Skip to content
Open
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
32 changes: 32 additions & 0 deletions CONVENTIONS.md
Original file line number Diff line number Diff line change
Expand Up @@ -204,3 +204,35 @@ controls and the client cannot influence. The test to apply:
The same question applies to any set built from headers, query parameters, body keys, filenames or
metadata: derive from the schema, the manifest, the catalog, the code — the authority — and use
the request only as the thing measured against it.

## A brief names the instruction it supersedes, and a lane that cannot spawn says so

Two halves of one failure, and the second is what made the first visible.

**A brief states which instruction it replaces.** A coordinator was told to spawn subagents and
run two review lanes. Both lane briefs it wrote still carried an earlier *"do not spawn further
subagents"*, copied forward from a message that had already been superseded, and nothing in
either brief said which instruction won. Both lanes then ran as two scoped passes by one reader.

This is the same defect class as a superseded claim surviving into a record. The rule that
corrections are appended rather than rewritten exists because a stale sentence sitting beside a
true one is indistinguishable from a current one; a stale *instruction* sitting in a brief is
worse, because it is acted on rather than merely read. **The brief is a record, and it is subject
to the same rule: say what it supersedes, or it is not a correction, it is a contradiction the
reader has to resolve alone.**

**A lane that cannot spawn says so in its verdict.** Both lanes here did, unprompted, and that is
the only reason the defect was caught before the record shipped: each wrote plainly that it was
one reader making two passes rather than two readers. Had they written "two lanes" and stopped,
the record would have claimed independence it did not have — and *independence is the entire
content of a second lane*. A second pass by the same reader shares its assumptions, its blind
spots, and whatever the first pass concluded.

So the verdict names the condition it was produced under. "Two lanes" is a claim about how many
parties looked, and a run that cannot make it true says which it was, in the verdict rather than
in a footnote somewhere else.

**Derivation.** The failure cost nothing here, and that is the argument for the rule rather than
against it: the lanes reported the limitation themselves, so the record is honest. The next one
may not, and the phrase that would have shipped — "two lanes, three rounds" — reads exactly the
same either way.