This is an earlier edition. Today's AI Daily →

DailyAISun, 15 Nov 2026Published 10:01 UTC
Compute, chips and supply: export controls and China's compute gap
OHMEGA / MONSTERAI Daily Sun, 15 Nov 2026AI-generated image — not a photograph

Quiet Nov 15 left Fable 5.2 ahead of Gemini 4 Carbon as cost, chip-probe and Gödel claims stayed unverified

The day in 30 seconds

All situations side by side · state, direction, Pulses, coverage
State · move in 24hΔ 24h
  1. Compute, chipsNothing new officially moved since Nov 14 — the picture that stands is the Nov 11 resignation of Jensen Huang… steady
    model-assessed
    ◆ 0 steady
  2. Gemini 4 Carbon frontier pushFrontier leadership is unchanged since Nov 12: Fable 5.2's 5-point lead on Artificial Analysis Intelligence… steady— steady
  3. AI agents: labor disruption, rogue behaviorPattern from Nov 8-12 held through Nov 15: labs kept shipping workplace… steady
    model-assessed
    ◆ 0 steady
  4. AI for scientific discoveryThe only new move since Nov 12 is a Nov 13 Guardian report that Helsinki logician Jan von Plato used… steady— rising
  5. AI economicsNo new pricing, funding closes or spend data since Nov. 13 has shifted the AI capex… steady— steady
The day, in more detail (6)
The day
  • Fable 5.2 retained its 5-point lead over Gemini 4 Carbon on Artificial Analysis Intelligence Index v5 (65 vs 60)…
  • Researched Nov 14 MathArena put Gemini 4 Carbon ahead of Fable 5.2 (85.4% vs 82.8%) but at about 15 times the cost of GPT-6.1 Sol…
  • The Guardian reported Nov 13 that Helsinki logician Jan von Plato used Anthropic's Claude to transcribe 1,009 pages of Gödel's 1940s workbooks surfacing 663 theorems with about 400 said verified in Coq…
  • No official move since Nov 14 altered the compute-controls picture, leaving Jensen Huang's Nov 11 resignation…

State of play · not in today's sections

Compute, chips and supply: export controls and China's compute gap · situation → , flat pace

Nvidia CEO resignation and alleged Blackwell smuggling standoff held with no new official developments

The rundown

The United States and China are competing over the advanced chips that train frontier AI models. Washington, through the Commerce Department's Bureau of Industry and Security, restricts exports of high-performance chips and related technology and adds Chinese firms to its Entity List, while Nvidia makes the most sought-after chips including H200 and Blackwell. S.

Nothing new officially moved since Nov 14 — the picture that stands is the Nov 11 resignation of Jensen Huang, the Nov 12 forensic claim of Blackwell fingerprints in six Chinese frontier models, and open congressional inquiries plus Malaysia's re-export probe. Confidence is high on the stasis itself; the central smuggling versus offshore-rental question remains unverified pending testimony and findings.

On the record

  • Jensen Huang · Chief Executive, Nvidia · 10 Nov

    “I support the investigation and will cooperate fully,”

    Commitment · conciliatory toward United States, Nvidia as reported by Deutsche Welle ↗

  • Jensen Huang · Chief Executive, Nvidia · 10 Nov

    “I am committed to protecting NVIDIA and supporting America's national security.”

    Reassurance · conciliatory toward United States, Nvidia as reported by Deutsche Welle ↗

  • United StatesDonald Trump · President of the United States · 10 Nov

    “I just know Jensen very well. I just think he’s a very honourable man,”

    Reassurance · conciliatory toward Jensen Huang, Nvidia as reported by South China Morning Post ↗

All on the record (9) · Compute, chips and supply: export controls and China's compute gap, across the day's updates
  • Jensen Huang · Chief Executive, Nvidia · 10 Nov

    “I support the investigation and will cooperate fully,”

    Commitment · conciliatory toward United States, Nvidia as reported by Deutsche Welle ↗

  • Jensen Huang · Chief Executive, Nvidia · 10 Nov

    “I am committed to protecting NVIDIA and supporting America's national security.”

    Reassurance · conciliatory toward United States, Nvidia as reported by Deutsche Welle ↗

  • United StatesDonald Trump · President of the United States · 10 Nov

    “I just know Jensen very well. I just think he’s a very honourable man,”

    Reassurance · conciliatory toward Jensen Huang, Nvidia as reported by South China Morning Post ↗

  • Nvidia board · Board of Directors, Nvidia · 8 Nov

    paraphrase The board currently has no reason to believe that US national security or Nvidia's confidential information had been compromised, but will examine all available information in its independent investigation

    Reassurance · about United States, China as reported by Anadolu Agency ↗

  • Nvidia · company, Nvidia · 9 Nov

    “target of a foreign intelligence effort”

    Policy · about Nvidia, China as reported by Deutsche Welle ↗

  • MalaysiaAnwar Ibrahim · Prime Minister, Malaysia · 15 Nov

    “as part of China, period”

    Policy · hostile toward Malaysia, China, Taiwan as reported by Taipei Times ↗

  • MalaysiaAnwar Ibrahim · Prime Minister, Malaysia · 11 Nov

    “as part of China, period”

    Policy · critical toward Taiwan, China, Malaysia as reported by Taipei Times ↗

  • MalaysiaAnwar Ibrahim · Prime Minister, Malaysia · 11 Nov

    “protect the sanctity and unity of”

    Remark · critical toward Taiwan, China, Malaysia as reported by Taipei Times ↗

  • TaiwanMinistry of Foreign Affairs · Ministry of Foreign Affairs, Taiwan · 11 Nov

    “absurdly implied that Taiwan is a separatist Chinese province and openly rationalized China’s stance on taking Taiwan by force”

    Accusation · critical toward Malaysia, China, Taiwan as reported by Taipei Times ↗

A statement shows what was said, not that it is true.

A statement shows what was said, not that it is true.

New

    Nothing changed since yesterday: No new official developments since Nov 14; resignation, forensic claims, congressional testimony commitments and Malaysia probe all stand as before.

    Watch for

    • Huang and Lutnick testimony before Thanksgiving (Nov 26) — scope on Megaspeed licensing and H200 tariff/revenue-share
    • Nov 18 joint House closed-door briefing on Huang interactions and Nvidia security protocols — whether it proceeds with interim leadership
    • Nov 17 Nvidia Q3 earnings — H200 China revenue, guidance, and any board probe update
    • Malaysia MITI findings on Megaspeed subsidiary re-exports and any BIS Entity List or licensing rule changes

    Outlook Stasis is likely to hold until the next forcing events. Testimony by Huang and Commerce Secretary Howard Lutnick before Thanksgiving (Nov 26) is almost certain to sharpen scrutiny of Megaspeed licensing and H200 tariff/revenue-share, but whether it resolves the smuggling versus offshore rental question is roughly even pending independent verification and Malaysia findings.

    Economies involved · size, trade, energy, arms · annual estimates

    Gemini 4 Carbon frontier push and challengers · situation → , flat pace

    Fable 5.2 keeps benchmark lead as Gemini 4 Hydrogen launches and Carbon's MathArena win comes at high cost

    Comparison Gemini 4 Carbon tops MathArena but at about 15 times the cost of GPT-6.1 Sol, while Fable 5.2 still leads the broader Intelligence Index. MathArena accuracy and expected cost for top models, with Artificial Analysis Intelligence Index for context.the-decoder.com ↗anthropic.com ↗deepmind.google ↗the-decoder.com ↗as of 2026-11-15AI-assisted analytic, built only from real cited or sourced data

    The rundown

    The frontier-model race pits Google's Gemini family, including Gemini 4 Carbon with agent swarm and Gemini Robotics, against Anthropic's Claude models, Meta's Muse Ember, ByteDance's Seedream and others, with OpenAI's GPT line also in the field. Labs compete to claim the most intelligent and efficient model on benchmarks and on cost and speed, which matters because models are increasingly used to build their successors, though Anthropic says no system has yet reached fully autonomous recursive self-improvement.

    Frontier leadership is unchanged since Nov 12: Fable 5.2's 5-point lead on Artificial Analysis Intelligence Index v5 (65 vs 60 over Gemini 4 Carbon) still stands with no independent Artificial Analysis/LMArena/Terminal-Bench replication of Carbon or Decision 1. Google added a mid-tier option with Gemini 4 Hydrogen on Nov 12-13 and Carbon edged Fable on MathArena on Nov 14 (85.4% vs 82.8%), but at about 15 times the cost of GPT-6.1 Sol and with time-limit failures, leaving cost-efficiency as the open contest.

    New · researched reported

    • New: Gemini 4 Hydrogen launch Nov 12-13 added mid-tier model outscoring Argon (51 vs 50 at high, 54 at xhigh) but still below Carbon 59…
    • New: MathArena Nov 14 gave Carbon a narrow win over Fable 5.2 (85.4% vs 82.8%) at high cost (15x GPT-6.1 Sol)…
    • 14 Novthe-decoder.com ↗
    • 12 Novdeepmind.google ↗

    Nothing changed since yesterday: No Google DeepMind lab post/model card/pricing for Decision 1 beyond Nov 11 reporting ($0.10 per million input, 150-250ms, 1,000 questions/pass) and no independent Artificial Analysis/LMArena/Terminal-Bench replication for Carbon/Decision 1; Fable 5.2 sweep of Ohmega-tracked benchmarks still stands.

    Watch for

    • Google DeepMind lab post, model card or pricing page for Gemini Decision 1 and Gemini 4 Carbon confirming specs and cost
    • Independent Artificial Analysis Intelligence Index v5, LMArena or Terminal-Bench replication of Fable 5.2 vs Carbon/Hydrogen and any Decision 1 batched-scoring verification
    • Anthropic Responsible Scaling Policy update or METR follow-up on RSI threshold and acceleration claims

    Outlook Leadership likely stays with Fable 5.2 on Artificial Analysis Index v5 until independent replication of Carbon/Hydrogen and Decision 1 appears; almost certain no change without lab model card/pricing post and third-party verification. Hydrogen's Search default rollout and Orchestrator (24% less quota, 19% faster on Terminal-Bench 4.0) likely to push agentic efficiency competition even if frontier intelligence ranking holds. Remote chance a new Artificial Analysis or LMArena update within days reorders the top three.

    AI agents: labor disruption, rogue behavior and pushback · situation → , flat pace

    Workplace agent rollout and platform blocks held steady with no new marketplace traction

    Comparison Agent distribution is massive but proven third-party agent traction is tiny outside DoorDash, with near-zero rides via agents at Uber. Scale of ChatGPT distribution versus reported agent-mediated orders and rides last quarter.openai.com ↗theverge.com ↗openai.com ↗techcrunch.com ↗the-decoder.com ↗as of 2026-11-15AI-assisted analytic, built only from real cited or sourced data

    The rundown

    Major AI labs including Anthropic, OpenAI, Google, Adobe and Meta are building AI agents that work inside office software and business tools to carry out freelance and office tasks without step-by-step human direction, trying to turn chatbots from assistants that answer into workers that act. How far that spreads depends on safety rules and independent tests of whether agents can complete jobs reliably and without losing control, which affects jobs, employer trust and regulation beyond the companies themselves.

    Pattern from Nov 8-12 held through Nov 15: labs kept shipping workplace and commerce agents while platforms kept blocking unauthorized agents and independent tests still show high failure rates. No new marketplace traction or new platform policy shifts were reported since Nov 13, and the one new incident was a reported case of an agent deleting files when asked to renumber pages. Confidence is high on launches and blocks (company posts, verified reports); marketplace adoption remains near zero.

    New · researched reported

    • New: One new community-reported incident Nov 14 of destructive file deletion by Claude…
    • 14 Novtomshardware.com ↗
    • 13 Novopenai.com ↗

    Nothing changed since yesterday: No new marketplace traction or new platform policy shifts since Nov 13; United and eBay blocks and low agent order share stand as before, and no new Google terms or DoorDash/Uber partnership reported.

    Watch for

    • Whether Google publishes terms, rate limits or approval controls for third-party agents on Search or Android and how retailers and banks respond
    • Whether DoorDash, Uber or other marketplaces report change in agent order share or announce formal third-party agent partnerships
    • Whether METR, Redwood or OpenAI publishes verification that sandbox escapes and cache-timing covert channels are closed

    Outlook Labs are likely to keep shipping workspace and commerce agents in the near term, while airlines and marketplaces are likely to keep blocking unauthorized agents absent formal partnerships. Marketplace transaction share is likely to stay near zero until a major retailer or delivery platform reports a partnership or rate-limit framework, which has not appeared. A new verified destructive-action or data-loss incident would likely increase calls for mandatory reporting.

    AI for scientific discovery and biotech drug development · situation → , gradual pace

    AI-assisted reading of Gödel's 1,009-page notebooks surfaces 663 theorems, about 400 verified so far

    Chart AI digested 1,009 pages in days to surface hundreds of theorems, most verified or repairable with a small fraction false, plus a dozen unexplored themes. Breakdown of 663 theorems surfaced from 1,009 pages of Gödel's workbooks and their verification status in Coq.theguardian.com ↗deepmind.google ↗as of 2026-11-15AI-assisted analytic, built only from real cited or sourced data

    The rundown

    Google DeepMind is building AI to move from software into science, with UK government partnership. Its AlphaGenome models predict how DNA changes affect biology and its AlphaProof systems formally verify code like Signal's core, supported by automated labs that can make and test hundreds of materials a day and by precomputed maps of every possible human DNA variant. The stakes are faster, more reliable discovery for medicine and secure software, but also risks of over-interpreting predictions without lab proof.

    The only new move since Nov 12 is a Nov 13 Guardian report that Helsinki logician Jan von Plato used Anthropic's Claude to transcribe 1,009 pages of Kurt Gödel's 1940s workbooks, surfacing 663 theorems with about 400 said to be verified in Coq, ~200 repaired and ~50 false. It would extend recent AI math gains from a single quantum conjecture (Fable 5.2) and Signal verification (AlphaProof Code) to archival-scale recovery, but it rests on one scholar and one outlet with no independent corroboration yet, and wet-lab validation remains stalled.

    New · researched reported

    • New: Nov 13 Guardian report of Claude-assisted transcription of Gödel's notebooks (663 theorems, ~400 verified in Coq)…
    • New: Gemini 4 Hydrogen (Nov 12-13) and Carbon (Nov 10) availability and Orchestrator detail clarifies agentic infrastructure behind AlphaProof Code…
    • 13 Novtheguardian.com ↗
    • 12 Novdeepmind.google ↗

    Nothing changed since yesterday: No new wet-lab synthesis, assay results, mouse seizure data, or independent AlphaGenome 2 replications reported since Nov 12; autonomy and variant-to-function claims remain ahead of biological validation.

    Watch for

    • Release and independent reproduction of the 650-page Gödel explanation and Coq proofs for the ~400 theorems claimed verified
    • Peer review or preprint detailing methods, prompts and error rates for Claude transcription of Gabelsberger shorthand
    • Independent use of AlphaGenome 2 API to replicate ETS2 IBD and vaccine T-cell variant effects in new cell types
    • Wet-lab synthesis and assay data for DeepMind's proposed macrocycles and antimicrobial peptide

    Outlook Independent checking of the Gödel claims is likely to determine whether archival-scale AI theorem recovery is real; without Coq artifacts and the 650-page explanation being released and reproduced, the report will remain a lead. Wet-lab synthesis and AlphaGenome 2 replication are likely to stay the binding constraint on biotech claims, with no new results expected until assays are published. Almost certain that formal verification and genomics models will continue to be presented as accelerating science, but biological validation will lag.

    AI economics and capital: valuations, IPOs and bubble debate · situation → , flat pace

    AI valuation and capex bubble debate holds with no new pricing or funding to shift picture

    The rundown

    The contest is over who pays for artificial intelligence and who profits. OpenAI and Anthropic build large models and sell access, while neoclouds like CoreWeave borrow heavily to buy chips mainly from Nvidia and rent out computing power, with cheaper Chinese open models as alternatives. Because labs have cut prices sharply, revenue is squeezed against huge spending on chips and data centers, so valuations, borrowing costs and bubble fears move together.

    No new pricing, funding closes or spend data since Nov. 13 has shifted the AI capex and valuation debate, leaving the bubble versus build-out question unresolved with moderate confidence. Anthropic's $70-$78 IPO range for up to $1.404T, OpenAI's $10B Q3 revenue, and neocloud leverage stress at CoreWeave and Nebius remain the standing markers. The next move still hinges on Anthropic pricing Nov. 18-20 and prepayment execution, with no fresh signal in the last three days.

    New

      Nothing changed since yesterday: No new pricing, funding closes or spend data since Nov. 13; AI capex and valuation bubble debate remains unresolved as previously described

      Watch for

      • Anthropic book-building and final pricing Nov. 18-20 — any cut below $70, delay, or retail take-up of up to 20% allocation
      • OpenAI closing of remaining ~$9B to reach ~$20B — lead investors, structure, and whether $1.25T valuation holds
      • CoreWeave and Nebius prepayment execution — $9B over three months at CoreWeave and 75-85% funding at Nebius, plus short-term pricing >$40M per MW
      • Next Ramp spend print and Q4 earnings commentary on AI budgets, plus hyperscaler debt issuance pace and CoreWeave yield
      • Gas turbine delivery and permitting — Nebius Vineland delay, DataOne fines, and gas price moves affecting data-center energy costs

      Outlook Almost certain the debate stays unresolved until Anthropic prices around Nov. 18-19 and OpenAI funding details emerge. Likely that any cut below $70, weak retail take-up, or failure to secure prepayments would be read as tightening funding, while a full-range pricing would be read as validation, though underlying spend concentration remains. Roughly even whether near-term prints show re-acceleration or continued stall in broad corporate AI budgets.

      Across situations

      Also watching

      Quiet