the week in AI, briefly. then briefly again.
Artificial Intelligence — briefly, then briefly again · tbb.ceo
listen tothe daily 0:00 –:––
/daily ·12 NOV 2026 ·THURSDAY ·3 MIN READ ·7 STORIES

Oust Day

The world's most valuable chip company may have a new CEO by Friday; Google shipped a model that doesn't generate, it decides.

01 / The Day

THURSDAY 12 NOV 2026, ranked

07

Nvidia board moves to oust CEO Jensen Huang amid spy scandal

Nvidia's board is preparing to remove Jensen Huang — its co-founder and 32-year CEO — following his admitted affair with a woman US intelligence identifies as a foreign agent, plus a separate affair with a subordinate. CFO Colette Kress is being discussed as interim replacement; the White House has been briefed; Congress is demanding a closed-door hearing.

  • Board working with an outside firm to identify candidates; Colette Kress cited as likely interim
  • Congress requested a closed-door briefing from Huang on chip exports and the affairs
  • Alice Huang, the Megaspeed founder at the centre of the spy allegation, has returned to Beijing
Why it mattersThe founder-CEO of the world's most critical AI chip company may be replaced within days — a governance crisis with direct implications for AI supply-chain continuity and US-China compute policy.

Google launches Gemini Decisions API: 1,000 questions, one pass

Google DeepMind released Gemini Decision 1 and its companion Decisions API — a model built not for generation but for scoring. It evaluates up to 1,000 typed questions against multimodal inputs in a single forward pass and returns calibrated probabilities, billed at $0.10 per million input tokens with no output charges.

  • 150-250ms end-to-end latency; designed for agentic routing, rubric scoring, and policy enforcement
  • Accepts text, images, audio, video, and PDF inputs in a single batch call
  • Output tokens are free; model scores rather than generates, cutting orchestration inference cost
Why it mattersA purpose-built decision model at this price point changes how developers build agentic pipelines — cost-of-inference for orchestration just dropped by an order of magnitude.

OpenAI quarterly revenue crosses $10B for the first time

OpenAI's Q3 revenue reached approximately $10 billion — a roughly 50% jump from $6.7B in Q2 — driven by enterprise API demand. Operating losses reportedly widened despite the surge, and The Decoder notes Anthropic is achieving higher revenue with better operating efficiency.

  • $10B quarterly equates to roughly a $40B annualised run rate ahead of a likely 2027 IPO
  • Enterprise API and ChatGPT Enterprise are the primary growth drivers
  • Operating losses still widening; exact figures not public as OpenAI remains private
Why it mattersThe milestone sets a credible revenue baseline for OpenAI's forthcoming IPO while surfacing that Anthropic may be the more efficient operator at scale.

Amazon ships official AWS emulator written by AI agents

Amazon Web Services released an open-source AWS emulator — a local test environment covering hundreds of AWS APIs — with the codebase substantially written by autonomous AI coding agents rather than human engineers, making it one of the largest production software systems credibly attributed to agent-generated code.

  • Covers hundreds of AWS APIs; works locally without a live AWS account
  • Codebase written, validated, and maintained by AI agents under human-specified guardrails
  • Available at github.com/aws/aws-emulator; requires Docker on macOS and Windows
Why it mattersWhen AWS ships production infrastructure written by its own agents, autonomous software engineering has crossed a credibility threshold — no longer a demo, now a shipped product.

Chinese AI firms rush overseas Nvidia compute as US tightens access

Chinese AI companies are urgently contracting capacity in overseas data centres — Singapore, Malaysia, UAE — as a hedge against tightening US rules, while the Commerce Department finalises a rule that would directly block remote access to US AI compute by Chinese entities and close chip re-export routes through third countries.

  • Chinese firms securing Nvidia GPU capacity in third-country data centres before rules close
  • Commerce rule would restrict both remote access AND chip re-export via overseas intermediaries
  • Jensen Huang's DC lobbying has collapsed with the spy scandal; bipartisan controls now have tailwind
Why it mattersThe US-China AI compute separation — long telegraphed — is now executing in real time, compressing a years-long policy timeline into weeks.

New York AI safety law takes effect: register, disclose, or face fines

New York's AI safety statute activated on November 12, requiring developers of major AI systems — covering OpenAI, Anthropic, Google, and similar large-scale deployers — to register with the state, disclose safety evaluation practices, and report critical incidents to regulators within 72 hours.

  • Registration, disclosure, and incident-reporting obligations now enforceable in New York
  • OpenAI and Anthropic specifically named in coverage; thresholds based on compute usage
  • New York joins California and the EU in having substantive AI safety compliance requirements
Why it mattersState-level AI safety regulation with enforcement teeth is a first at this scale in the US — it sets a compliance floor that other states and the federal government will reference.

Anthropic: Mythos ported Claude to AMD GPUs in 3 days, not months

In its IPO roadshow, Anthropic disclosed that its internal Mythos model compressed a multi-month AMD GPU migration project to roughly three days with minimal human engineering. Claude now runs on four chip families: Nvidia, AMD, Trainium, and TPUs.

  • Switching from Nvidia CUDA to AMD ROCm previously required months of engineering effort
  • Mythos completed it in ~3 days with minimal human involvement
  • Anthropic now runs Claude on four chip families; using this as an IPO investor pitch
Why it mattersIf AI can automate chip porting at scale, the Nvidia software moat — the main justification for CUDA premium pricing — becomes structurally weaker over time.
Be subscriber #028 weekly AI email interest list
Past days

The Daily Archive