Verifiable draw · one draw per catalogue · 100+ tasks

Dataset Draw

Paste your dataset.toml. The moment it lands here is the commitment: the draw uses the first public drand randomness round published after that moment, a number that does not exist yet when you upload. Neither you nor I can steer or re-roll it, and anyone can recompute the result.

Step 1 · Commit your dataset.toml

0 tasks parsed · min 100

One draw per catalogue. Every draw is recorded under your company name, and the history is visible to labs.

How to verify a draw

verifiable_draw.py
  1. The receipt names a drand round. Fetch it yourself, from any drand mirror, and check the randomness matches: api.drand.sh/public/<round>
  2. Check the round’s publish time is after the receipt’s received_at: the randomness did not exist when the catalogue was committed.
  3. Recompute: seed = SHA256(manifest_hash ‖ randomness), score each task with SHA256(seed ‖ task_sha), sort ascending: the lowest n are the draw. Same inputs, same result, every time.