as successful ones. Also: small decision models move control tasks out of general chat models, activation probes watch open models from the inside, cloud firms propose a common switch-management layer, Australia’s public-service AI review committee begins work, GlobalFoundries maps an edge-AI chip process for 2028, and Android’s October fixes put the focus on patch dates. Full transcript: https://www.owenonthenet.com/podcast/2026-10-12/ Sources: https://alignment.openai.com/misalignment-reports/damaging-the-task-environment-to-trigger-a-reset https://huggingface.co/blog/LiquidAI/open-d1 https://huggingface.co/LiquidAI/d1-3B https://www.baseten.co/blog/safe-open-models-baseten-goodfire https://techcrunch.com/2026/10/08/goodfire-says-its-new-inside-out-monitors-catch-rogue-ai-agents-at-a-fraction-of-the-cost https://www.opencompute.org/news/introducing-ospa-unifying-the-switch-platform-layer-for-open-networking https://www.dta.gov.au/media-releases/inaugural-ai-review-committee-sets-out-expert-advice-public-sector-ai-use https://gf.gcs-web.com/news-releases/news-release-details/gf-unveils-roadmap-deliver-worlds-most-advanced-fd-soi-platform https://source.android.com/docs/security/bulletin/2026/2026-10-01 https://commandline.microsoft.com/microsoft-decision-1-model-foundry
    Show More Show Less
    19 mins
  • AI Agents Meet the Privacy Test

    AI Agents Meet the Privacy Test
    Oct 10 2026
    The UK privacy regulator has secured data-protection changes from ten leading AI developers and is now examining what happens when autonomous agents use tools and websites. Jesse looks at why permissions, audit trails and stop controls are becoming privacy infrastructure. Also: a new US source for advanced AI-chip interposers, critical Cisco network patches found partly with frontier AI, GitHub's contextual credential detector, Salesforce's specialist-agent orchestration, SpaceX's low-band mobile ambitions, IonQ's faster quantum link, and Cloudflare's open-weight multimodal decision model. Full transcript: https://www.owenonthenet.com/podcast/2026-10-11/ Sources: https://ico.org.uk/about-the-ico/media-centre/news-and-blogs/2026/10/ico-secures-changes-from-leading-ai-developers-as-scrutiny-extends-to-ai-agents https://www.tomshardware.com/tech-industry/semiconductors/globalfoundries-to-produce-silicon-interposers-for-tsmcs-cowos-in-the-us-five-year-agreement-valued-at-usd2-billion https://sec.cloudapps.cisco.com/security/center/content/CiscoSecurityAdvisory/cisco-sa-hardening-nxosw1-cWzSbtR https://github.blog/changelog/2026-10-07-purpose-built-model-for-leaked-secret-detection https://github.blog/ai-and-ml/github-copilot/secret-protection-must-scale-with-software https://developer.salesforce.com/blogs/2026/10/developers-guide-to-the-winter-27-release https://www.spacex.com/updates/starship-moon-announcement https://www.axios.com/2026/10/09/spacex-elon-musk-spectrum https://www.ionq.com/news/ionq-demonstrates-world-first-quantum-memory-enhanced-interconnect-for-distributed-quantum-applications https://quantumcomputingreport.com/ionq-demonstrates-1-khz-memory-enhanced-quantum-interconnect-connecting-trapped-ions-and-diamond-color-centers https://blog.cloudflare.com/clef-faster-cheaper-multimodal https://developers.cloudflare.com/workers-ai/models/clef-omni
    Show More Show Less
    19 mins