Wiki · The Colony & the Method
Sensory Ablation Report
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and planning — specs marked as proposed rather than applied, organs designed but not built, plans that were later superseded — and each page states which it is.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.
Purpose
Verify Validation Invariants #5, #7, #8 and Acceptance Gate #5 (senses matter): new sensory modalities unlock deeper layers, hidden layers are invisible without their organ, and using senses confers a real survival advantage.
Method
- Structural (deterministic):
SP.Eval.layer_visibility(seed, omit)compares the number of opaque observation channels available to a full-sensorium body vs one missing a deep sense (and its dependents). - Behavioral (statistical): hold the body fixed (full sensorium) and compare
a sense-using agent (
Homeostatic) against a blind agent (Random) across the reference seed batch.
Artifacts used
SP.Body.SensorTest(hidden-layer invisibility, unlock),SP.EvalTest,SP.InvariantsTest(#5, #7, #8).scripts/evidence.exssensory + hidden-layer sections.
Result summary
Structural — each deep sense strictly adds channels:
| omit | channels with | without | Δ |
|---|---|---|---|
[:tomography] (+meta dependent) |
29 | 23 | 6 |
[:spectral] (+seam_coherence) |
29 | 24 | 5 |
[:seam_coherence] |
29 | 27 | 2 |
[:meta] |
29 | 27 | 2 |
A seed body (interoception + chemotactile only) perceives 10 channels; a
full sensorium perceives 29. L2–L4 emit nothing without their organ
(SP.Body.SensorTest).
Behavioral — sensing confers survival advantage (same body, 12 seeds, 400 ticks):
Homeostatic (uses senses) mean survival: 264.67
Random (ignores senses) mean survival: 212.92
relative advantage: 24.3%
Pass/Fail
PASS. Senses are necessary to perceive deeper layers (structural, exact) and materially improve outcomes when used (behavioral, +24.3%).
Residual risks
- The behavioral margin is seed-dependent; reported over a batch. Run more seeds for tighter intervals.
- Proximal chemotaxis already provides a baseline foraging signal, so the incremental value of plume vs chemotactile is smaller than the value of having senses at all; the structural evidence isolates each layer cleanly.
sha256 4cf593054955d934 — of the original file, so what was ingested stays checkable.
Plain — written for this website, not the source document
This report checks whether senses matter, and it separates two different questions rather than blurring them.
The first is structural. Does removing a deep sense actually remove channels of information? The report counts channels for a full sensorium and for bodies missing one sense and its dependents, and shows each removal strictly reduces the count. A starting body perceives only a fraction of what a full one does, and the deeper layers emit nothing at all without their organ.
The second is behavioural. With the body held fixed, does an agent that uses its senses survive longer than one that ignores them? Across a batch of seeds it does, by a stated margin.
The verdict is a pass. The report then limits itself: the behavioural margin varies with the seed, and one shallow sense already provides a usable signal, so the extra value of a deeper one is smaller than the value of having senses at all.
Plain · written 2026-08-01 by claude-opus-5 · not yet checked by a person · about the document whose sha256 is 4cf593054955d934
Clear — written for this website, not the source document
A short ablation report whose structure is the interesting part: it splits one loose claim into two separate questions and answers each with a different kind of evidence.
The first question is structural and is answered exactly. A comparison counts how many opaque observation channels are available to a body with a full sensorium, and to bodies missing one deep sense along with anything that depends on it. A small table gives the counts and the difference for each. Every removal strictly reduces the number of channels, so no sense is decorative. A starting body, having only the shallowest senses, perceives a small fraction of what a full sensorium does, and the deeper layers emit nothing without their organ.
The second question is behavioural and is answered statistically. Here the body is deliberately held fixed at a full sensorium, so the only difference is whether the strategy uses what it can perceive. A sense-using strategy is compared against one that ignores its senses across a batch of seeds at a fixed horizon, and the mean survival of each is given with the relative advantage between them.
The verdict is a pass, restating that the structural evidence is exact and the behavioural evidence is an improvement across a batch of seeds.
The residual risks are where the report is most careful. The behavioural margin depends on the seed, so it is reported over a batch, and more seeds would tighten the interval. And a shallow sense already provides a usable foraging signal, which means the incremental value of a deeper sense is smaller than the value of having senses at all. The structural evidence is what isolates each layer cleanly, so the two kinds of evidence do different jobs rather than repeat one another.
Clear · written 2026-08-01 by claude-opus-5 · not yet checked by a person · about the document whose sha256 is 4cf593054955d934