Wiki · The Colony & the Method
SP.Lab — Validation Report (QA / UAT / Science cross-check)
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-four pages about the colony. Each agent is an Elixir process holding a generative model and doing inference, attached to a body that logs into a Minecraft world as an ordinary player. Around that sit the broadcast suite that films them and the runbooks that keep the whole thing running. There are typed specifications for each organ of the model, plus the world and genome specs. There are also the adversarial review personas used to attack a proposed change before it ships.
It is for the reader curious how a running system is put together and how it is held to account. The accountability half is the more distinctive. There is a lab protocol governing evidence and attribution, and a claim fence that restricts the vocabulary a claim is allowed to use. There is a public gate log. And there is a standing invitation to reproduce any verdict from the commit and the seed named in its receipt.
Start with the public read, then the lab protocol, then the falsification invitation. If you want the mathematics rather than the operations, go straight to the typed organ specs.
What it is not: a description of a mind, and not all one kind of document. A large part of this corpus is design and planning — specs marked as proposed rather than applied, organs designed but not built, plans that were later superseded — and each page states which it is. A specification is not a running system, and these pages are careful about the difference; the reader should be too. Eight documents were withheld from publication because they describe private infrastructure.
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.
Branch: lab/ozone-life-uni-hard-science · Date: 2026-06-11 · Toolchain: Elixir 1.19 / OTP 28
Purpose
Independently harden the science lab (SP.Lab.*) before any downstream artifact is built on its
numbers: run the CI-grade gates, build a falsifiable cross-check that re-derives every documented
number from the code, fix-forward every real delta, map the 20 hard tests to evidence, and capture
the result. The repo's own discipline applies — the math is allowed to say "contradicted," and
nothing is labelled "proven."
Method
- Gates —
mix compile --warnings-as-errors --force;mix format --check-formatted(lab files);mix test(full +test/sp/lab/). - Cross-check — new
mix sp.lab.validate(SP.Lab.Validate) re-derives 24 code-backed quantities (the constants the modules use + every worked example / dossier figure derived from them) and asserts each against its documented value within a declared tolerance; exits non-zero on any delta. - Remediation — fix-forward; see
lab/remediation_log.md(append-only, including a tried-and- reverted change, recorded honestly). - UAT —
docs/reports/lab_uat_checklist.mdmaps the 20 hard tests to evidence.
Artifacts
| Artifact | Path |
|---|---|
| Gate output | lab/evidence/captures/a1_gates.txt |
| Cross-check report | lab/evidence/captures/lab_validate_report.txt |
| Computed reports | lab/evidence/captures/computed_reports.txt |
| Remediation log | lab/remediation_log.md |
| UAT checklist | docs/reports/lab_uat_checklist.md |
Result
| Gate | Result |
|---|---|
mix compile --warnings-as-errors --force |
clean |
mix format --check-formatted (lab files) |
clean (pre-existing non-lab drift noted, not touched) |
mix test (full suite) |
413 tests, 4 doctests, 0 failures |
mix sp.lab.validate |
24 checks, 0 failed, ALL GREEN |
| UAT (20 hard tests) | 16 PASS, 2 HOST (out of lab scope), 1 NOT-BUILT (flagged), 1 visual PASS — 0 failed |
Key numbers, code-confirmed (from computed_reports.txt): Newtonian g = GM/R² fails 0/7 at
2% (max residual 0.36% at the Moon); the pressure-weight model g = k·P fails 5/5 out-of-sample
(Venus ~100×, Titan ~10.5×, airless bodies ~14 orders); ozone τ(300 DU) = 88.77 → transmittance
2.8×10⁻³⁹; Nernst slope 59.16 mV/pH; PMF(150 mV, 0.5 pH) = 120.42 mV; radiative floor 893 W.
Tolerances
Each cross-check carries its own tolerance (exact equality for defined constants; ±0.01–±5 for
derived worked examples; band checks for order-of-magnitude claims). The Newtonian 2% test
tolerance reflects rotation/oblateness omitted by GM/R²; the gravity-vs-pressure verdict is
decided by a ~14-order-of-magnitude gap, far outside any tolerance.
Residual risks (honest)
- Scope of the cross-check: it recomputes only code-backed numbers. ~40 literature-sourced
ledger entries (Chapman rate constants, biosignature figures, geologic dates) are
provenance-checked in
lab/evidence/, not recomputed here — stated explicitly, not implied. - Two UAT rows are HOST (#13/#14, active-inference belief updating — lives in the host system, not the lab) and one is NOT-BUILT (#18, no time-evolution simulator). These are recorded gaps.
- Declared roundings: G, Faraday, gas constant, Stefan–Boltzmann are 4-s.f. roundings of the exact SI values (cited alongside); immaterial to every verdict, and the cross-check confirms the whole corpus is consistent with them.
Verdict
The lab is internally consistent and falsifiable: code, tests, ledgers, proofs, and dossier
agree on every code-backed number, and the gate mix sp.lab.validate will break CI if that ever
stops being true. No scientific verdict changed during remediation — pressure-replaces-gravity
remains contradicted-by-test, ozone-as-shield remains Class B / supported-within-model, and
nothing is labelled "proven."
sha256 7af6a432fbe28919 — of the original file, so what was ingested stays checkable.
Plain — written for this website, not the source document
This is a validation report for a science lab component, written before anything downstream was built on its numbers. Its aim was to harden the lab rather than to praise it.
The method has four parts. Run the strict compile, formatting and test gates. Build a cross-check that re-derives every documented number from the code and fails if any disagrees with what is written. Fix each real discrepancy, recording even a change that was tried and reverted. And map the hard tests to their evidence.
The results are given as a table of gate outcomes, followed by the key numbers the code produced. Two scientific readings are named as unchanged by the work: one account remains contradicted by test, another remains supported only within its model.
The residual risks are the part to read. The cross-check only recomputes numbers the code produces. A large group of figures taken from the literature is checked for provenance and not recomputed, and the report says so explicitly.
Plain · written 2026-08-01 by claude-opus-5 · not yet checked by a person · about the document whose sha256 is 7af6a432fbe28919
Clear — written for this website, not the source document
This is a validation report on a science lab component, and its stated purpose is to harden it independently before anything downstream is built on its numbers. It says up front that the project's own discipline applies: the mathematics is allowed to say a thing is contradicted, and nothing is labelled with the word the project forbids.
The method is in four parts. First the gates: a strict compile that treats warnings as errors, a formatting check limited to the lab's own files, and the full test suite. Second a new cross-check that re-derives a set of code-backed quantities, comparing each against its documented value inside a declared tolerance and exiting non-zero on any difference. Third remediation, fixing each real discrepancy forward and keeping an append-only log which, notably, records a change that was tried and then reverted. Fourth the mapping of the hard tests to their evidence in a separate checklist.
The artifacts are listed with their paths, and the results are a table of gate outcomes. A clean compile, and a clean format check with pre-existing drift elsewhere noted rather than touched. A passing suite, a passing cross-check, and the checklist's counts across its categories.
A paragraph then gives the key numbers the code produced. It covers how a standard gravitational formula fares against several bodies, and how badly a rival pressure-based account fails out of sample. Then an optical result, an electrochemical slope, a derived potential and a radiative floor.
The tolerances section explains that each cross-check carries its own, ranging from exact equality for defined constants to band checks for order-of-magnitude claims. It explains why one tolerance is what it is, by naming the physical effects the simple formula omits.
The residual risks are labelled honest, and they earn it. The cross-check recomputes only numbers the code produces; a substantial set of figures taken from the literature is checked for provenance rather than recomputed, and the report says this is stated explicitly rather than implied. Two checklist rows belong to the host system and one was never built. Several constants are declared roundings of exact values, cited alongside, and said to be immaterial to every verdict.
The verdict claims internal consistency and falsifiability rather than truth: code, tests, records, proofs and dossier agree on every code-backed number, and the cross-check will break the build if that stops being so. It records that no scientific verdict changed during the work, and repeats which account remains contradicted by test and which remains supported only within its model.
Clear · written 2026-08-01 by claude-opus-5 · not yet checked by a person · about the document whose sha256 is 7af6a432fbe28919