Wiki · Evidence & Verdicts
RED pre-registration — forage-pureworld-graduation
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 — forage-pureworld-graduation
- Gate name:
forage-pureworld-graduation - Phase: Phase 2 → graduation
- Pre-registered: 2026-07-13
- Runner:
runs/pureworld_qa_gate.exs - Related:
runs/nursery_forage_gate.sh,runs/pureworld_qa.exs
Motivation
The forage runway (metab_scale=0.2) is closed — deep-body UNIs survive by their own hunting at reduced metabolic pressure (development / womb-wean). Pure-world graduation is the discriminator claim: the trained brain forages+survives in a PURE world (metab_scale=1.0, no runway) on every seed AND the untrained twin does NOT — zero VOID. Verbatim from docs/receipts/emergent_forage_cure1.md:70-71. Task #25.
PASS condition (verbatim)
The trained brain forages+survives in a PURE world (metab_scale=1.0, no runway) on every seed AND the untrained twin does NOT (the discriminator) — zero VOID.
FALSIFIES condition (verbatim)
Any seed on which the trained twin dies AND the untrained twin survives.
Protocol
- Seed sweep: N ≥ 8 world seeds (choose ones with realistic biome mix, no forced-easy-food spawns). Freeze the seed list in the receipt before the run.
- For each seed:
a. Start a fresh MC world at
metab_scale=1.0, no gives, no props, kin memory sha-locked from the runway-closure lineage. b. Twin A = trained lineage (homeostat_colony/0or the closed-runway lineage). Twin B = untrained control (freshdefault/0genome, no memory). c. Run for T ≥ 4 hours in-world (or the pre-registered N ticks). d. Record: energy trajectory, food events (killed/collected/eaten), death events, forage-loop closure receipts. - Verdict:
- PASS iff every seed shows Twin A alive at T AND some seed shows Twin B dead by T.
- PARTIAL iff some seeds PASS the discriminator but not all.
- FAIL iff any seed shows Twin B alive AND Twin A dead.
Ship-gate discipline
- Independent confirmation: RCON
listat 1-hour intervals + brain probes against the live registry. - Claim fence: no "proven survival," no "life," no "graduation" language in the receipt outside the verdict word.
- One cure at a time: this runs standalone. If concurrent RED (motor-shuffle, consummation-Run3) is in flight, defer.
Non-goals
This gate does NOT establish self-awareness / experience / life. It establishes ONLY that the trained lineage's foraging behaviour survives full metabolic pressure while a proper untrained control does not.
sha256 af682234fdbcd007 — of the original file, so what was ingested stays checkable.
Plain — written for this website, not the source document
This is a plan for a test, written before the test. It asks whether a trained system can keep itself going in a harder world where nothing has been made easy for it, while an untrained twin, given the same world, does not. That contrast is the whole point: without the untrained twin, surviving on its own would say nothing about the training. The verdict is withheld. The page ends with a short list of things this test would not settle, including anything about awareness or experience, and the authors put that limit in writing themselves.
Plain · written 2026-08-01 by claude-opus-5 · not yet checked by a person · about the document whose sha256 is af682234fdbcd007
Clear — written for this website, not the source document
A pre-registration — the conditions written down before the run — with the verdict withheld. It describes a run that had not yet happened when the page was written.
The background is a gentler setting that had reduced the pressure the bodies were under while they learned to feed themselves. This test takes that help away. The claim under examination, quoted word for word from an earlier receipt, the file recording that run, is that the trained lineage feeds itself and stays going in the harder world on every seed, while an untrained control does not. The second half of that sentence is what makes it a discriminator rather than a success story.
The protocol asks for a spread of world seeds, chosen for a realistic mix rather than for easy food, and frozen into the receipt before the run. For each seed: a fresh world at full pressure, nothing given, no props, and two twins, one carrying frozen memory from the earlier lineage and one starting from nothing. Both run for at least a stated stretch, and a fixed list of things is recorded. Three verdicts are defined in advance, and the result that would show it wrong gets its own line: any seed where the trained twin dies and the untrained one keeps going.
Then comes the discipline, which is unusually explicit. Confirmation is to come from an independent channel rather than from the system's own report. The receipt is fenced against particular words. Only one such test may be in flight at a time, and if another is running this one waits. And a closing section states what the test would not settle: it speaks to behaviour under full pressure against a proper control, and to nothing beyond that.
Clear · written 2026-08-01 by claude-opus-5 · not yet checked by a person · about the document whose sha256 is af682234fdbcd007