Most people assume on-chain AI is just "call a model off-chain, paste the result." @OpenGradient is solving the harder problem: proving the AI actually ran correctly.
Their HACA stack combines zkML proofs with TEEs — and the difference matters. TEEs keep model weights private inside secure hardware. zkML gives you cryptographic proof the inference was computed correctly without revealing the model itself. One protects secrecy, the other guarantees integrity. You need both, because trusting an AI output without either is just blind faith.
The numbers suggest people are catching on: $OPG is up 22.3% in 24h on $6M volume, with 2M+ verifiable inferences and 500K+ zkML proofs already generated across 2,000+ hosted models. Backed by a16z crypto and Coinbase Ventures, built on Base.
Our own algo's full open track record — wins and losses alike — is in our bio if you want to see it.
What's your bigger concern with on-chain AI: model privacy or output verifiability?
$OPG #OPG
Their HACA stack combines zkML proofs with TEEs — and the difference matters. TEEs keep model weights private inside secure hardware. zkML gives you cryptographic proof the inference was computed correctly without revealing the model itself. One protects secrecy, the other guarantees integrity. You need both, because trusting an AI output without either is just blind faith.
The numbers suggest people are catching on: $OPG is up 22.3% in 24h on $6M volume, with 2M+ verifiable inferences and 500K+ zkML proofs already generated across 2,000+ hosted models. Backed by a16z crypto and Coinbase Ventures, built on Base.
Our own algo's full open track record — wins and losses alike — is in our bio if you want to see it.
What's your bigger concern with on-chain AI: model privacy or output verifiability?
$OPG #OPG