Loop-completion

Description

A specialization of feedback-loop with its own diagnostic shape: gaps become visible precisely because the journey is coherent enough to reveal them. In a pre-coherent system (one where the basic flow is not yet established), you don’t know what’s missing because there’s no complete path to reveal the absence. In a post-coherent system (where the end-to-end path works), the missing pieces become identifiable because the loop’s expected shape is clear and the gaps stand out against it.

The loop-completion form names the diagnostic move of inspecting a working journey for its gaps rather than debugging a broken journey for its failures. The move is: “now that this works end-to-end, what parts of the feedback loop are missing or incomplete?” This is different from “does it work?” (which is a correctness check) — it’s “is the loop closed?” (which is a structural check on the feedback path).

The pre-coherent vs. post-coherent regime distinction is the form’s strongest instance: before a system reaches coherence, the question “what’s missing from the loop?” can’t be productively answered. After coherence, it becomes the most important question.

Encounters

• Post-coherence audit in agent frameworks — once the basic Claude Code flow worked end-to-end, the missing loop-completions became visible: no confirmation that insights were saved, no feedback on whether projections were acted on, no signal when a session produces no save-worthy output.
• Email confirmation flows — “did the thing happen?” is a loop-completion question. A system that triggers an action but doesn’t confirm receipt has an incomplete loop; the gap is only visible once you have a working action path.
• Analogy engine evaluation loop — the engine projects candidate inferences, but without a mechanism for James to log “this was useful” or “this missed,” the feedback loop is incomplete. The system works (pre-coherence cleared) but the learning signal is missing (loop not closed).
• KCC PR review cycle — PR submitted, review requested — but the “did the review happen?” confirmation step was missing. The gap was invisible until the PR flow was coherent enough to reveal it.
• Testing loops — a test harness that runs tests but doesn’t surface failures prominently has a broken loop-completion: the feedback exists but doesn’t close back to the developer.
• Documentation loops — “did the doc get read?” is an open loop. Some teams add “did you read CLAUDE.md?” checks at session start as a loop-completion device.

When it applies / triggers on

User-initiated: User has a working end-to-end system and is asking “what’s missing?” or “how do I know it worked?” or “where does the feedback go?” Also when a user notes that something happened but they didn’t hear back / didn’t get confirmation.

Agent-initiated: Engine detects a working flow (post-coherent) and notices an absent or weak feedback path. Candidate inference: “this loop isn’t closed — what’s the completion signal that would close it, and who or what receives it?” Also: when a user is in a pre-coherent phase, the agent can flag that loop-completion diagnostics apply after coherence is achieved.

Vocabulary cues: “close the loop,” “complete the loop,” “did this land?”, “confirmation,” “feedback,” “where does the signal go?”, “how do I know it worked?”, “missing step,” “end-to-end,” “coherent journey,” “pre-coherent,” “post-coherent.”

Situation-shape signals: A system where something is triggered but the outcome isn’t confirmed. A working pipeline that lacks monitoring. A review process without a “reviewed” acknowledgment. The form is most useful when the basic path is established but the feedback arc is incomplete.

Composes with

• feedback-loop (specialization relationship) — loop-completion is a specialization of feedback-loop focused on the structural completeness of the loop rather than its dynamics. A feedback-loop can be closed but slow or noisy; loop-completion asks whether it’s closed at all.
• cadence (composition relationship) — loop-completion requires cadence: how often does the completion signal arrive? A loop with no cadence on the completion step is structurally open even if the mechanism exists.
• graduation-promotion (composition relationship) — the pre-coherent → post-coherent transition is a graduation move. Loop-completion diagnostics are appropriate after the graduation; before, they’re premature.
• shape (composition relationship) — the loop’s shape is the reference against which gaps become visible. You need to have a clear model of what a complete loop looks like before you can identify what’s missing.
• load-bearing (composition relationship) — the completion signal is often load-bearing for the system’s ability to improve over time. An open loop is a system that can’t learn; identifying the load-bearing completion step is the key move.

When it doesn’t apply

• Pre-coherent systems — asking “is the loop complete?” before the basic path works is premature. The form requires a working end-to-end journey as its precondition.
• One-shot processes — a process that runs once and has no expected recurrence has no loop to complete. Loop-completion applies to recurring flows.
• When the completion signal is trivially obvious — systems with synchronous, immediate feedback (function call returns a value) have trivially closed loops. The form earns its keep when the feedback path is asynchronous, distributed, or easy to miss.
• When “closing the loop” adds more noise than signal — over-instrumented systems that confirm every micro-action obscure the signal in confirmation noise. Loop-completion is about the right completion point, not exhaustive confirmation.

Sources

• Control theory: closed-loop vs. open-loop systems. A closed loop has a feedback path from output back to input; an open loop does not.
• UX design: “close the loop” as a user communication principle — confirm that an action was received and acted on.
• Agent framework design: the gap between “trigger an action” and “know the action completed” is a canonical open-loop pattern in async systems.
• The pre-coherent / post-coherent regime distinction: observed in James’s corpus as the regime shift that enables loop-completion diagnostics.
• Noted as a candidate with caveat: may be a specialization of feedback-loop. Promoted on the strength of its distinct diagnostic shape (visibility conditional on coherence).

Canonical exemplars from corpus (T2 2026-05-17)

Mined via scripts/mine_new_form_exemplars.py from data/backfilled-insights.jsonl. 17 backfill-only matches at score ≥ 2 — mid-support. Pre-coherent/post-coherent vocabulary is rare in the corpus by name; most matches surface via “close the loop” / “did this land” / structural diagnostic phrasings. Co-occurrences are sparse (1-2 with each of grain, cadence, doctrine).

• Dropping the confirmation step shrinks code AND simplifies UX (cwd: campconnect, 2026-05-10): “Dropping the confirmation step actually shrinks the code AND simplifies the UX. That’s the load-bearing-test signal: if removing structure improves both code and UX, the structure wasn’t doing real work. The confirmation step was risk-aversion theater pretending to be a feature.” — loop-completion in inverse: the confirmation was a loop-closure attempt that turned out to be unnecessary (the recovery surface elsewhere closed the loop adequately).
• Pre-coherent regime diagnosis on the analogy engine itself (cwd: analogy, 2026-05-17): “The ‘data investigation > pure-design’ pattern is itself a form. Compositionally: pre-coherent-vs-post-coherent-regime (you’re in the pre-coherent regime; the feedback the data gives you is what makes the coherence) + cargo-cult anti-pattern (designing without measuring is shape-without-mechanism).” — the regime classifier aspect of loop-completion applied to the project’s own methodology.
• Market filter as the natural recovery (post-coherent) path (cwd: campconnect, 2026-05-10): “The market filter on /programs is the natural recovery path. Parent whose GPS got the wrong market lands on /programs filtered to that market, sees no relevant programs, naturally reaches for the market filter.” — loop-completion via a downstream recovery surface that closes the loop without an inline confirmation step.
• Spatial proximity encodes semantic association (cwd: campconnect, 2026-04-28): “Spatial proximity encodes semantic association whether you intend it or not. ‘Manage kids’ sitting next to ‘Share’ implies they’re both view-level actions… Moving Manage kids to the left next to the kid pills makes the spatial story match the conceptual story.” — the gap between intended action category and perceived one is only visible because the surface is otherwise coherent; the missing-completion is conceptual alignment between layout and semantics.

Trigger pattern (T2): Loop-completion surfaces after a system reaches end-to-end coherence; the diagnostic move is “now that this works, what parts of the feedback loop are missing?“. The form is rarely lexically marked — agents reach for it after observing a working path that has an implicit acknowledgment gap. Caveat: 17 matches is modest; some corpus prevalence likely sits in phrasings the heuristic misses (e.g., “post-ship learning,” “what’s still open”).