hi again my dear friend

So excited for my Summer'25 Lineup 😤 Begins with the YC AI Startup School, in SF the next 5 days Then, starting as a Research Scientist Intern at Meta London for the next 4 months Finally, Presenting 5 projects at ICML in Vancouver Looking forward to meeting ppl, hmu!

MegaSaM got an award! Big congrats to the team!!!!! 🥳🥳🎉🎉 Zhengqi Li, Richard, Forrester Cole, Linyi Jin, Qianqian Wang, Vickie Angjoo Kanazawa, Noah Snavely

My notes on Anthropic's substantial essay about how they built their multi-agent research system, which has finally talked me around to taking multi-agent LLM prompt engineering seriously simonwillison.net/2025/Jun/14/mu…

Association for Computing Machinery Council has reaffirmed yesterday that all ACM publications and related research artifacts in the ACM Digital Library (ACM Digital Library) will be fully openly available after 1-1-2026. Computing science material of the highest quality freely available to all! #OpenAccess

“Having a good time” at #CVPR2025 :)

Terence Tao says today's AIs pass the eye test -- but fail miserably on the smell test. They generate proofs that look flawless. But the mistakes are subtle, and strangely inhuman. “There's a metaphorical mathematical smell.. it's not clear how to get AI to duplicate that.”

Last chance to check out the #CVPR2025 AI Art Gallery in Exhibit Hall A1 🤖🖼️ Open until 3pm #CVPR2025 #CVPRAIart

AAAI'26 (abs): 40 days. WACV'26 (R1,reg): 26 days. WACV'26 (R1,paper): 33 days. AAAI'26 (paper): 47 days. 3DV'26: 63 days. WACV'26 (R2,reg): 89 days. WACV'26 (R2,paper): 96 days.

It’s a hefty 206-page research paper, and the findings are concerning. "LLM users consistently underperformed at neural, linguistic, and behavioral levels" This study finds LLM dependence weakens the writer’s own neural and linguistic fingerprints. 🤔🤔 Relying only on EEG,

LLMs are massive cheating machines. All the intelligence comes from the millions of intelligent human beings that wrote the texts that LLM practitioners scrape from the internet. There is no new science of intelligence in LLMs. Linguists have known for a long time that

⚠️ Everybody who uses LLMs should read this thread: “LLM users consistently underperformed” on multiple cognitive and neural measures.

I don't wanna say 'I told you so', but I told you so.

The #Turing test was designed to test for artificial intelligence, but it ended up testing for human gullibility.

One could not believe that, in the practice of “artificial intelligence” in the past decade, how many different names people have given to the same thing, or have reinvented the same thing so many times, often even in a lesser form! Deep networks, for example, are largely

"Once an irrational idea is embedded in a culture, it creates psychological pressure on any member of that culture not to allow themselves to question that idea: people whose respect you psychologically depend on will despise you if you do. So an anti-rational meme emerges."

You don’t need a PhD to be a great AI researcher, as long as you’re standing on the shoulders of 100 who have one.

This solo founder built an open-source competitor to Perplexity with no team, no funding, and no permission. What started as a weekend project now powers over 1M searches, with 60K+ monthly users. This is the story of Scira AI and how Zaid Mukaddam built what others wouldn’t.