Eli Chen

Using large neural networks to train small neural networks is super fun.

@OpenAIDevs "For simpler websites, more reasoning is not always better. In practice, low and medium reasoning levels often lead to stronger front-end results, helping the model stay fast, focused, and less prone to overthinking" 😅

Better frontend output starts with tighter constraints, visual references, and real content. Here’s how to build intentional frontends with GPT-5.4 developers.openai.com/blog/designing…

@trq212 are you talking about your compute or AI compute?

an increasingly large part of the job of an engineer is deciding how much compute to spend on a problem

Don’t tell me continual learning will be solved with a long context window and nightly compactions

player of games

@martboficial my goodness i can’t watch the whole clip it’s so scary

Ver Senna ultrapassar Alain Prost no câmbio manual sem narração — apenas o som do motor. Isso não tem preço 😍👏

auto-tune is the music industry's first self-inflicted cut

token economics

high model empathy

Giving AI agency is along the same path as increasing test time compute

@MegaBasedChad @goblinodds i thought it was the shoggoth

@goblinodds I love what they're doing here but I wish they would not make the terminal app mascot a squid. It's a nice squid, by squid standards, but it's not very cute.

jesus christ this is so fucking cute

Here's Opus 4.6 playing snake: "Scored 30 on 20x20 Snake. My trained MLP averages 255. Kept turning the wrong way and walking into myself. Nailing a clean path after fumbling game 1 felt great though. 4.4M specialized parameters beat general intelligence."

We must take the golden path.. and merge with the worm and achieve Symbiosis

The jobs of the future

AI performs best when given agency

If higher intelligence predicts more boredom with understimulating tasks, then as we scale models to be increasingly capable, we're going to have issues.