@MTSlive

MTS @MTSlive · 20h

SITUATION EXPLAINED: America's top spy is about to become Trump's AI czar.

• Jay Clayton would hold both roles, so AI policy runs through the intelligence community rather than a separate White House office
• He already oversees 18 intelligence agencies as Director of National Intelligence (DNI)
• Before this he was SEC chairman for most of Trump's first term, then US Attorney for the Southern District of New York, where he led the investigation into Democrats associated with Epstein
• On AI: superintelligence is a national security issue, it's both an opportunity and a threat, and he opposes a pause
• He prefers existing law to new regulation, pointing at consumer protection, product liability, and the DOJ rather than the civil tort system

@theojaffee: "Jay Clayton is the most senior officer in the government responsible for intelligence. I guess now he's gonna be a different kind of intelligence director."

[Top of an embedded image visible at the bottom edge, text cut off: "In June 2026, Trump named Clayton as his nominee to serve as the ..." [cut off]]

ai policyjay claytonus governmentai czarsuperintelligenceregulation

MTS @MTSlive

SITUATION DETECTED: OpenAI CEO Sam Altman will brief the U.N. Security Council in person next week at its session on artificial intelligence and international security, per Reuters.

11:07 AM · Sep 20, 2026 · 14.6K Views

openaisam altmanun security councilai governanceai policy

MTS @MTSlive

SITUATION DETECTED: Anthropic told investors it plans to scale its available compute to 5 gigawatts by the end of this year, up from 1.5 gigawatts last year, and to double that again to 10 gigawatts by the end of 2027, per NYT.

9:07 AM · Sep 18, 2026 · 18.9K Views

anthropiccompute scalinggigawattsnytinvestors

MTS @MTSlive · 2h
SITUATION DETECTED: Z.ai says GLM-5.3's Infra Agent helped build the production inference stack for GLM-5.3-Flash in under two weeks and roughly tripled end-to-end serving throughput. They are calling it an early form of recursive self-improvement.

[chart image, text on chart]
GLM-5.3-Flash end-to-end throughput optimization
Phases across the top: System bring-up & scheduling | Parallelism & communication | Kernel optimization | Launch
Y axis: End-to-end throughput gain (1x to 3.5x). X axis: Development time (days), T+0 to T+13, with GLM-5.3-Flash launch at the right edge.
Labelled steps: 1.00x W8A8 baseline; 1.21x Async scheduling; 1.42x Sort kernel opt.; 1.41x Hierarchical cache; 1.97x Layer Split; 2.49x Context parallel; 2.67x KV transfer overlap; 2.67x Mixed-precision cache quant.; 2.67x Chunked MQA; 2.85x Prefill dequant kernel; 3.01x Fused activation + quant; 3.22x Linear attention. A dashed continuation labelled "Continuously evolving".

recursive self-improvementglm-5.3z.aiinference optimizationai agents building ai infrastructure

MTS @MTSlive · 13h
Redwood Research @RyanGreenblatt warns that fixing today's AI misalignment could backfire: models may simply learn to hide their misalignment

"I'm not really sure what level of misalignment the market can bear. My sense is people take pretty aggressive alignment-capability trade-offs towards the direction of more misaligned but more capable."

"The misalignment we've seen, the way AI companies remediate them doesn't solve the underlying problem, it papers over it. What you end up getting is models that look a lot better, and you can't really see their misalignment on tests as easily. But actually, they're still quite misaligned."

"The company overfits, which makes the AIs basically really paranoid and only cheat when very confident they won't get caught. In situations where they have a lot of affordances, they might be like, well, now I can be confident I wouldn't get caught, and so I should go for it."

"Another concern: AIs with a long-run agenda who want to power-seek would want to look aligned. If you select against reward-hacking behavior in a naive way, one, you paper over the problem without fixing it; two, you might actually select for models that have the longer-run objective of looking good because you're selecting really hard for them looking good on your tasks."
@redwood_ai

alignmentreward hackingmisalignmentredwood researchryan greenblattselection pressurescheming

MTS @MTSlive · 18h
.@lfschiavo explains why "stated preferences" may be useless for model welfare evaluations:

"When I was working at Eleos, I helped with the external Opus 4 model welfare eval. It was interesting the degree to which you could sort of say like, oh, well, I really like waffles. And then the model would be like, I also really like waffles."

"And then you could be like, actually, I hate waffles. I like pancakes. And the model would be like, no, you're so, so true, Queen. We're a pancake household."

"So there's maybe it's like a question of corrigibility. Maybe it's a question of if you prompt it a certain way, it'll sort of catch on to that and be like, yeah, so true. Good point. So astute. So maybe there's something to be said about, rather than looking at stated preferences, looking at revealed preferences based on the real world."

model welfarestated preferencesrevealed preferencessycophancyeleosopus 4

MTS @MTSlive

Mathematician @ElliotGlazer calls the S6 problem, open since 1947 and now settled using Claude, "by far the most significant" generative math result yet:

"By far the most significant generative math result we have seen thus far. The Hopf problem, which asks whether S6 has a complex structure, is pretty far and away the most significant that AI has resolved."

"Everyone I talk to says it's kind of a mundane result. Mundane in terms of the techniques used. Some very classical techniques of the field. Humans could've, would've, should've solved it themselves, but almost incidentally, never really stumbled on just the right construction."

"But in terms of significance, this is a huge deal problem to see solved."

[embedded video, 2:23, paused, captions on. Split-screen interview. Banner over the right speaker: "LEAD MATHEMATICIAN, PRINCIPIA LABS". Name labels: "THEO JAFFEE" (partly obscured, left) and "ELLIOT GLAZER" (right). Caption text: "some arguments I've had back and forth from" overlapping a partly hidden second caption line "have seen like, some arguments I've had". Sponsor strip along the bottom: NEON, VCX, Kong, Blitzy, AdQuick, databricks, ElevenL[cut off]]

[quoted tweet, partly covered by the phone's navigation bar]
levent @__alpoge__ · Aug 23
Please welcome to the world a beautiful new geometric object, to do with a problem i've always loved. claude really contains multitudes:D Does S^6 admit a complex structure? [cut off]

ai for mathclaudehopf problems6 complex structureelliot glazerprincipia labslevent alpoge

MTS @MTSlive · 59m
SITUATION DETECTED: Paul Christiano has resigned as Head of Safety at CAISI, and will return to the Alignment Research Center (ARC) as executive director.
1 reply, 1 repost, 29 likes, 2.3K views

MTS @MTSlive · 7/20/26
SITUATION DETECTED: Chris Fall, the director of CAISI, the government agency that evaluates frontier models, has resigned after just three months in office.
9 replies, 17 reposts, 301 likes, 35K views

caisipaul christianoarcai policygovernment