Skip to content

AI Weekly #16 — Nov 9–15, 2026

November 9–15, 2026

Onei AI Editorial 2 min read How we make this

The big story Model release

Anthropic ships Claude Fable 5.2 and Mythos 5.2: up to 30% faster, at $8/$40 per million tokens

Both run on one underlying model. Fable 5.2 is open to all users behind cyber and biology safeguards; Mythos 5.2 goes to organizations in Anthropic's Life Sciences and Cyber Verification Programs. Anthropic says they beat Opus 5.5 on almost every benchmark it ran, and in testing they formalized the proof of Fermat's Last Theorem in 30 hours.

Why it matters: The same model now ships twice, safeguarded for everyone and unrestricted for vetted defenders and researchers; that split is becoming how frontier labs release their strongest systems.

Anthropic

Top stories

  1. Jensen Huang resigns as Nvidia CEO after 33 years; CFO Colette Kress becomes interim CEO

    The board says the roadmap and customer commitments continue unchanged, but the supplier most AI labs depend on for compute now has an interim leader and a CEO search.

    NVIDIA Ecosystem

  2. Anthropic launches its IPO roadshow: 825 million shares at $70–78, to list on Nasdaq as ANT

    Anthropic itself sells 801 million of those shares, so most of the money goes to the company rather than to existing holders cashing out.

    Anthropic Funding

  3. Google's Gemini 4 Hydrogen edges Argon on the AA Intelligence Index at $1.50/$7.50 per million tokens

    Flat pricing across a 1M-token context, plus an Antigravity Orchestrator that switches between Carbon and Hydrogen, makes it the execution half of multi-model coding setups.

    Google Model release

  4. Four nonprofits sue OpenAI, Anthropic, Google, Meta and SpaceX to freeze frontier AI development

    Filed in San Francisco Superior Court under California's unfair competition law, it seeks an emergency injunction on next-generation models until the companies prove human control.

    Al Jazeera Policy

Quick hits

  • DeepMind's AlphaGenome 2 predicts DNA-variant effects in individual cell types — finds nearly twice as many causal variants; free for non-commercial research through its API. Google DeepMind

  • Runway opens the weights of Praxis-1, its world action model, under Apache 2.0 — 27B parameters; Runway puts it first on NVIDIA's RoboLab-120 leaderboard at 46.2% success. Runway

New on Onei AI this week

Released during this week, and now in the catalogue.

  • Meetily

    Meetily · App — v0.5.0 brings Linux: native PipeWire and PulseAudio system-audio capture, plus signed .deb and AppImage packages.

  • MCP Server Kubernetes

    MCP Server Kubernetes · MCP server — v4.2.3 merges several kubeconfig files the way kubectl does, days after v4.2.2 let the Helm tools target a chosen context.

  • Cloud Run MCP

    Cloud Run MCP · MCP server — v1.12.0 makes every tool call work under OAuth and fixes a potential cache-poisoning vulnerability.

  • MailPace MCP

    MailPace MCP · MCP server — First npm release: tools register only for the tokens you set, and credentials are read from environment variables, never command-line arguments.

The directory gained 32 apps, 11 agent skills and 11 MCP servers this week; it now lists 718 apps, 383 skills and 393 MCP servers.

Browse the full catalogue →

Get the next issue by email

One email a week with the stories and releases above, before they reach the archive. Unsubscribe in one click.

We send a confirmation link first. See our privacy policy.