@kortxfi

Kortx is building a verification layer for AI inference on Solana, recording cryptographic evidence (input hash, model fingerprint, sampling seed) on-chain at different attestation tiers (attested, sampled, proven).

The project addresses AI output provenance — enabling independent verification that a specific model produced a given output.

Currently running on Solana devnet with a working implementation visible via GitHub and a live indexer, though still experimental (labeled EXP-01).

Green flags: Genuinely early-stage: 238 followers, brand-new X account (Aug 2024), devnet deployment · Technical substance: open-source on GitHub, working Anchor program with three-tier verification model · Real problem: AI inference provenance gap in current API model landscape · Clear architecture: input hash + model fingerprint + sampling seed committed on-chain with explicit coverage limits per tier · Transparency: publicly states what each tier does NOT prove, not overselling guarantees

Red flags: Very early devnet stage, no mainnet deployment yet · Token mention ($KORTX described as collateral) but unclear if live or planned · Extremely low social traction (238 followers, 1 tweet with modest engagement) · Team/founder information not visible on website or Twitter