NOVA Blueprint Update: Nesso-1 & Antitarget Scoring
We’re upgrading NOVA Blueprint with Nesso-1 as its new oracle model, replacing Boltz-2, and reintroducing antitarget scoring to reward molecules with stronger predicted binding selectivity.
Why it matters:
• Higher throughput: Nesso-1 enables significantly faster binding affinity predictions, allowing miners to evaluate more molecules and explore chemical space more efficiently within the same competition time budget.
• Greater selectivity: Reintroducing two antitarget proteins encourages algorithms to identify molecules that bind strongly to the intended target while minimizing predicted binding to undesired proteins.
The first competition will establish a new champion in a single round, with a bounty capped at one day’s emissions. Standard competition rules resume afterward.
The oracle is open source, so miners can run their own benchmarks and testing. We’ll also publish performance metrics for our selected hardware.
Rollout: November 9 | Epoch 20767
Code available for review:
Oracle:
Blueprint: