UNI Universal Natural Intelligence

Wiki · Evidence & Verdicts

RED pre-registration — homeostat-colony-live

Evidence & Verdicts · docs/receipts/red_preregistration_homeostat_colony.md @ 44baf03d5041 (gen2-runtime) — opens the published snapshot ac338733bbba

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 — homeostat-colony-live

  • Gate name: homeostat-colony-live
  • Phase: Phase 2
  • Pre-registered: 2026-07-13
  • Runner: runs/homeostat_colony_red.exs
  • Related: CLAUDE.md:162-164, lib/sp/brain/genome.ex:homeostat_colony/0

Motivation

Genome.homeostat_colony/0 ships unproven per CLAUDE.md:162-164 ("streamed genome ... offline-green but NOT RED-validated live; it ships unproven only per explicit owner go-ahead"). This is the paired-twin RED that either promotes it to PASS or WITHHELDs it.

PASS condition

Paired-twin live RED against a properly-controlled sibling:

  • Twin A: homeostat_colony/0 lineage.
  • Twin B: matched sibling (same organs enabled, matched dirichlet counts at t=0, but homeostat_colony/0 streaming-parameter disabled).
  • PASS: Twin A shows the same runway-closure verdict as Twin B (no regression) AND at least one measurable signature (energy trajectory smoothness, satiety-attenuation of positive C on the whitelist) is distinguishable in favor of Twin A.

FALSIFIES condition

  • Twin comparison shows homeostat_colony/0 matched or beaten by the properly-controlled sibling; OR
  • default_genome byte-identity breaks.

Protocol

  1. Freeze the two lineages at identical seeds + memory shas.
  2. Same MC world, same spawn.
  3. Run T ≥ 4 hours; collect energy/satiety/kill/eat trajectories.
  4. Verdict:
    • PASS: no regression + ≥ 1 favorable signature.
    • PARTIAL: no regression but no distinguishable favorable signature.
    • WITHHELD: regression (Twin B beats Twin A) → withdraw homeostat_colony/0.

Ship-gate

  • No FE code changes to lib/sp/brain/genome.ex before this RED runs.
  • MERGED VERDICT required to promote from "ships unproven" → PASS.

sha256 6208f41295904d88 — of the original file, so what was ingested stays checkable.

Plain — written for this website, not the source document

Written for this website — not the document. This is a plain-language retelling, written to help you meet the document. It is not the source, and it is not evidence. It has not yet been checked by a person. (or choose Precise in the reading-level control above)

This page sets out a test that had not been run when it was written. It concerns a piece of the system that was shipped while still unproven, by explicit permission, and this test is the thing that would either promote it or withdraw it. The design is a paired comparison: the same setup twice, differing in one switch, run side by side under matched conditions. If the version without the new part does as well or better, the page says in advance that the new part should be withdrawn.

Plain · written 2026-08-01 by claude-opus-5 · not yet checked by a person · about the document whose sha256 is 6208f41295904d88

Clear — written for this website, not the source document

Written for this website — not the document. This is a clearer retelling, written to help you meet the document. It is not the source, and it is not evidence. It has not yet been checked by a person. (or choose Precise in the reading-level control above)

This is a pre-registration — conditions written down before the run — and its verdict is withheld. The subject is a lineage that was allowed to ship while still unproven in live conditions, on the owner's explicit go-ahead, and this page is the test that would settle it either way.

The header names the gate, the phase, the date it was registered, the script that would run it, and the two places where the claim it is checking already lives. The pass condition is a paired twin: one lineage with the new streaming parameter switched on, and a properly matched sibling with the same organs enabled and the same starting counts, but that parameter switched off. To pass, the new lineage must not regress against its sibling, and must also show at least one measurable difference in its favour.

What would show it wrong is the sibling matching or beating it, or an unrelated default behaviour ceasing to be byte-identical. The protocol freezes the seeds and the world, runs both twins for at least a stated stretch of hours, and collects the same trajectories from each. Three outcomes are written out in advance, including one that withdraws the lineage altogether. A closing rule stops related code from changing before the test has run.

Clear · written 2026-08-01 by claude-opus-5 · not yet checked by a person · about the document whose sha256 is 6208f41295904d88