AI progress — 19 November 2026
Google DeepMind’s Gemini 4 Neon is the most consequential release of the day: it is a new frontier-class model family from a major lab, and GitHub shipped it in Copilot within hours. DeepMind also made Genie 4 available, and Microsoft announced seven models at Ignite. Outside releases, Nadella’s remark that much critical infrastructure is unpatchable stands out against fears of AI-driven attacks. The candidate feeds gave little detail beyond titles, so items below are kept to what the sources state.
New models and papers
- Introducing Gemini 4 Neon — Google DeepMind announces a new Gemini model, Gemini 4 Neon. Why it matters: It is the day’s main release, and it already reached GitHub Copilot the same day, so developers choosing a coding model have a new option to test. One independent blog post compares it to Opus 5.5; that is a single author’s view, not a benchmark.
- Genie 4 is now available — DeepMind says Genie 4 can now be used, with a Project Genie page linked from the announcement. Why it matters: Access moves from announcement to something people can try; check the post for what is offered and to whom.
- Seven new models at scalable price-performance — Microsoft AI’s Ignite 2026 post announces seven new models. Why it matters: The title frames them around price-performance, which matters for anyone comparing model cost; read the post for specifics before relying on it.
New tools and software
- Connect Supabase, Box, Calendly, and more to Copilot — GitHub adds connections from Copilot to Supabase, Box, Calendly and other services. Why it matters: Copilot can now be pointed at those services directly instead of through custom glue.
- Wait for pull request events with GitHub CLI — The GitHub CLI can now wait for pull request events. Why it matters: Scripts and agents that gate on PR state can use the CLI rather than polling the API themselves.
- Full Fine-Tuning (FFT) on SageMaker AI Serverless Model Customization — AWS describes full fine-tuning in SageMaker’s serverless model customization. Why it matters: Teams already on SageMaker can do full fine-tuning without managing the training infrastructure, according to the title; check the post for supported models.
Overall digest
- Satya Nadella says lots of critical infrastructure is ‘unpatchable’ — Speaking amid fears of AI-driven cyberattacks, the Microsoft CEO said much critical infrastructure cannot be patched. Why it matters: It is a statement from a major vendor that the defensive answer to AI-assisted attacks cannot rest on patching alone.
- Coalition challenges Anthropic’s safety branding ahead of IPO — A press release from a coalition of scientists, legal scholars and tech leaders disputes Anthropic’s safety positioning. Why it matters: This is one group’s advocacy release, not a finding; it is relevant to anyone weighing the company’s safety claims ahead of its reported IPO.
- Muse privacy invasions are freaking out Meta staffers — Ars Technica reports, from internal chats, that Meta staff are alarmed by privacy problems with Muse. Why it matters: Internal concern at the builder is a signal for anyone planning to adopt or integrate the product.
Get the AI digest by email
One email per weekday morning: new models and papers, new tools, and the overall picture. No ads, unsubscribe in one click. See the privacy policy.