AI Development Methodology

P1 — Doc completeness

Pillar goal: for every common situation an adopter encounters, the methodology has explicit, findable guidance.

Last updated: 2026-08-19

Related:

• Brief: Capability layer 1
• Strategy phase: Phase 1 — Foundation (primary)
• Depends on: nothing (foundational)
• Feeds into: P2 — Doc clarity (clarity is improvement on top of completeness)
• Delivering epics: (none yet)

1. Overview

This is the first pillar because nothing else works without it. Gaps in coverage send adopters away — a methodology that’s good on 8 of 10 needed surfaces feels incomplete; adopters either improvise (and get burned) or leave for a more complete option.

Doc completeness asks: for each common adopter situation, does the methodology say something specific? Not “we have a section about that area” but “an adopter who lands here gets actionable guidance.” The bar is findable + actionable, not just present.

2. What this pillar covers

Surface	What “complete” means here
Planning (strategy / pillars / epics / items)	Every layer has a doc that explains its shape, lifecycle, and required content.
Discipline (working principles / DoD / memory)	Every discipline has a doc with rules + anti-patterns + concrete examples.
Operations (locks / git workflow / testing)	Every operational surface has rules + protocols + recovery patterns.
Human dimension (roles / decision ownership)	The human side of AI-driven workflow is named, not assumed.
Templates	Every AI tool with significant adopter share has a template (or adapted template).
Worked example	At least one concrete worked example exists (the self-development folder).
Operational adopter questions	“How do I cut a release?” “What if I’m doing brownfield adoption?” “How do I handle a hot-fix?” — each has explicit guidance.
Edge cases	“What happens when the lock TTL expires?” “What goes in HUMAN_NEEDED.md?” “When should I deviate from the ROI heuristic?” — each named.

3. Exit criteria

The pillar is delivered (not done — it’s evergreen) when:

• All 14 methodology docs (00–13) have been reviewed in the most recent semi-annual self-evaluation pass with no “missing” gaps surfaced by cross-AI review. This criterion read “12 docs” until 2026-08-19, which scoped it to 00–11 and silently excluded 12_milestone_evaluation.md and 13_ai_safety_and_prompt_injection.md — both added after the criterion was written. The 2026-05-25 pass ran against the old scope.
• Each of the 6 supported AI tools (Claude Code, OpenAI Codex, Google Antigravity, Cursor, Aider, Continue.dev) has a usable starting template — either native (3: CLAUDE.md, AGENTS.md, and AGENTS.md again for Antigravity) or adaptable (3, from AGENTS.md). Adaptation is the accepted permanent answer; the epic that proposed native files for the latter three (E04) was parked 2026-08-14.
• At least one worked example folder exists in the repo (the self-development/ folder satisfies this as of v1.7.0; examples/example-project/ added in v1.15.0).
• Every Discussions / Issues thread tagged as a “how do I” question has a recorded response (from maintainer or community) within 30 days that either links to an existing doc as the answer or is logged as a methodology gap candidate for the next release.

Re-tested: every semi-annual self-evaluation pass (per methodology/07_definition_of_done.md). The fourth criterion above only becomes testable once Discussions activity exists; until then, treat as N/A.

Health indicators (not binary, but worth watching):

• Adopter questions repeat across multiple threads with no existing answer → likely a gap to file.
• “I had to read this three times” feedback → completeness might be OK but pairs with a clarity (P2) issue.

4. Dependencies

Depends on: nothing. This is the foundational pillar.

Feeds into:

• P2 Doc clarity — completeness is necessary but insufficient; the next pillar makes the complete docs actually readable.
• P3 Doc currency — completeness must be maintained as the methodology evolves; currency keeps it true.
• P5 Adopter discoverability — discoverability is downstream of having complete enough docs to make discovery worthwhile.

5. Anti-patterns

• Writing a doc to fill a section heading rather than because adopters need it. Empty stubs are worse than missing sections.
• Repeating content across docs because each section “should be self-contained.” Cross-references > duplication. Drift will follow duplication.
• Adding sections without testing whether adopters find them. A complete-but-buried doc is functionally missing.
• Over-specifying — writing 800 lines when 200 would do. Completeness is “covers the surface,” not “exhausts the surface.”
• Treating completeness as binary — the surface area grows as adopters use the methodology in new contexts. Re-evaluate continuously.

6. Current state (last reviewed 2026-08-19; entries below date from v1.8.0 unless noted)

Strong:

• 14 methodology docs (00–13) cover planning, discipline, operations, human roles, milestone evaluation, and AI safety.
• 6 templates cover the major AI tools and the working phases (CLAUDE.md, AGENTS.md, AGENT_KICKOFF.md, AUTONOMOUS_LOOP.md, PROJECT_STRUCTURE.md, ROLE_BRIEFS.md — the last added in v1.30.0).
• v1.5.0 added 4 substantive new sections (stdlib growth loop, verification taxonomy, brownfield onboarding, decision matrix).
• v1.6.0 added 5 more (self-evaluation cadence, lock-file management, squash-vs-merge, AI-autonomy-in-git, release tagging, hot-fix workflow).
• self-development/ worked example exists.

Known gaps (as of v1.8.0):

• No anonymized real-project examples beyond the self-development folder. (Tracked as a potential Phase 2 epic.)
• No CHEATSHEET.md / one-page reference. (Deferred; the brief and TL;DR sections serve this partially.)

Honest: completeness is the most-developed pillar currently. The next ~3 months of Phase 1 work is incremental polish, not major additions.

7. Delivering epics

(None yet. Epics get added here as Step 2 of the self-development bootstrap charters them.)

This site is open source. Improve this page.