Daily Brief Nov 6, 2026
How we checked every item today except "Also happened"
For readers who want to dig in. What happened, in brief, is on the Today tab.
This issue draws on material from November 5 to early November 6; main-line item 1 is also checked against Anthropic's November 2 amended IPO filing and Daniel Kokotajlo's November 3 post, and item 4 against a Marketplace report of October 29. Named commentary item 2 is a November 1 post, and the From the archive item comes from a SecondSource call opened August 2 and settled October 3. The report source is SecondSource's own research brief. Last night's collection took in 885 pieces, and this issue uses 25 outside sources with checkable links; the later sections also draw on material from the past week.
1. [Today] METR subjectively estimates Anthropic's models are gaining capability about twice as fast as from March 2024 to October 2025, and says Anthropic redacted most of the evidence METR found most useful
Not proven yet： both reports share one set of authors, with no second independent line of evidence
How we checked
The sentences from Anthropic's risk report and Responsible Scaling Policy all come as quoted in METR's review. The two reports have the same authors and come from the same evaluation, so they don't count as two sources, and there is no second independent line of evidence so far.
2. [Today] OpenAI writes down what its own campuses will give communities and asks states to make large power users bear grid costs, while campuses held by partners stay outside the pledge
Not proven yet： the documents don't say what share of OpenAI's capacity is owned and operated
How we checked
These terms come from the full text of the two PDFs. The announcement and both PDFs are all OpenAI's own documents, and the Effingham County and Pike County figures also come from the compact. Both documents cover only future sites OpenAI owns and operates, so campuses held or run by partners are outside them. Neither document says what share of OpenAI's capacity is owned and operated.
3. [Today] OpenAI releases GPT-6 Cyber and says the model beats GPT-6 Astra on most security tasks, while one third-party test in the system card has Astra slightly ahead
Not proven yet： most of the figures are OpenAI's own measurements
How we checked
The only independent figures are Irregular's, as cited in the system card. Whether Irregular has a commissioning relationship with OpenAI is unclear so far. The ExploitGym and Irregular figures appear only in OpenAI's announcement and system card, with no other source to compare them against yet.
4. [This week] (event date November 4) GPU cloud provider Lambda says it will move into an Alabama campus with an estimated US$33.1 billion investment, with power still awaiting regulatory approval
Not proven yet： the US$33.1 billion is Lambda's own estimate
How we checked
The amount and timeline come only from Lambda's announcement and haven't been checked by a third party. The county's lack of zoning and the count of signs come from Marketplace alone, and the signs are the reporter's observation.
[Today] NVIDIA argues that inference can scale with almost no added HBM capacity, and its two same-day posts put Rubin at up to 30 times GB300, per dollar in one and per megawatt in the other.
How we checked
NVIDIA's two same-day posts each self-report Rubin at up to 30 times GB300, one on throughput per dollar and one per megawatt, and the two can't be converted into each other.
[Today] In Microsoft's internal test of Cobalt 200 on chip-design software, Xcelium, the tool that simulates chip designs, reaches 4.15 times the per-logical-core throughput of Cobalt 100.
How we checked
[Today] (November 5) Ryan Greenblatt roughly estimates that AI software R&D already runs about 4.5 times as fast as with no AI automation at all, and admits being wrong to expect 1.15 to 1.3 times for this year.
How we checked
Greenblatt derives the 4.5 times from METR's roughly twofold figure as a rough personal estimate on three of the author's own assumptions, and discloses working at METR. Greenblatt's input is METR's own number, so this doesn't count as a second independent source.
[This week] (November 1) John Burn-Murdoch, chief data reporter at the Financial Times, says many people cite a productivity chart as proof that AI is showing up in the data, though the reporter made that chart to rebut exactly that claim.
How we checked
[Today] Anthropic analyzes 207,000 research conversations and projects that, once existing capabilities spread, research output on a fixed budget reaches about 1.96 times in computer science and about 1.35 times in the arts.
How we checked
[Today] According to a preprint's abstract, a finance evaluation the authors built requires passing 15 checks and satisfying a simulated user, and four coding agents plus one model without tools failed all 150 runs.
How we checked
[Today] Thinking Machines, the AI-model company founded by Mira Murati, releases Inkling 2 Small, says the model was trained entirely in NVIDIA's 4-bit floating-point format, and plans to release the weights after outside safety testing.
How we checked
[Today] In Perplexity's own test of the medium effort level of its Agent API, the score on the public research benchmarks BrowseComp plus DSQA rose from 55% to 77% while cost fell nearly 90%.
How we checked
Agent scores and costs swing widely with the step limit, and this is Perplexity's own evaluation with no third-party rerun.
[Today] Google makes Gemini Omni 1.2 Flash generally available, raising the single-clip limit from 10 seconds to 15 seconds and pricing 720p 30% below 1.1 Flash.
How we checked
[Looking back] (SecondSource call of August 2, settled October 3) Investor Bill Gurley predicted analysts would spend eight straight quarters asking how much of CoreWeave's unsold compute NVIDIA has committed to buy, and on the first call no one asked and the company gave no figure.
How we checked
(card)
Sources & accounting
The past 24 hours. 885 new pieces came in from November 5 to November 6: 530 academic papers, 149 social-platform posts, 70 linked articles, 64 company and personal blog posts, 55 from a second academic-paper feed (counted separately from the academic papers), 8 subscription newsletters, 7 company filings and 2 show transcripts. Of those 885, 276 were read in full, 228 were screened out as off-topic (3 of them before fetching), 6 were logged with a stated reason for not reading closely, 374 haven't been read yet, and 1 is being refetched.