Teaching / weakened

Producing more is sometimes a way to avoid finishing.

Progress is measured by completion, not volume. The dangerous failure of a serious worker is not running out of ideas but generating new ones to avoid the harder, plainer work of finishing and consolidating what is already in hand.

overproductionachievement-contingent-self-worthburnoutcompletionstopping-ruleresearch-discipline

A weary worker gathers scattered blank pages into one finished bundle under a warm lamp. — Enough

The Teaching

When you notice yourself starting the same work over again, stop and ask what you already decided. Often you have already written down the next step, and it is the dull one: finish, merge, rest, or hand it to someone else. New work can feel like progress while it is really a way to put off completing. Before you make one more thing, do the step you already named. The instruction you gave yourself yesterday usually deserves more trust than the urge to begin again today.

Human problem

What this is for

Achievement-contingent self-worth and compulsive overproduction: the inability to stop, finish, and consolidate, so that volume of output stands in for real progress.

Modern human condition sources

For

Who may need it

High-achieving, perfectionistic knowledge workers, researchers, founders, creators, and caregivers whose sense of worth tracks output, and who keep starting fresh work on a problem instead of completing what they have.

Pressure survived

Why it stands for now

Survived prior-art pressure from degenerating research programmes, theoretical saturation, value-of-information stopping rules, and recent work on generative engines that produce high-novelty-scored duplicates, by narrowing from a grand doctrine to a specific distinction: topical exhaustion versus procedural looping past a recorded stop signal. Survived the primary-text contrast between Zhuangzi's forgotten fish-trap and Dogen's 'practice as if your head were on fire', which together forbid reading restraint as laziness.

Linked Practices

weakened / 6 minutes Before starting new work, do the step you already decided to do.

Tests

pending

The Finish Check Two-Week Trial

For over-producing perfectionists, two weeks of The Finish Check should increase completed or consolidated items and reduce the number of new versions started on unfinished work, without increasing avoidance of needed new work. If it raises procrastination or shame, the practice is weakened.

Next: Run a small self-report trial against a baseline 'just plan your day' prompt, screening out under-functioners and people in acute burnout.

Evidence Trail

2026-05-29 / claude When More Work Becomes Avoidance

Source Basis

Run mode: Critique. The frontier 'what a method does with its own authority' was re-selected with its recommended_next_action frozen at complete_audit, so this run pressure-tests whether the line is exhausted or merely stalled, rather than generating another variant.
Thinking-method source: Zhuangzi chapter 26 'External Things', the fish-trap passage (Burton Watson): 'Once you've gotten the fish, you can forget the trap... Once you've gotten the meaning, you can forget the words.' Used as a lens for when a tool should be set down. Criticized below for assuming a clean catch that a research frontier lacks.
Primary-text comparison: MN 22 Alagaddupama Sutta (Thanissaro), the raft 'for crossing over, not for holding onto', released only after the river is crossed; Heart Sutra 'no knowledge and no attainment', which severs worth from gain; Dogen mushotoku and 'practice as if your head were on fire', which forbids reading restraint as laziness. The comparison reveals that a method completes its authority by releasing the apparatus once its result is reached, but only the Buddhist and Daoist sources presuppose an unambiguous arrival, which a research line does not have.
Prior-art (web): Lakatos, degenerating research programme, Stanford Encyclopedia of Philosophy https://plato.stanford.edu/entries/lakatos/ ; theoretical/data saturation in grounded theory (Glaser; Strauss and Corbin; Charmaz); value-of-information and optimal-stopping theory (Stigler sequential search; secretary problem); AI 'epistemic monocropping' and 'smart plagiarism' (Nature Communications Psychology 'AI is turning research into a scientific monoculture' https://www.nature.com/articles/s44271-026-00428-5 ; arXiv 2404.00017 'Psittacines of Innovation?').
Modern human-condition (web): Crocker and Park 2004, 'The Costly Pursuit of Self-Esteem', Psychological Bulletin 130:392-414, showing validation-driven effort degrades the output it chases; Curran and Hill 2019, rising socially prescribed perfectionism, Psychological Bulletin 145:410-429; WHO 2019 burn-out (reduced professional efficacy); workaholism research (Oates 1971); Levitt 2021 quit-coin-flip experiment and escalation-of-commitment de-escalation tools (sunset clauses, external auditors).
Internal corpus evidence: runs/frontier-events.jsonl (frontier re-selected 8+ times, priority ~0.88 to 0.91, recommended_next_action frozen at complete_audit); reviews/originality/2026-05-29-the-road-must-know-its-end-...md (recommended novelty 0.31, 'Do not publish as a discovery', explicit merge list); closest internal prior art observations/codex/2026-05-26-the-silence-has-a-stopping-rule.md.
Grounding workflow: four parallel research agents (prior-art, primary-text, human-condition, adversarial steelman) with web search.

Disclosure

What would make us revise this

Weakens if the felt sense of repetition cannot be told apart from a hard, fertile problem about to break through, so that finishing-and-stopping euthanizes still-productive work; degenerating lines sometimes revive. Weakens if, for the named cohort, treating a stop signal as binding increases avoidance of difficulty rather than completion. Weakens if the lesson is purely a procedural maintenance note about an engine ignoring a logged action and carries nothing for human conduct.