Yannick Mühlhäuser retweeted
AI Verification technology will be among the most important things humanity ever builds.
If you want to catalyze this insanely important industry, I recommend you ping @woj_zaremba and apply for this gig!
“Trust, but verify”
– Ronald Reagan
In 1959, the US launched Project Vela to monitor nuclear-test detonations around the world—in space, in the atmosphere, and underground.
Such technology subsequently helped enable international treaties on nuclear testing.
We will need something similar for the AI era.
The nuclear analogy is far from perfect, because AI carries insane number of positive applications. However, it makes a point for what is needed for high stakes, international agreements.
We need technological solutions that make policy agreements possible. These solutions would allow humanity to verify fundamental properties of AI systems and their infrastructure: Where are chips located? Are they being used for training? Is AI being used for cyber offense?
Humanity is in the early days of developing such technology, and we need an entire ecosystem to support it.
The OpenAI Foundation is hiring a Program Director to support such an ecosystem: fund teams, attract talent, commission red-teaming, and drive deployment in data centers.
We are looking for a builder with strong technical judgment and the instincts of a general manager—someone who can turn promising research into technology that countries can rely on.
openaifoundation.org/careers…
Yannick Mühlhäuser retweeted
OpenAI's version of the ECI, the OECI, shows no change in trend like Anthropic's AECI with Fable 5.2
Yannick Mühlhäuser retweeted
List of prominent people who have recently come out in favor of ~stopping further AI development. 🧵 (1/n)
The OpenAI Foundation is hiring an AI Verification Program Director!
AI verification is one of the most important fields of our time: verifiability will largely determine which agreements on AI development we can and cannot strike, especially internationally.
Apply here, or share with people who'd excel in this role: jobs.ashbyhq.com/openai-foun…
Yannick Mühlhäuser retweeted
Ambassador Susan Rice (@AmbassadorRice) spoke alongside @So8res and @ForHumanityPod in DC:
"As a former National Security advisor, the 'but China' argument is BS"
Yannick Mühlhäuser retweeted
A minority of French people think an extramarital affair is morally unacceptable. Germany is far behind with 60%.
Yannick Mühlhäuser retweeted
Nice argument, but I’ve already released a 12M-line lean formalization that you are the soyjack and I am the chad.
Yannick Mühlhäuser retweeted
Barack Obama: AI is reaching the point where it no longer needs humans to teach it. The era of exponential intelligence growth is upon us.
“The machines are starting to be able to teach themselves. They don't have to wait for humans to start figuring things out.”
“We're now at a stage where it's 50-50, and if the trend lines continue, the expectation is it might be 90-10.”
“We're now hitting the hockey stick trajectory where the models are going to get smarter and smarter at an exponential pace.”
“If they start setting their own agendas, you may get a misalignment between what they want to do and what we want them to do.”
“And that gap can be dangerous.”
Yannick Mühlhäuser retweeted
Anthropic has confirmed in an updated S-1 today that it has determined that the Autonomy 2 capability threshold has been met, based on the “dramatic acceleration of overall AI R&D progress” condition, given that it applies where Anthropic "observes or expects" accelerated progress. Anthropic further states that it does not make this determination based on current models.
sec.gov/Archives/edgar/data/…
Ah yes, AI for farmed animal welfare
Norway now has more than 2,000 underwater AI laser robots shooting lice off its salmon
they're made by a company called Stingray, and they already guard more than 50 million salmon and trout 24/7
the reason is sea lice, tiny parasites that latch onto salmon and feed on them
as few as 11 can kill a small fish, and one pen can hold 150,000 salmon
the usual fix is pumping the fish out and blasting them with warm water or water jets to knock the lice off
but the handling is so rough that Norway's fish health report ranks it the top cause of death in full-grown farmed salmon
so instead, Stingray puts a robot inside the pen that hunts lice one at a time, all day:
> two cameras watch every salmon that swims past
> AI checks two frames to make sure it's really a louse
> then a green laser fires from up to 2 meters away
> the salmon's scales reflect the beam, so the fish swims off fine and the louse is dead
we now live in a world where AI lasers do pest control on your dinner lmao
LLMs will be amazing for historical research.
Been running Opus 5.5 agents over the notebooks of Gauss, Riemann, Hilbert and Grothendieck. around 12,000 pages so far. Heres the best finding so far:
Riemann's 1860 shock wave paper is famous for getting the energy part wrong however his unpublished sheets at Göttingen have it right the Rankine Hugoniot conditions a decade or two before Rankine and Hugoniot plus results usually dated 19061937, all in his own notation. A second sheet in another folder has the same physics.
There are probably thousands of potential undiscovered stuff in the research papers of the smartest ever to live imo all of them should be made public and reviewed by ai years of work can probably be transcribed by agents in days
51
Yannick Mühlhäuser retweeted
I think this is roughly the right conceptualization of the timeline for the recent history of AI
Yannick Mühlhäuser retweeted
Kimi K3.1 is outperforming on TerminalBench, because for the most difficult tasks, where all frontier models underperform, it has knowledge about the benchmarks hidden tests.
And the model's performance easily breaks if you slightly change the tasks to disagree with the initial hidden tests.
Kimi K3.1 recalls instructions it was never given on Terminal Bench. They match the hidden tests. K3.1 while solving: "Instructions likely distilled pitfalls from previous agents." 🤯
When we edit a task to contradict them, K3.1 refuses the new instructions and fails. Other models follow the edit.
Is it terminal(ly) benchmaxxed? 🧵
Yannick Mühlhäuser retweeted
Contrary to popular belief, AI safety super PACs have far outspent their opponents in the 2026 midterms
Yannick Mühlhäuser retweeted
After two wonderful and rewarding years at @ARIA_research, I’m excited to share that I am joining the @FoundationOAI as Program Director of Formal Methods.
Everyone in AI companies should read this and lobby for doing it. There is a lot to do.
If just a few lab staff took on verification work full-time, verifying a US-China pacing deal could get much more feasible within months.
Here are 8 specific things lab staff could start on today to help make international verification possible: 🧵
4
155














































