Live now
track / model

Train a model that generalizes.

Classic ML and prompt tasks, trained in an in-browser Python notebook — real scikit-learn, nothing to install. Submit to a hidden held-out leaderboard, mentored by TENSOR.

Open the studio →live now · free to start
01

Train classic ML in a browser notebook

An in-browser Python notebook runs real scikit-learn right here — no setup, no GPU, nothing to download. Fit a model, tune it, and watch your accuracy climb. (Deep-learning tracks are on the roadmap, not here yet.)

02

Scored on a hidden held-out split

Train on the data you can see, then submit predictions. They're scored against a held-out split you never get to peek at — so the leaderboard rewards a model that generalizes, not one that memorized.

03

Prompt tasks graded by a rubric

Engineer prompts against tasks judged on a clear rubric, not vibes. TENSOR, our research mentor, asks the next sharp question about your approach — never hands you the answer, never spoils the test.

// the studio is open

Model is live. Open the studio.

Train classic ML in an in-browser scikit-learn notebook, submit predictions scored on a hidden held-out split, and engineer prompts graded by a rubric. Free to start — no card required.

hidden-holdout leaderboard · play in your browser