Skip to content

Give the probe-writing knowledge one home - #9

Merged
wormeyman merged 2 commits into
mainfrom
knowledge-docs
Aug 18, 2026
Merged

Give the probe-writing knowledge one home#9
wormeyman merged 2 commits into
mainfrom
knowledge-docs

Conversation

@wormeyman

Copy link
Copy Markdown
Contributor

Build-order step 8, and the last item on the list. The probe-writing knowledge that four repos have been accumulating separately now has one home instead of copies that drift.

factorio-blueprint-editor's README already said its method was "borrowed from the sibling project" by hand. That borrowing had drifted, which is the argument for doing it once here rather than a fourth time.

The three documents

docs/order-of-attack.md - which source to ask first. factorio-data, then the oracle, then the binary. Most behaviour is Lua shipped in the clear, and reaching for a disassembler usually means steps 1 and 2 were not exhausted: after eighteen probes, factorio-blueprint-editor has never needed it.

docs/method.md - the epistemics, and the part that transfers to any repo. A control must be able to fail while the hypothesis holds. Last man standing is not a measurement. A probe entity is part of the question, not a neutral instrument. Transcribe a proposed rule into the probe before writing code.

Every rule there was paid for by a probe that produced a clean, confident, wrong answer, and each names the issue where it happened. The rail-on-rail control that could only ever agree. The rail support placed on the wrong parity, which read as "supports do nothing" across all 16 orientations and 16 distances. The zoom rule that matched to nine digits and was the wrong rule.

It also records something worth expecting rather than being surprised by: four times, the measurement refused the change it was capturing evidence for. A probe that only ever confirms the plan is not being pointed at anything.

docs/gotchas.md - the facts, roughly 40 of them, each of which cost a run. Grouped by when they bite: writing the mod, launching the game, reading the answer back, fields that do not mean what their name says, version and seed traps, and numbers.

Attribution is the point, not a courtesy

A reader has to be able to tell "this repo checked it" from "borrowed and unverified", because otherwise there is no way to tell a typo from a finding.

So anything marked measured names the version and the date. Anything else names the sibling repo and issue it came from. An accuracy review against both source READMEs confirmed every transcribed number matches - the counts, ratios, tile offsets, issue numbers and hashes - and found four bullets carrying neither tag, which the second commit fixes.

The worst of those four claimed 2.0.77 ships five bundled mods rather than six, with no trail anywhere in this repo or either source. It is true, and it now says how it is known.

Scope

Docs only. No code changed and no test moved: 167 tests, same as main.

These are prose for a person, not input to anything. Nothing parses them, and nothing should.

🤖 Generated with Claude Code

https://claude.ai/code/session_01Du2g8GxpozoXhtZFu8MF82

Build-order step 8. The probe-writing knowledge that four repos have been
accumulating separately now has one home instead of copies that drift.

docs/order-of-attack.md - factorio-data first, then the oracle, then the
binary. Most behaviour is Lua shipped in the clear, and reaching for a
disassembler usually means steps 1 and 2 were not exhausted.

docs/method.md - the epistemics, and the part that transfers to any repo.
A control must be able to fail while the hypothesis holds. Last man
standing is not a measurement. A probe entity is part of the question, not
a neutral instrument. Transcribe a proposed rule into the probe before
writing code. Every rule here was paid for by a probe that produced a
clean, confident, wrong answer, and each names the issue where it happened.

docs/gotchas.md - the facts, roughly 40 of them, each of which cost a run.
Marked measured where this repo measured it, with the version and date,
and attributed to the sibling repo otherwise.

factorio-blueprint-editor's README already said its method was borrowed
from FactorioMapWebUI by hand. That borrowing had drifted, which is the
argument for doing it once here rather than a fourth time.

Docs only. 167 tests unchanged.
An accuracy review against both sibling READMEs confirmed every number
transcribed correctly - the counts, ratios, tile offsets, issue numbers and
hashes all match. What it found instead was four bullets carrying neither a
"measured" tag nor an attribution, which is the exact ambiguity the file's
own opening paragraph promises will not happen.

The worst of the four claimed 2.0.77 ships five bundled mods rather than six,
with no trail anywhere in this repo or either source. It is true, and it now
says how it is known: measured on a standalone 2.0.77 install, and confirmed
independently by the fixture diff, where recycler appears in loadedMods only
on 2.1.14. A durable reference that asks to be trusted is worse than one that
shows its working.

Also: the f32 round-trip claim is attributed to the notes file it actually
comes from rather than to an oracle README that does not contain it; the
zoom probe regains its second issue number; the cross-check bullet regains
the versionDifferences field the source names; and a dropped lesson is
restored - re-capturing footprints on 2.0.x found four entities that move
between versions, which nothing was looking for.

Docs only. 167 tests unchanged.
@wormeyman
wormeyman merged commit c01309f into main Aug 18, 2026
1 check passed
@wormeyman
wormeyman deleted the knowledge-docs branch August 18, 2026 02:36
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant