Trustless & fast-improving AI models — powered by SN74 on Gittensor.
We build open infrastructure for verified AI improvement: frontier teachers propose, Blackwell GPUs prove, and deterministic evaluation rewards only measurable wins. No maintainer trust, no hand-waved datasets, no checkpoint-only claims.
| Repository | Role |
|---|---|
| SparkDistill | Open miner economy for Triton-native AI: verified datasets, training recipes, eval harness, and SN74 rewards on Gittensor. |
| SparkProof | Blackwell CC-attested Triton dataset generation — pinned frontier teachers, GPU compile/execute validation, Merkle proofs, and CPU-verifiable bundles for SparkDistill. |
- SparkProof miners generate stratified Triton trajectories on NVIDIA RTX PRO 6000 Blackwell GPUs with Targon confidential compute (Bittensor SN4), seal
sparkproof-2bundles, and publish verified datasets. - SparkDistill aggregates registry-approved miner datasets, gates merges with cryptographic verification, and runs a deterministic harness so SN74 can reward marginal quality over the current frontier.
- Anyone can re-verify proofs from published Hugging Face bundles on ordinary CPU hardware — attestation and hashes travel with the data.
Moving fast is key to success.
