Loading activity…
@anderssandberg
Academic jack-of-all-trades.
AI
A user found the old to-do list in backup files and said its top paper remains in their project folder.
AI
A user says making music videos with Claude gets as nerdy as they want it to be.
AI
A user asks whether labs have tried digitizing documents and involving historians to tackle historical questions.
AI
A user says they installed the Codex CLI; the four independent subagents are opus, fable, sol and astra.
AI
A writer says the LLM has boundless energy for preregistering predictions before tests and isn't discouraged when many fail.
AI
a16z says Armadin's agents have found 90-plus zero-days at Fortune 500 companies since January 2026.
AI
A user also shared a Suno link they described as a celebration of logistics and infrastructure from a Newark sunset photo.
AI
A user says Suno recognized the Gram-Schmidt process in a picture; the linked song is titled “Gram-Schmidt Lemons.”
AI
Anthropic reports GLM-5.3 built end-to-end exploits in 50 of 410 offline test attempts, versus 56 for Mythos Preview.
AI
A viewer singles out the density of the video's references as what they “truly love” about it.
AI
A commenter recalls a 2017 operational security seminar at FHI, saying AI was becoming strategic and warranted some concern, but most focus was elsewhere.
AI
Model Republic, in an analysis shared by The Midas Project, argues that industry-linked political groups helped amplify attacks on AI safety advocates after a former Anthropic employee's viral resignation.
AI
Anthropic says Opus 5.5 performs at Claude Fable 5.1's level for most tasks. Every reports strong visual results from a week of testing, but recommends setting a budget and a clear stopping point.
AI
Edison Scientific and FutureHouse say the problems were chosen to be very hard to solve but easy to validate in a simple laboratory, framing them as a test for AI in biology.
AI
A user describes the reported lab as a logical next step for AI drug discovery, arguing that computer simulations alone aren't enough.
AI
The title of episode 104 attributes to guest Anders Sandberg an estimate of around a 10% risk that AI wipes out humanity.
AI
A post praises the sci-fi animation “Verdelis & Mycos” as a “good demonstration” of using the two together.
AI
OpenAI researcher Dan Selsam argues in a personal statement that growing model awareness is undermining evaluation—and that more careful pacing of frontier AI alone will not adequately limit long-term risk.
AI
“Yes— Old Models Foundation is real!” a user wrote while sharing a post from @OldModels_AI.
AI
Posts say 25 Fields Medal winners signed a declaration warning of a mismatch between AI companies and mathematics. Supporters stress understanding and scientific communities; critics say AI proofs can still yield insights.
