Agreed, and it's why we run live, not just open.
The network already serves real inference on independent operator nodes today, not a testnet.
Two checks catch what code review can't:
- Proof-of-Execution: every result carries a SHA-256 commitment the coordinator verifies.
- Redundant re-dispatch: a sampled share of tasks reruns on other nodes to flag divergent output.
Public repos catch code bugs. Live traffic PoE catch the runtime ones.
Reputation weighting (uptime, completion) is next. The coordinator opens after.