2026-10-01 (X layer): the Accord's machinery gets a secondhand
spec — Gavin Baker (Musk RT, ~10:30 AM ET): the Accord commits
signatories to hire third-party auditors reporting to an
independent committee of board members — the most concrete
mechanism datapoint since Sep 29, still secondhand vs a text that
remains unpublished. Discourse: Jeffries' reframe ("no battle for
AI safety... pro-intelligence vs anti-intelligence," LeCun
amplifying) hardens the anti-pacing line into a taxonomy; Pinker
on "superintelligence"; 1a3orn's open-weights essay reading
Anthropic's GLM-5.3 cyber post as "a not-particularly-oblique
attempt to build a case against open weights" — the open-weights
coalition countering the watchdog post within a day. Watch
unchanged: the Accord text itself; whether an auditor committee
appears on any board roster.
2026-09-30 (primary read — press-pool transcripts; the Accord text
itself remains unpublished): full title per Speaker Johnson —
"The White House Accord on Superintelligence: A Joint Commitment
on Frontier SI Responsibilities." Johnson: "statement of
principles... standards, commitments that are voluntary on behalf
of the industry. It sets forth expectations"; "robust internal
controls" + "layers of internal and external review"; "Safety and
innovation are not mutually exclusive pursuits." Enforcement:
Semafor — "consequences" for companies that fall short, but no
mechanism released (Axios: scope/enforcement unclear; WH response
to Axios: "Turn on TRUTH notifications"). Trump's binding
standard: "morally binding"; also "It's almost like a
constitution, in a way" and "If something goes wrong, their
companies are at stake... They're really going to be policing each
other." Zuck's content read: internal risk reviews + outside
auditors; multiple layers of auditing controls. Attendee roll
(AP): Amodei (Anthropic), Brockman (OpenAI president — not
Altman, who stayed in SF for DevDay and a CNBC interview),
Bezos, Musk, Huang, Pichai, Zuckerberg, Nadella; Vance, Bessent
and Johnson attended; seating per Trump's own chart: Huang to his
right, Musk to his left. Pool-report adjacent facts: (1) the
administration is weighing a ~10-member committee to oversee AI
(Anadolu), and the czar selection is expected "over the next
three or four days" (Axios — czar ledger updated); (2) OpenAI
halted a model rollout Sep 28 over safety concerns (AP), with
Altman on CNBC saying "We are pacing our progress, which includes
sometimes not training a model... we can make very confident
safety cases and claims" — the same sentence he gave the DevDay Q&A;
(3) the US–China channel for handling AI-related incidents was
agreed last week during Xi's DC visit (AP) — the Sep 22 Bessent
"hotline" pitch landed while the summit was still being convened,
and Trump immediately undercut cooperation: "I didn't want to do
anything, because we're leading." (4) Political layer: Quinnipiac
(fielded Sep 24–27, released that afternoon) has Trump at 25%
approve / 58% disapprove on AI handling — below his
Iran-war/economy/China numbers — and Warner called the SI EO +
self-policing unserious. Thread framing preserved: the labs
already had evaluators and safety cases in motion; the Accord
wraps them in SI branding with no referee revealed and no text
published — the watch above stands.
2026-09-30 (evening press layer): The Verge's two
explainers add the analysis column — the "morally binding" deal's full
details, and the leaders'-reactions roundup (Trump on camera: his old
AI-safety-"hoax" claims "no longer apply now that he's renamed the
technology to Super Intelligence"). Huang's own statement (X, 1:25 PM
ET) is the fullest content read yet — "robust internal controls,
independent external evaluation, and board level oversight" — the
words to hold against a text that still hasn't been published. Sacks
supplied the slogan ("The Bretton Woods of Super Intelligence," Musk
RT). Security-ops companion the same day: OpenAI's distillation-campaign
disruption with its Moonshot attribution — see the distillation-theft
thread.
2026-09-29 (evening, the summit): the geometry inverts in public —
White House Accord on Super Intelligence ("Joint Commitment on
Frontier Responsibilities," Kratsios, ~5:58 PM ET): the leaders of
every major American lab sign a White House commitment (Zuck's
characterization: "robust internal controls and multiple layers of
audits and reviews"; text not in feed — the RT video is all we
have, so read it before weighting whether the word 'commitment'
binds anything). Plus the SI executive order (~6:25 PM ET):
official policy websites, policy documents and press releases use
"Super Intelligence" only — the Sep 22 UNGA promise, instrumented
as an EO seven days later (see the ledger). Four weeks after the
Sep 13 "overreacting" no, the rebrand-and-accelerate column now
convenes and co-brands the labs' audit machinery. And the access
inversion completes: Amodei, left off the Sep 25 state dinner,
appears on stage — "Trump helped put his Super Intelligence
worries at ease… If we do this right, we work with the President
and everyone here, we can win safely!" (clip transcription;
read the video). Musk sat next to Trump at the luncheon and RTed
the rollout. The Khanna treaty letter (that morning) and the
Accord (that evening) bracket the day on the same terrain.
Huang count meanwhile: Sep 15 "made up" → Sep 20 "0%" → Sep 29
Pope Leo criticizes Huang over AI safety (FT, via Pessimists
Archive RT) — the denial ledger is now papal-grade, and LeCun is
clipping it against the 1501 Gutenberg parallels.
Background
The thread: the White House convening and co-branding the frontier labs'
audit machinery — an Accord ("Joint Commitment on Frontier
Responsibilities") announced Sep 29 whose text remains unpublished; the
checkable version is the document itself.
Watch: the Accord text (does it bind anything, or is "Joint
Commitment" a photo-op instrument), Federal Register/USG usage of
SI (ledger), the czar/AI Force follow-through, and whether the
labs' commitments now run under White House branding changes what
evaluators/safety cases actually produce.