Attest a human score

Flat validator stake (ROOT) + $100 of ROOT burned per project, one score per project every 14 days. Neither stake nor burn size buys influence.

Scoring is gated by stake + burn; it is never paid — human scorers receive no monetary reward by design (ADR-0005). Validators are rewarded only from the tokens they scored — half of that token's ROOT-buyback lane, paid in its pair asset (ETH / USDG / RWA), never in ROOT.

Preview assumes a full-quality, original thesis. At epoch close the AI grade (below 0.3 ⇒ 0) and the copy check (derivative ×0.5, copy ×0) scale your weight; your raw score and thesis stay public either way.

Anti-exploit note

Near a systematic 0 the envelope drives human weight to ~0, so sock-puppet wallets can't lift a marginal project past the gate; near the top it protects a proven performer. Before any systematic score exists, humans can move a project at most 20 + 5 per later score, capped at 50. Every move is scaled by the AI-graded thesis quality, and copied theses (earliest submission wins) carry no weight.

Re-fetching a metric proves it exists in the client's system, not that the activity is economically real. Every weight-vector component is tiered by how expensive its signal is to forge. The eligibility gate requires nonzero T1-or-T2 output — a T3-only project scores 0 no matter how large its self-system metrics are.