the week in AI, briefly. then briefly again.
Artificial Intelligence — briefly, then briefly again · tbb.ceo
listen tothe daily 0:00 –:––
/daily ·13 NOV 2026 ·FRIDAY ·2 MIN READ ·7 STORIES

Friday 13 November 2026

Google ships a faster Gemini, Nvidia's post-Huang chaos compounds, and Anthropic joins your meetings.

01 / The Day

FRIDAY 13 NOV 2026, ranked

07

Google launches Gemini 4 Hydrogen: smaller, faster, cheaper

Google released Gemini 4 Hydrogen, the lowest-cost member of the Gemini 4 family and effectively a replacement for the Flash line. Early benchmarks put it close to Claude Sonnet 5.5 on the Artificial Analysis index while costing less per token than the larger Argon model it was released alongside.

  • Hydrogen outperforms Gemini 4 Argon on the Artificial Analysis index despite being smaller
  • Rolls out first to Google Search worldwide, then to developers via API
  • Simon Willison added Gemini 4 Hydrogen support to llm-gemini 0.37 on the same day
Why it mattersA capable, cheap Gemini model reaching Search at scale reshapes the cost economics of AI-powered products.

Nvidia fallout: petition, Congress, third affair, Tim Cook

Four separate new developments hit the Nvidia-Huang story on Friday: employees started a petition for his return, Reuters reported a third workplace affair, Huang agreed to testify before Congress after a threatened subpoena, and Trump publicly floated Tim Cook as a successor.

  • Nvidia employees petitioned the board to reinstate Huang as CEO
  • Huang agreed to December congressional testimony after threat of a Senate Banking subpoena
  • Democrats launched a formal Nvidia investigation into China chip sales
Why it mattersWhat started as a personal scandal has broadened to a congressional accountability question for the company that holds the keys to US AI compute.

Nscale needs more capital to deliver on $45B Anthropic deal

Nscale's updated IPO prospectus disclosed it needs additional financing to meet 2027 commitments tied to Nvidia chip orders placed for a $45 billion Anthropic compute deal. Q3 revenue was $307.5M, up from $140.6M in Q2, but net losses widened to $1.69B.

  • Nscale filed a new IPO prospectus after the FT reported its financing gap
  • Q3 net loss of $1.69B, up from $1.02B in Q2, despite near-doubling of revenue
  • The company's planned $35B IPO remains uncertain pending new capital
Why it mattersIf a neocloud cannot fund its commitments mid-IPO process, the whole model of prepayment-funded AI infrastructure comes under scrutiny.

Anthropic ships Claude Meetings: AI joins your video calls

Anthropic launched Claude Meetings, a beta feature that adds Claude to Google Meet sessions, generating documents, whiteboards, and notes in real time. It is live for Team and Enterprise plans, with more platforms planned.

  • Claude joins live Google Meet calls and produces documents and whiteboards during them
  • Currently in beta for Team and Enterprise plans only
  • Additional video platforms listed as coming soon
Why it mattersPutting a frontier model inside video calls rather than beside them is a different integration bet — it owns the meeting artifact instead of summarising it after.

Apple opens Foundation Models framework to developers

Apple published its Foundation Models developer framework, letting third-party apps run on-device foundation models directly. The page hit 197 Hacker News points and signals Apple's bet on local AI over cloud-dependent inference.

  • Framework enables apps to use on-device foundation models without cloud calls
  • Available via Apple's developer portal as a new API surface
  • No per-call pricing or cloud costs for developers beyond device capability
Why it mattersOn-device foundation model access at the app layer is the quiet alternative to the API-price treadmill, and Apple's distribution makes it meaningful.

OpenAI launches Daybreak Red Team and Defense Factory

OpenAI released two enterprise security products: Daybreak Red Team for automated vulnerability testing of web applications and Daybreak Defense Factory, which finds, validates, and patches security flaws end to end. Both are in early access.

  • Red Team automates web app vulnerability scanning; Defense Factory closes the fix loop
  • Available in early access, not general release
  • Positions OpenAI directly against cybersecurity incumbents
Why it mattersOpenAI entering the security tooling market with AI-native products is a new revenue vector — and a signal that model-company and security-company boundaries are blurring.

Mirendil raises $1B Series A for frontier AI R&D tools

Mirendil, founded by former Anthropic researchers, closed a $1 billion Series A led by Kleiner Perkins and Clearlake Capital, six months after a $200M seed round. The company builds tools to accelerate frontier AI R&D itself.

  • $1B Series A follows a $200M seed round just six months earlier
  • Co-founders are former Anthropic researchers building infrastructure for frontier labs
  • Kleiner Perkins and Clearlake Capital led the round
Why it mattersRaising $1B to build tools for the labs building the models is a new layer of the AI stack — and suggests the R&D acceleration bet is now fundable at scale.
Be subscriber #028 weekly AI email interest list
Past days

The Daily Archive