Every model, benchmark, leaderboard, RL environment and paper on one screen, with the signals that show what is rising and what moves a score.
Every leaderboard, ranked in real time.
Models side by side as a benchmark matrix.
Envs and datasets ranked by the benchmarks they move.
One typed stream. Save the filters you live in as views.
⌘K across models, evals, tools, papers and labs.
Context, price, modality and full benchmark coverage.
Star papers, models and evals to track. Browser-local, no account.
trained with a verifiable reward signal across envs
Mark up a paper's PDF in four colors. Pin a note to any passage.
Notes
Reproduces the SWE-bench claim. Compare reward shaping against v2.
Only stored in your browser.
Jot a thought on any paper. Stored only in your browser.
Links-first, no lock-in. 298 env-to-benchmark lift links.