Wiki · Evidence & Verdicts
RED pre-registration — consummation-honest-cure2 (Run 3)
How to read this page
Three ways to read this page. Precise is the document itself, exactly as it is written in the repository. Plain and Clear were written for this website to help you meet that document — they are about it. They are not it, and they are not evidence.
Eighty-seven dated pages: receipts, pre-registrations, handoffs, validation records and review verdicts. A receipt is written at the moment a piece of work was checked. It names what was claimed, the commit and the seed, what was actually run, and the outcome in one of a small set of controlled words. Then it names what the work did not achieve. That last part is what makes it a receipt rather than an announcement. A pre-registration is the same discipline run in advance: the conditions that would count as a pass and the conditions that would falsify the claim are written down before the run, so neither can be adjusted once the numbers arrive.
That is why so many small dated stubs are an audit trail rather than noise. No one of them is meant to be a good read. The value is in the sequence and in the dates, because you can watch a prediction be registered, then the run happen, then the verdict land — sometimes against the prediction. Pages here record a falsified result, a rejected fix, a retracted overclaim, and a green receipt that turned out not to be reproducible from the commit that carried it. A record that carried only successes would be worth a good deal less than this one.
A gentle way in is to read a pre-registration first, so the shape becomes familiar, then a result page, then one of the corrections. This section sits off the main navigation on purpose: it is the record you check the rest of the site against, not the place to begin.
What it is not: documentation, and not a summary. Nothing here has been tidied in hindsight. Every entry reads as of its date, a later entry may overturn an earlier one, and the presence of a page is not a claim that its result stood.
Your browser cannot switch reading levels, so the document itself is shown.
Precise — the source document
This is the document. Rendered from the repository at the commit above, with nothing rewritten for the web. A gate re-renders it on every deploy and fails the build if a single byte differs.
verdict: WITHHELD evidence_class: pending
RED pre-registration — consummation-honest-cure2 (Run 3)
- Gate name:
consummation-honest-cure2(Run 3 supersedes Run 2) - Phase: Phase 2
- Pre-registered: 2026-07-13
- Runner:
runs/consummation_run3.exs - Related:
docs/receipts/forage_honest_consummation_RED.md:87-100
Motivation
Run 2 of the honest-consummation Cure-2 RED was CONFOUNDED: aggregated arms + non-isolated per-bot drop attribution meant Cure-2's distinct benefit vs Cure-1 could not be established. Run 3 fixes the confounder.
PASS condition
- Isolated arms: Cure-1 (baseline motor) and Cure-2 (honest-consummation motor: attack only closes the reach gap; kill only when reach + strike land) run in ISOLATED worlds (separate seeds, no shared spawn).
- Per-bot drop attribution: every drop event is attributed to a specific body via RCON
data get, with time-of-death and location. - PASS: Cure-2's per-bot drop rate is distinguishably better than Cure-1's on isolated arms (ε ≥ 2×).
FALSIFIES condition
Cure-2's per-bot drop rate is NOT distinguishably better than Cure-1's on isolated arms.
Protocol
- Two lineages, freshly seeded at identical prior:
- Cure-1: baseline motor.
- Cure-2: honest-consummation motor (existing implementation, offline PASS receipts on file).
- Two isolated MC worlds (different seeds, matched biome mix). One lineage per world.
- Run T ≥ 4 hours; per-bot drop attribution via RCON.
- Verdict:
- PASS: Cure-2 rate ≥ 2× Cure-1 rate under matched attribution.
- PARTIAL: Cure-2 > Cure-1 but ratio < 2×.
- FAIL: Cure-2 ≤ Cure-1.
Ship-gate
The Run-2 receipt (docs/receipts/forage_honest_consummation_RED.md) stays as the honest record of the PARTIAL. Run 3 either promotes to PASS or explicitly withdraws Cure-2's distinct-benefit claim.
sha256 419725bd8cca5829 — of the original file, so what was ingested stays checkable.
Plain — written for this website, not the source document
This page exists because an earlier run of the same test did not work. That run mixed its two arms together and could not tie outcomes to individual bodies, so it could not settle whether the newer design had any distinct benefit. This is the replacement plan, written before the new run: two separate worlds, one design in each, and every relevant event traced back to a particular body. The verdict is withheld. The page also says what happens if the new run does not come out well, and that answer is not a quiet retreat.
Plain · written 2026-08-01 by claude-opus-5 · not yet checked by a person · about the document whose sha256 is 419725bd8cca5829
Clear — written for this website, not the source document
This is a pre-registration — the conditions written down before the run — and it is also a correction. The verdict is withheld.
The motivation says what went wrong before. The second run of this test pooled its arms and could not attribute outcomes to individual bodies, so the question it was meant to answer could not be settled. This third run is designed to remove that confounder rather than to argue around it.
The changes are specific. The two designs run in separate worlds, with different seeds and no shared start, so neither arm can help or hinder the other. Every relevant event is attributed to a named body through an independent query, with a time and a place attached to it. The pass condition is a stated ratio between the two per-body rates, with a middle verdict for an improvement that is real but smaller than that, and a failure if the newer design does no better than the older one.
The closing rule is the part worth noticing. The older, weaker receipt — the file recording that run — is not deleted and not quietly replaced. It stays on the record as the honest account of what was found, and this run either promotes the claim or withdraws it.
Clear · written 2026-08-01 by claude-opus-5 · not yet checked by a person · about the document whose sha256 is 419725bd8cca5829