Revenue Search: Inside Bittensor Subnet Session with Max from Score: Subnet 44
38 snips
Dec 1, 2025 Max Sebty, the founder of Score, dives into the innovative world of computer vision beyond sports. He details a new incentive mechanism using VLMs to enhance performance metrics through twin-track competitions. The conversation touches on Score's evolution, featuring insights from cricket strategist Nathan Leamon on integrating AI into sports analytics. Max reveals Score's expansion into various fields like retail and car washes, and the exciting launch of 'Vision GPT' for automated video analysis. Notably, he shares Score's first recurring five-figure invoice, highlighting real commercial success.
AI Snips
Chapters
Transcript
Episode notes
Pseudo Ground-Truth From VLMs
- Score uses state-of-the-art VLMs to generate a pseudo-ground-truth from footage so validation no longer requires humans.
- That unlocks scalable competitions and lets miners be judged against a consistent, verifiable baseline.
Consensus Replaces Manual Annotation
- Replacing human annotation with consensus and pseudo-ground-truth reduces attack vectors and scales validation.
- Miners can be evaluated objectively without a fragile 'trust me bro' approach.
Run Twin Public And Private Tracks
- Run twin tracks: an open, verifiable public competition and a private, TEE-backed client track.
- Use the open track to push model quality and the private track to onboard paying clients securely.