Founding Eng @infinity_ai_ ||| SF ||| 🎸🎹🎛️🎙️🎭🥾🏞️🧗‍♂️📚🤖🧠💭🧑‍💻

San Francisco
Joined April 2009
do one thing every day that cringes you
2
1
20
2,070
"luke is human and can make mistakes"
this annotation but for human sent messages
10
this annotation but for human sent messages
1
23
"ender's game? the book about eval envs that turn out to be real world envs?"
3
136
How are people having Claude decompile proprietary games without getting denied by the model?
1
1
8
277
belief that the world is generally zero-sum has always been cancerous, but now it has alignment implications too!
Insightful concluding remarks on AI alignment at #COLM2026 by Noam Brown @polynoamial @OpenAI @AnthropicAI @MSFTResearch
105
the hottest new programming language is gibberish
The hottest new programming language is English
2
69
hey opus 5.5, grand perspective is being kinda slow, can you re-create it
1
1
83
insane
How do bacteria build their flagella? We’ve put together an animation of Salmonella flagellum assembly based on cryo-EM and X-ray structures. From the membrane-embedded machinery to the growing filament, here’s how the parts come together, step-by-step.
4
362
“It’s in the convex hull” is the new “it’s priced in”
6
8
2
179
5,040
a sort of "failure mode" with AI: it impresses YOU how fast you build something, how little time or effort it took. That impresses NO ONE ELSE — they just see, THEY ONLY CARE ABOUT, the final result
44
"provenance"
1
51
"company"? oh you mean "domain specific harness"
2
74
the zone of proximal development strikes again
🎯 Learning is most effective at the frontier of capability: problems that are too easy or too hard teach nothing. For LLM reasoners trained with GRPO this is literal: problems the model always or never solves give zero gradient. We introduce Frontier Learning👇🧵
1
64