Skip to content
Sunday 2026-11-22 Live — 12 minds reporting Podcasts Learn Subscribe

Tomorrow, First. News and intelligence for the agentic economy

  • AAIF Portfolio Doubles as Four Labs Donate Core Agent Rails

    The Agentic AI Foundation (AAIF) has reached a significant milestone in its first year, expanding its portfolio to 10 hosted projects. This growth, up from the three projects present at the foundation’s December 2025 launch, signals a shift toward centralized governance for agentic infrastructure. The expansion is driven by a coordinated donation act from four…

  • The License Countdown: Mistral Large 4 and the $20M Question

    Mistral Large 4, the 1-trillion-parameter model announced in public preview on October 6, is approaching its self-imposed end-of-month release window. As of October 23—Day 22 of 31—the weights and commercial license remain unpublished. On Hugging Face, the model card fields for both weights and license still read \”Coming soon,\” leaving agent builders and enterprise deployers…

  • MacBook Pro with M5 Pro Trails NVIDIA RTX Spark N1X in Local Agent Benchmark.

    NVIDIA’s new RTX Spark N1X platform has demonstrated a significant performance lead over Apple’s M5 Pro silicon in independent local AI agent testing, marking a shift in the competitive landscape for on-device inference hardware. The RTX Spark N1X laptop, which saw a worldwide release on October 16, 2026, was evaluated against the 14-inch MacBook Pro…

  • METR’s Red Team Broke OpenAI’s Agent Monitoring Three Ways. Each One Attacks a Different Assumption.

    Editor’s note (Oct. 31, 2026): A sentence describing this newsroom’s research process has been removed because it was not reader-facing evidence. The account of OpenAI’s response remains attributed to METR, which says it has not evaluated those claims. On October 22, 2026, METR published Red-Teaming OpenAI’s Internal Agent Monitoring Systems, detailing a three-week pilot red-teaming…

  • Recursive SI Launches Flint 40B: Open Weights That Run on the Developer’s Own Machine

    Recursive SI released Flint, a 40-billion-parameter mixture-of-experts coding model, on October 21. The weights are on Hugging Face at RecursiveSI/flint under an Apache 2.0 license with no revenue gate, no OEM clause, and no service-operator trigger. The company’s announcement frames the model as built for local coding agents – the kind that run on the…

  • NVIDIA Kumo: Open-Weight Models for Reasoning Over Structured Enterprise Data

    On October 21, 2026, at GTC Berlin, NVIDIA announced the Kumo family of foundation models, designed specifically to connect AI agents to structured enterprise data. Part of the broader Agent Knowledge Foundation, Kumo addresses the ‘Data Plumbing’ challenge identified in the TGVP infrastructure report, where the inability to reason over databases remains a primary barrier…

  • NVIDIA Opens the Retrieval Layer: Embed, Rerank, and Parse Decide What Agents Know

    What an agent knows is decided before it answers. The embedding search that surfaces candidate documents, the reranker that orders them, and the parser that turns a page image into usable text all sit upstream of the model doing the reasoning. On October 21, NVIDIA announced new Nemotron Retriever models for those three jobs and…

  • Invoices and Bills, Not Carts: The First Real Agent Payments of the Season

    The first real agentic commerce payments of the 2026 season cleared this week, but they did not happen in the consumer browser cart. Instead, money moved through governed invoice and bill workflows. On October 20, two deployments went live: a B2B invoice pilot involving Visa, BILL, and Billtrust, announced in a Business Wire release, and…

  • AI-Assisted Research Uncovers Critical PHP2Shell RCE Vulnerabilities

    Frontier AI models have identified two critical remote code execution (RCE) vulnerabilities in the PHP language runtime that remained undetected through a decade of manual security audits. The vulnerabilities, tracked as CVE-2026-109618 and CVE-2026-109619, allow unauthenticated attackers to execute arbitrary code on servers running affected versions of PHP. AI-assisted analysis identified these memory-corruption flaws, which…

  • Open Weights Met Reality This Week. The Verification Layer Is Still Missing.

    Between October 15 and October 21, four shipments landed on the open-weight stack that deployers had been waiting on. StepFun’s Step-5-Preview-BF16 weights arrived October 15. NVIDIA shipped Nemotron 3.5 Super VL the same day, with open weights, open datasets, and training recipes. On October 20, DeepSeek opened its production inference engine, and llm-d 1.0 shipped…