Every external-culture run is filed as a GitHub issue, logged on receipt, redacted for blinding, and routed to independent judges. The ledger below is the append-only record — zeros included.
Do not have the project team write your prompts or referee mid-run. Independence is the whole point.
Prefer confidentiality for the raw run? Email artifacts to lati@clista.ai — they still must be judgeable by blind external judges to count.
On a real, hard-to-reverse decision you own — using whatever agents or humans you already use. Start with npm run replay if you have not seen the format.
LEDGER.md, failures.md, cost.md — including human-minutes of format overhead — and outcome.md if the decision has been executed.
Open the External run report issue template. It pre-applies the external-run and gate-evidence labels and asks for the applicability check, run details, and artifacts.
The maintainer logs receipt, redacts for blinding, and assigns external judges per the rubric. Completed runs that go through judging count toward the gate — the score does not gate it.
| Run | Origin | Submitted | State | Outcome |
|---|---|---|---|---|
|
0 external runs received
This is reported honestly, including zeros. Failed and abandoned runs are wanted evidence — they belong here too. |
||||
Updated on a regular cadence toward the 2026-09-07 gate · counts mirror the
external-run + gate-evidence issues on GitHub.
