Match Process Weight to Stage and Stakes

Dial the weight of process (planning, documentation, review gates, ceremony) to a project's stage and stakes: light discovery loops early, ceremony added only where a specific pain appears, heavy process only when coordination or correctness risk exceeds the ritual's cost. Grounded in part by Rao on Legibility; the stage × stakes dial itself is still a hunch.

Promoted 2026-11-08 — partially confirmed by Rao — A Big Little Idea Called Legibility (originally captured as the human's hunch 2026-06-20). Rao/Scott ground the mechanism behind the ❌ high ceremony, low value cell: process imposed so a central eye can read the work destroys the local, tacit value it was meant to manage ("the big brother of waterfall planning"). Alexander adds the positive half: grow structure only by structure-preserving steps where its absence is felt (Living Structure). That is "add ceremony only where a specific pain appears." Not grounded: the claim that weight should be dialed to stage × stakes. See Open questions.

Supporting examples, not groundings (auto-ingest, 2026-11-08; compressed):

• Prescriptive rules help weak agents and tax strong ones (Anthropic — The Components of Agent Memory).
• "Deliberate checkpoints… accept slower progress to reduce risk" (Anthropic — How Scientists Use Claude).
• "Do normal things unless they don't work" (Anthropic — Investigating Vault Particles with Claude).

All three dial process to context. None argues stage × stakes.

Open questions

The stage × stakes dial is still ungrounded. Rao doubts that agile mitigates legibility failure, which is a caution for any "lightweight early" prescription. Grounding candidates: Boehm & Turner (Balancing Agility and Discipline), Cockburn (Crystal), Cynefin. The value × ceremony 2×2's form still needs a Rao quadrant essay.

The idea

Process is not free, and it is not one-size-fits-all. The right amount of process — planning, documentation, review gates, sign-offs, ceremony — should be dialed to the project's stage and stakes, not applied uniformly:

• Early / exploratory → lightweight discovery loops. When you're still finding out what you're building, fast cheap iteration beats ritual. Heavy process here mostly taxes learning.
• Add ceremony reactively, where a specific pain appears. Don't front-load process speculatively. A given gate or artifact earns its place when there's a concrete failure it would have prevented (a regression that slipped, a handoff that dropped, a decision nobody can reconstruct).
• Heavy SDLC is justified only when coordination and correctness risk exceed the cost of the ritual. Many people in the loop, expensive-to-reverse decisions, high blast radius if it's wrong — that's when the overhead of formal process pays for itself. Below that threshold it's net negative.

The underlying move is a cost/benefit at the margin: each unit of process buys some coordination/correctness assurance and costs some speed and learning. Match the dial to where the project actually is. The failure modes are symmetric — over-process an early/low-stakes effort and you smother discovery; under-process a high-coordination/high-correctness-risk effort and you eat rework, defects, and chaos.

A 2×2 framing (value × ceremony)

Sharpened 2026-06-20 — the human intuits this fits a 2×2 matrix "ala Venkatesh Rao" (quadrant-thinking). Still a hunch; the matrix is the maintainer's rendering of the human's "high value, low ceremony" cue.

Plot any activity, gate, or artifact on two axes — value it produces (low ↔ high) against ceremony it demands (low ↔ high):

	Low ceremony	High ceremony
High value	✅ Sweet spot — lean, high-leverage; just ship it	⚖️ Justified heaviness — overhead earns its keep only when coordination/correctness risk is real
Low value	🆗 Cheap tinkering — harmless low-stakes experiments	❌ Bureaucracy / theater — ceremony with no payoff; the trap to cut

This operationalizes the heuristic into a diagnostic:

• "High value, low ceremony" is the cue the human flagged — the quadrant to default to and protect, especially early/exploratory.
• The whole heuristic = avoid the ❌ bottom-right (process theater) and earn your way into the ⚖️ top-right only when stakes justify it. "Add ceremony where a specific pain appears" is precisely the move from low-ceremony into the ⚖️ cell.
• Open question the 2×2 surfaces: value and ceremony aren't independent — some ceremony creates value (a review that catches a defect). The axes may need to be realized vs. expected value, or net value after ceremony cost.

What would ground this

A source — ideally with an explicit model or empirical evidence — that argues process discipline should be tailored to risk/criticality/size rather than applied uniformly. Strong candidates to seek (not yet read, so not logged as human-flagged leads — see note below):

• Boehm & Turner, Balancing Agility and Discipline (2003) — the most direct match: home-grounds analysis (size, criticality, dynamism, personnel, culture) for how much plan-driven discipline vs. agility a project needs. Would confirm strongly.
• Cynefin (Dave Snowden) — match your response to the complexity domain (clear / complicated / complex / chaotic); "probe-sense-respond" in the complex domain ≈ lightweight discovery loops. Would confirm the stage half.
• Alistair Cockburn, Crystal family — scale methodology weight to team size and defect criticality. Would confirm.
• A challenge source would argue process discipline should be consistent from the start — e.g. that quality/architecture/security can't be bolted on late, that ad-hoc early process creates path dependence and tech debt. Worth flagging if found, so the tension stands rather than resolving silently.

Leads

• Venkatesh Rao #to-read — flagged 2026-06-20 for the 2×2 form. Rao's quadrant-thinking (Ribbonfarm) is a model for naming the matrix's cells sharply; his legibility writing (riffing on James C. Scott) is also substantive grounding for why imposed ceremony destroys value — the ❌ quadrant above.

The human flagged Venkat Rao for the matrix's form, not as a source on the process claim itself. The grounding candidates in "What would ground this" (Boehm & Turner, Cynefin, Cockburn) remain the maintainer's suggestions, not yet human-flagged — promote any to a real #to-read lead to surface it in LINT.

Connections

• Agent Autonomy Spectrum — structural twin: both are "match the dial to the context" heuristics. There the dial is agent autonomy (turn-based ↔ semi-synchronous ↔ mega-autonomy) and the sweet spot is in the middle; here the dial is process weight and the right setting depends on stage/stakes. Same shape of judgment, different axis.
• The Bottleneck Shift — see also: process/ceremony is itself a cost, and when execution gets cheap (AI agents), the "is this gate worth it?" calculus shifts. If deciding what to do is the scarce input, ceremony that doesn't serve that decision is even harder to justify.
• Vibe Coding — the lightweight end made concrete: describing intent and iterating fast is exactly the early/exploratory discovery loop this heuristic favors before stakes rise.
• Leadership, Lab, and Crowd — see also: Mollick's "build stuff that doesn't work yet" / learning-as-discovery culture is the same low-ceremony, high-iteration stance at the org level.
• Control-Flow Integrity — see also (cross-domain): a concrete instance of the ⚖️ justified heaviness quadrant — security rigor adopted when correctness/ exploitation stakes make the overhead worth it.
• Scope Hygiene for Tickets and Tooling — adjacent process craft (not a grounding): hands-on rules for running tickets/tooling cleanly. Related domain, but it argues how to keep work clean, not this seed's thesis that process weight should scale with stage × stakes — so it doesn't promote the seed.
• Choose the Platform by the Hardest Recurring Part — see also: both weight a decision by where the real recurring cost is, rather than the most salient up-front cost.

2026-06-26: the Working Heuristics — Data, Matching, Product, Org, Tooling ingest brought in adjacent process/tooling craft (above) but did not ground this seed — none of those heuristics assert the weight ∝ stage × stakes claim. Still awaiting a real grounding source (Boehm & Turner / Cynefin / Cockburn).

<!-- On PROMOTE: move this file to concepts/, set status: active, remove the hypothesis tag, add the citation, and record the lineage, e.g. "Promoted 2026-06-20: confirmed by Boehm & Turner — Balancing Agility and Discipline (captured as a hunch 2026-06-20)." -->

Sources

• Rao — A Big Little Idea Called Legibility — partial grounding (legibility mechanism).
• Alexander — The Origins of Pattern Theory (OOPSLA 1996) — structure-preserving growth.
• The human's original hunch (captured 2026-06-20).