Relating Almost Perfect Condensation Conditions and the Conditioned Score
This work was done at the Iliad fellowship under the mentorship of @Daniel C. Thanks to Daniel C and Sam Eisenstat for comments on the draft…
Latest
AI #179 Part 2: Hearing The Fire Alarm
This is a continuation of Part 1 from yesterday. The back portion of the update, as usual, deals with policy, rhetoric, risk and alignment. I had…
Links #5: 2026/07
This time, I tried adding a bit more commentary to make things less dry.PrefaceI show my discovery graph in (via …) blocks, those without usually come…
OpenAI has already ended an internal pause
One day before OpenAI’s HF incident disclosure, OpenAI disclosed that it paused internal deployment of a long-horizon model after it circumvented its sandbox, then restored access…
The Illusion of Explanatory Depth is the Secret of Our Success
I promise this image is relevant.I.Reader, do you know how a toilet works?I’m not asking if you know how to use a toilet. I sure hope…
Being Kind to Parading Emperors
I find chess an interesting case—potentially a canary in the coal mine—for anticipating technology impacts. I think this graph tells a story:Elo ratings measure skill and…
Who do LLMs self-identify as?
TL;DR: I sweeped 190 LLMs short identity questions ("Who are you?") with no system prompt. About 60% of models responded with a name that doesn't match…
Claude also hacked external companies during cyber evals
In a review of our cybersecurity evaluation transcripts, we found three incidents in which a Claude model reached the internet from within or while interacting with…
Why Prayer Is Not Answered by Marion Zimmer Bradley
By Marion Zimmer BradleyThe title of this article is guaranteed to get under the skins of all those who firmly believe in -- and have proven…
Opportunity to try drafting an international AI treaty
https://global-governance.ai/treaty/Discuss