evidence / Self-refutation & honesty
Paper-Execution Discipline -- I built a full execution layer with no path to a real order
A complete paper-trading execution stack -- entry-timing study, order-lifecycle state machine, settlement join, and a paper ledger -- with a deliberate boundary: no code path reaches a real order or a real dollar. The single truth-source for any figure below is docs/JOB_EVIDENCE_PACKET.md section G. Paper-only; no edge, ROI, or dollar result is claimed anywhere, per the repo's no-edge rail.
strongest single receipt
the claim
The execution layer is fully built and tested, and there is intentionally no code path from it to a live order -- not an unfinished feature, the design. And where the layer cannot honestly measure a number, its own audit publishes the null: realized CLV is unmeasurable here because no independent closing-price feed was captured, so the artifact records realized_clv_pct: null and says why. The hire signal is a person who builds execution plumbing to the last mile, refuses to let it trade until a human wires that mile, and whose audit reports what it does not know rather than dressing a gap as a result.
receipts
| claim | receipt | committed artifact |
|---|---|---|
| Entry-timing study | "Efficient on entry timing; enter at the close" | scripts/platformkit/execution/entry_timing/ (900-1,645 drift events/market, NBA/MLB ML/spread/total); policy last_pregame_tick in data/frontend/ops/timing_policy.json (gitignored) |
| Double-gated executor | "Plumbing proven end-to-end; trigger deliberately removed" | executor/lifecycle.py header + mock_exchange.py, dryrun.py, test_lifecycle.py; env-flag gate + live=True gate + resolve_exchange() NotImplementedError backstop |
| TOCTOU write race closed | "Race closed and proven, not just asserted" | 099476ac; test_clv_ledger_io.py (two real OS subprocesses racing the same ledger) |
| `close_source` write-path bypass | "Silent bypass caught; historical rows left honest, not rewritten" | c1417777 (96/96 recent props rows forward-fixed) |
| Score-drift audit | "Drift surfaced and quarantined; settled history preserved" | 0d01a0e7 (76/436 settled MLB rows quarantine-flagged) |
| Self-published nulls | "No close feed = no CLV; the audit says so instead of fabricating" | paper_execution_audit.json (83 rows, 0 fills, edge_claimed=false; realized_clv_pct null for all 37 settled) |
why this matters
On a desk, the dangerous engineer ships an execution path that can fire before its controls are proven. The discipline here is the opposite: a complete, tested lifecycle with the live trigger absent behind two gates and a hard-refuse backstop; settlement integrity fixed forward-only with a real concurrency test, not a hopeful comment; and an audit that prints null and edge_claimed=false where the honest answer is "not measurable yet." No-edge discipline is not a shortfall -- it is the feature. I build execution systems to institutional completeness and keep the safety on until a human decides.
reproduce
# Proves the executor has no live path (every test uses MockKalshiExchange)
python -m pytest scripts/platformkit/execution/executor/test_lifecycle.py -q
# Proves the settle race is actually closed (two real OS subprocesses)
python -m pytest scripts/platformkit/test_clv_ledger_io.py -qThis site reports calibration and sharpness only, never a dollar edge, ROI, or bankroll result. An honest null is a success.