merry christmas

what a time to be a part of this team 🙏 incredibly grateful to see this project live - massive shoutout to Andrew Peng

preference fine-tuning, you can tell the model what to do and what not to do 😎

finished my reading list today: 1. zen & the art of motorcycle maintenance (tbh would prefer other forms of philosophical literature) 2. lora vs. full fine-tuning: an illusion, shoutout Scott Cory for the rec 3. random blogs on how to get better at craps (*pre-vegas prep)

🎉Applications are open for Neo Accelerator 2025, with incredible mentors, competitive funding, an immersive bootcamp, and a dedicated workspace for the entire program. If you’re a technical founder or team with a bold vision for the future, apply now! (1/6)

happy sbd

walking through sf is worse than the stairmaster how do i optimize for flattest route on google maps

GPT-4.5 is the first time people have been emailing with such passion asking us to promise to never stop offering a specific model or even replace it with an update great work Kai rapha Amelia (Mia) Glaese

Small fine-tuning update: you can now attach and filter by metadata for fine-tuning jobs: platform.openai.com/docs/api-refer… Finally have a way to filter out all of Steven Heidel’s test jobs 😌

🚨 NEW INVESTMENT 🇺🇸 Sequoia led a $190M Series C in Peregrine. Peregrine makes software that helps mission-critical organizations (public safety, law enforcement, government, enterprises) integrate and analyze their data, and build operational applications on top of it.

super dope stuff Deniz Birlikci

check if your model passed the vibe check

the way that SF flirts with you

sometimes it feels like the only food worth eating is either 6 dollars or 60. everything inbetween an overpriced (at best) parody of the real thing. life is all about the poles: the 1 star or the 4 star, street taco or omakase, plastic stool or leather banquette

my code wasn't working so i decided to do some writing instead 🙃 revived an old draft karennli.substack.com/p/the-painters…

Remember reinforcement fine-tuning? We’ve been working away at it since last December, and it’s available today with OpenAI o4-mini! RFT uses chain-of-thought reasoning and task-specific grading to improve model performance—especially useful for complex domains. Take

The DHS order hurts not just current Harvard students but grads too. Even years after college, grads on OPT visas may get kicked out of the US. If your status is in jeopardy, @Neo can help you join a startup or start your own, get an O-1, and stay in the US. A few options: 🧵

got a 128gb computer at work a couple weeks back, qol is noticeably better when cmd click actually works in a monorepo 🙏