Skip to content

False Closure Typology

False closure is when the output looks complete by one closure dimension while failing another. The methodology distinguishes specific false-closure shapes that practitioners encounter. Each has a distinct response.

Build-Pass / Consumer-Fail

Build green, tests pass, source basis cited, but the consumer (reader, user, operator) rejects the output.

SymptomResponse
Build greenHold topic in pending, do not mark closed
Tests passAdmit a new wave addressing the consumer dimension
User rejectsDo not retroactively edit the closed wave

Real example: the previous public docs topic that produced this current docs remediation. Authority + semantic + drift-resistance all passed; consumer failed.

Source-Anchored / Unreadable

Every public claim has spec basis, but the prose is so source-anchor-flavored that no reader can build a mental model.

SymptomResponse
Source basis presentA follow-on rewrite wave
Reader cannot build mental modelReorganize by reader purpose

The follow-on does not reopen the source-basis decision; it addresses the readability gap.

Closed-But-Not-Accepted

A wave closes machine-side but human acceptance never recorded. The methodology's pending state with explicit close_trigger is for this exact case.

SymptomResponse
Wave closeout recordedTopic state stays pending until acceptance
User has not reviewedDo not advance to true-close
Acceptance gate unrecordedMaintain pending-note with reopen criteria

The topic does not advance to true-close until acceptance is recorded.

Overflow Vs PASS

A wave that did not finish in its packet boundary returns OVERFLOW, not FAIL and not PASS.

Overflow continuation is admitted only when:

  • direction is still correct
  • scope has not crossed into a new owner domain
  • the current state is acceptable but the packet boundary was too thin

Overflow continuation is rejected when:

  • shadow truth was introduced
  • a fallback or alias was needed
  • the packet crossed into a new owner domain

This distinction prevents the most insidious false closure: an overflowing wave being silently extended past its owner domain, accumulating implicit scope.

Premature True-Close

A topic is folder-closed without recording an explicit true-close audit.

SymptomResponse
Folder moved to closed/Verify topic.yaml.current_true_close_status
current_true_close_status: not_startedRoll back; record true-close audit first
Passed true-close revoked laterRecord revocation lineage

Passed true-close may itself be revoked by a later independent audit; revoked true-close requires follow-up lineage.

Pseudo-Progress

New wave names exist without new closure. The wave-DAG policy's anti-patterns explicitly catch this.

SymptomResponse
Wave admitted but no primary closure goalReject admission
Wave admitted, no bounded result reachedPause or re-preflight
Cascading wave namesStop; renaming is not closure

A wave needs a primary closure goal; if it cannot name one, it is not a wave, it is a planning detour.

Local-Requirement Trap

Small requests replace the mainline goal — a topic that started about A becomes a backlog of micro-fixes for B/C/D.

SymptomResponse
Topic accumulating unrelated micro-fixesRefuse; topic is one major iteration line
Each micro-fix becoming its own waveMove micro-fixes outside topic discipline

The development-rhythm rules say a topic is the home for one major iteration line, not a micro backlog.

Giant Planning Topic

A topic that never commits to a bounded wave.

SymptomResponse
Multiple consecutive planning wavesStop; planning-only waves cap at one consecutive
No bounded closure reached after planningPause or re-preflight rather than open new planning waves

The wave-limits policy is explicit: planning may harden one execution target, but it must not chain indefinitely.

Reader Scenario: Recognizing A Failure In Progress

You are managing a topic. The first wave's audit returns "PASS." But the user, reviewing the rendered output, says "it's correct but not impressive."

This is the build-pass / consumer-fail shape. The audit was correct (PASS by machine criteria); the consumer dimension failed.

Response:

  1. Do not retroactively edit wave-1's record. The audit was correct.
  2. Hold topic in pending.
  3. Admit wave-2 addressing the consumer gap.
  4. Wave-2's packet declares the new acceptance invariants based on the user feedback.
  5. Wave-2 closes when consumer dimension is satisfied (subject to next user review).

The methodology turns "I don't know what to do next" into a typed sequence.

Reader Scenario: Reading An Overflow Verdict

A wave returns OVERFLOW, not PASS or FAIL. The worker hit the packet boundary mid-implementation.

You evaluate:

  • Direction still correct? Yes — the work being done is the work intended.
  • Scope crossed into new owner domain? No — work stayed in the declared owner domain.
  • Shadow truth introduced? No — no parallel paths created.
  • Fallback or alias needed? No — work refused to fall back.
  • Current state acceptable, packet boundary too thin? Yes.

Continuation admissible. Admit a continuation packet that extends the boundary; complete the work.

If any answer had been "yes" to shadow / fallback / new domain, continuation would be rejected; the wave returns to revision.

Reader Scenario: A Topic Stuck In Planning

A topic has been ongoing for weeks. Three waves admitted; all three were planning-only; no bounded closure reached.

This is the giant planning topic anti-pattern.

Response:

  1. No new planning wave. Wave-limit policy forbids the fourth.
  2. Pause the topic with explicit pending-note, OR
  3. Re-preflight. Reframe and try once more with sharper stop line.

Pseudo-progress is what would happen if the manager just admitted wave-4 with another planning name. The methodology's limit is what stops it.

Source Basis

Nimi AI open world platform documentation.