TrainLoop

We partnered with @mercor_ai to test a simple idea: What if knowledge-work agents were just… coding agents? Result: +25% performance, 2x faster, cheaper, and new SOTA on APEX-Agents. @josancamon19

+48% from a single RL step, and 100k rollouts from a single policy. @LoganGrasby wrote up his findings that off policy training with OAPL is more robust than we thought!

Task-specific training can be far more efficient than we realize. This work by @hasith_v explores the LoRA-trained solution space for GSM8k, finding a massive "plane" of solutions, representable in a single rank.

Specialized models for the most important tasks!

🧵 Labs and VC's are throwing cash at RL environments, especially for computer and browser use. Yet, with just 4 customers and over 30+ vendors, is cloning every website in the world really the path to scale? of course not. Introducing TRACE: Trajectory Recording and Capture of Environments.

@FlyaKiet @jackson_stokes @mlpierce22 @avimakesrobots @superset_sh TrainLoop ❤️🤝 Superset

@TrainLoop_ai is one of those companies in our group office hour that has something special. They're hitting an exponential curve and are now hiring more cracked engineer! Had a great time onboarding @jackson_stokes and @mlpierce22 to Superset yesterday with @avimakesrobots

@jackson_stokes @gdb No way

@jackson_stokes chatgpt is merging with windsurf is merging with vim

🍿 OoOh we’re getting CODEX in ChatGPT

@elonmusk @EnhancedRadar

There is a shortage of top notch air traffic controllers. If you have retired, but are open to returning to work, please consider doing so.

@ThePrimeagen hold my beer

puppeteer is the single worst library ever written