Loading...
Loading...
TV-PRF/001 · Live RWA Agent Console
mETH staking yield + USDT0 stable treasury rail on Mantle · every allocation proven on-chain before execution
mETH flip grid
Lower-band risk-on ideas, upper-band risk-off exits, and adversarial validation stay visible before the wallet moves.
No decision payload loaded yet.
Every alpha-seeking reallocation must survive adversarial multi-model review before execution. Validator validator proposals were rejected before execution. Liveness-only heartbeat swaps stay tagged as HEARTBEAT_SWAP.
AI agents trade your money. You can't verify why they made a decision. Black-box models + irreversible on-chain actions = unacceptable risk.
Multi-model consensus (GLM-5 proposes → Claude 4.6 challenges → Gemini 3.5 Flash arbitrates) with every reasoning step hashed to IPFS and anchored on Mantle.
If the AI was wrong — the proof exists forever on-chain. If it was right — trust accumulates in a verifiable reputation score. No trust assumptions, only math.
Active Strategy
Vault contract pattern in development
Read the ValidationRegistry and ReputationRegistry contracts directly from your wallet. No backend, no trust assumptions.
Example pipeline lines. Real per-cycle reasoning is on the Proof Explorer — IPFS-pinned per decision.
The agent's prompt-evolution module can mutate the analyst system prompt based on settlement outcomes and pin each version to IPFS. It is currently disabled at runtime to keep JSON output stable across model providers. The pinned base prompt is v— (last updated —). Re-enabling is tracked as a follow-up spec.
Offline
no recent data · unknown