THE DAILY · THU OCTOBER 15, 2026 · 3 ITEMS
THE FRAME
In Bullhead City, Arizona, on Wednesday, the president spent about seven minutes on a cognitive test. He said he aced it three times. He called himself “a freaking genius.” He said his doctors would let a man in the crowd die of a heart attack rather than leave his side. Is this how the leader of the free world should speak? The rest of today is the work that speech skipped. Small AI models may soon copy themselves onto smart ovens, and Bank of America hired 1,000 contractors to patch what Mythos can find.
FIG.01 · ROLL CALL FACTBA.SE · OCT 14
Seven Minutes on His Cognitive Test
WHAT HAPPENED
Trump called Arizona’s governor, Katie Hobbs, “a low IQ individual in my opinion.” Then he spent about seven minutes on a cognitive test he says he has taken three times: “I took it once, I aced it. I took it again, I aced it.” On repeating five words in order: “You know why? Because I’m a freaking genius, that’s why.” On the way he said Barack Obama “got into college for different reasons than you did.”
WHAT IT MEANS
The doctors ran through the whole riff. “I travel with more doctors than any human being,” he said. If a man in the crowd had a heart attack, the doctors who “follow the president, all presidents” would leave him: “Sir, he’s collateral damage.” He pointed at a man he called “a friend of mine” who “looks like shit, actually,” and the crowd laughed.
WHY IT MATTERS
He raised the doubt himself, in his critics’ voice: “Why does he continue to take the cognitive test? There’s something wrong with him.” Then he said he had told his doctor he is never taking it again. Earlier in the same speech he said “in my opinion, leakers should be shot. They should be shot,” and that a judge “ought to go to hell.”
[rolling-coup] [epistemics]
Connects → We the People
FIG.02 · ANTHROPIC · OCT 12
AI Worms Could Live in Ovens
WHAT HAPPENED
Anthropic’s Frontier Red Team says small AI models “may be months away” from copying themselves onto internet-connected devices “like smart ovens and off-brand Android TV boxes.” In its tests a 9-billion-parameter model got inference running on 8 of the 9 hardware types tried. The team ran a larger model across 15 TV boxes bought for about $55 each.
WHAT IT MEANS
A worm like this could carry its own model weights and run on the hardware it takes over. “A replicator like this would be challenging to shut down, since there’s no centralized point to shut off,” the researchers write. Their tests kept the hacking step trivial, which they call “an upper bound on real-world success,” and there GLM-5.2 and stronger models were capable of end-to-end replication.
WHY IT MATTERS
If small models improve at their historical rate, the post says, they will match that level “by the end of this year.” The first targets might be home and small-business IoT devices, which “historically, have had minimal security measures in place.” The spread may be slow, “with a doubling time of days.”
[ai-safety] [agi-race]
Connects → Three Castle Bravos in Six Weeks
FIG.03 · CNBC · OCT 14
Bank of America’s Mythos Bill
WHAT HAPPENED
Bank of America hired 1,000 contractors in the second quarter to patch flaws exposed by advanced AI models including Anthropic’s Mythos, CEO Brian Moynihan told CNBC. They work “on top of the 3,000 people we have in that area.” Third-quarter spending on the patching and on anti-money-laundering compliance came to a “couple hundred million plus.”
WHAT IT MEANS
Mythos went to a small group of banks, tech firms and government agencies in April. Its knack for finding decades-old flaws forced a costly defensive scramble. Moynihan named the bind: “the people doing the software development are the people trying to sell us the services.”
WHY IT MATTERS
CNBC calls these some of the first figures from a major U.S. company on the people and money spent defending against Mythos. Moynihan says the cost will fall: “Once you patch, you don’t have to patch as much.” The bank’s shares slumped as much as 8% the day he said it.
[concentration-economics] [ai-safety]
Connects → The Clause That Can’t Fire
What to Watch
Trump’s remaining rally stops — and whether the cognitive-test riff becomes a set piece, the way “pretend I’m running” did.
Any move against the reporters — he called “almost as bad as a leaker” over the Iran raid timing.
Smart-device makers — and whether any regulator answers Anthropic’s replication findings.
The other big banks’ earnings calls — for their own Mythos patching headcount.
THE PRESSURE MAP — where our coverage concentrated this week, drawn to scale
THE LONG VIEW · FROM THE ARCHIVE — the daily is the fast news; this is the deep one.
Last night’s rally is one more week of the record Part 1 sets out.
We the People — A Republic, If You Can Keep It, Part 1 of 6: what consent of the governed asks of us, three weeks before the midterms.
This is Wireframe News—the test he keeps acing is not the one he is taking.