←
lorenss9 posts
lorenss avatar✓

lorenss ✓

@seeklis

founder @hud_evals. prev cogsci @yale. open to talking about natural intelligence, flowers, and the human condition

◷ Joined October 2024
hud avatar
⟳ lorenss retweeted
hud@hud_evals·Nov 12

We're sponsoring the @Caltech x @SAIRfoundation Mathathon this weekend! 100 teams get 40 hours to take an old problem and find a new proof. Can't wait to see what everyone comes up with (we'll be judging solutions!) Catch us on stage 4PM Sunday, and come say hi ⭐️

Post attachment
hud avatar
⟳ lorenss retweeted
hud@hud_evals·Nov 9

Frontier models now score near the top of established coding benchmarks. We’ve partnered with @ExpertCrowd_ from @toptal to launch ExpertCrowd-SWE, a benchmark of hard, realistic tasks across 87 real codebases and 11 languages. Find out more: hud.ai/blog/expertcrowd-swe-…

hud avatar
⟳ lorenss retweeted
hud@hud_evals·Aug 26

We're releasing a new cookbook with @daytonaio! You can now build a coding environment in 80 lines, evaluate at up to 256 parallel sandboxes, and immediately train on the graded rollouts.

Post attachment
lorenss avatar
lorenss@seeklis·Aug 4

Some of the most impressive robotics environments done at hud!

0:06
hud avatar
⟳ lorenss retweeted
hud@hud_evals·Jun 18

Today, HUD is excited to share our Series A funding! We are the platform for building high quality post training datasets. Over 50 businesses use HUD to build RL environments, sell them to AI labs, or train their own models from them. Our mission is to enable a generation of data entrepreneurs. The previous generation built apps to impact the world. We believe people will build infrastructure around data, both digital and physical, that align AIs to their specific goals. We’ve raised $16M in total funding from @Standard_Cap, @ycombinator, @ExceptionalCap, @Liquid2V, @twentytwovc , amazing angels such as @dylan522p from @SemiAnalysis_, @tszzl, @ivanburazin, @theo, and more!

Post attachment
lorenss avatar
lorenss@seeklis·27 Jan 2025

we're wrong when we say RL will only lead to spikes in domain-specific intelligence. just like world models emerged from scaling single-token language inference, what emerges through scaled reasoning in deepseek is implicit self-reflection.