AI progress — 10 November 2026

Claude Fable 5.2 reached AWS and GitHub Copilot on the same day, and is the main release of the window. Separately, the scandal around Nvidia CEO Jensen Huang dominated the news, with reports of a Chinese espionage probe. Ars Technica also reports Claude Code settings files leaking secrets on GitHub. Several claims below are single-source and are attributed as such.

New models and papers

• Claude Fable 5.2 on AWS — AWS announces availability of Claude Fable 5.2. A Hacker News post links an Artificial Analysis page listing it at 65, the highest score on that index so far. Why it matters: teams on AWS can call the new model without leaving their existing platform; the index score is a third-party claim worth checking against your own workload.
• Muse Ember (Meta) — Meta presents Muse Ember as its smallest, most efficient model. Why it matters: a small model from a major lab is a candidate for cheap or on-device workloads; check the post for sizes and benchmarks before relying on it.
• Next-Concept Modeling — a paper on fast, parallel, coarse-to-fine text generation. Why it matters: it targets the sequential bottleneck of token-by-token decoding; the speed claims are the authors’ and untested here.

New tools and software

• Claude Fable 5.2 in GitHub Copilot — GitHub adds Claude Fable 5.2 to Copilot. Why it matters: Copilot users can switch to the new model today.
• Secret scanning for GitHub Actions logs — GitHub secret scanning now detects secrets in Actions workflow logs. Why it matters: credentials leaked in CI output, not only in committed code, can now be flagged.
• Nemotron 3.5 Super VL in SageMaker JumpStart — NVIDIA’s Nemotron 3.5 Super VL vision-language model is now available in SageMaker JumpStart. Why it matters: AWS users can deploy it from JumpStart rather than self-hosting.

Overall digest

• Nvidia spy scandal targeting CEO Jensen Huang — The Verge summarizes what is known about the scandal; Ars Technica reports Huang admitted an affair while Nvidia probes possible Chinese espionage, and TechCrunch reports lawsuits are being prepared. Why it matters: it concerns the leadership of the main AI chip supplier and now carries a national-security angle; details are still emerging.
• Claude Code settings files expose secrets on GitHub — Ars Technica reports that Claude Code settings files committed to GitHub are exposing thousands of secrets. Why it matters: if you use Claude Code, check that settings files with credentials are not in your repos; the new Actions-log scanning above is a separate feature.
• Anthropic reportedly moving forward with IPO plans — TechCrunch reports, citing unnamed sources, that Anthropic plans to proceed with an IPO. Why it matters: it is unconfirmed; a listing would bring public financial disclosure for a major model vendor.