# williams-adlc-7-built-with-the-lifecycle-2026-06-12

## Veille

Seventh and final installment of the ADLC series: Williams presents an open-source toolkit of eighteen tools built *with* the cycle itself (build-prosecute-fix loop, parallel agents, a frozen `@adlc/core` core followed by fan-out — "pinned means merged"). The doctrinal core is "frontier-free": hitting precision targets with mid-tier models (Opus/Sonnet/Haiku-class) rather than frontier ones, via five substitutions (search replaces insight, decomposition replaces horizon, banking replaces presence, measurement replaces metacognition, the generator-verifier gap keeps the engine running), with the human remaining the "frontier" tier at the two spec gates. Throughline of the series: "replace trust with structure, and structure with measurement."

## Titre Article

The ADLC Toolkit

## Date

2026-06-12

## URL

https://www.voodootikigod.com/adlc-7-built-with-the-lifecycle

## Keywords

ADLC, toolkit, eighteen tools, @adlc/core, pinned means merged, frontier-free, mid-tier models, generator-verifier gap, search replaces insight, decomposition, banking, measurement vs metacognition, judge panels, premortem, parallax, lesson-foundry, adoption sequencing, replace trust with structure, npx, exit codes, github.com/voodootikigod/adlc

## Authors

Chris Williams (@voodootikigod)

## Ton

Profile: closing demonstration of a technical series (practitioner-builder perspective in English, synthetic and doctrinal register), very high technical level, targeting engineers and teams ready to adopt the cycle. The tone is proof-by-artifact: the toolkit was built with the cycle it tools, making the article both documentation and a self-referential demonstration. Authority rests on the shipped code (public repo, npm package, zero-dependency npx tools with deterministic exit codes) and on an "honest loss account" that explicitly owns what the doctrine sacrifices. Recapitulative and axiomatic style: five numbered substitutions, an ordered adoption sequencing, a named anti-pattern, and a summary formula closing the seven installments — "replace trust with structure, and structure with measurement."

## Pense-betes

- **Self-reference**: the toolkit (eighteen tools) was built *via* the ADLC cycle — parallel agents following the build-prosecute-fix loop, with a frozen core `@adlc/core` (LLM calls, git operations, CLI conventions, findings ledgers).
- **Structural principle**: "pinned means merged" — the shared core is merged **before** fan-out, to avoid dependency hell in parallel development.
- **Five tool phases**: Specify (`spec-lint`, `premortem`, `parallax`, `coldstart`); Rail + Build (`rails-guard`, `hollow-test`, `preflight`, `merge-forecast`, `model-router`, `flail-detector`, `consensus-fix`); Prosecute (`review-calibration`); Integrate (`behavior-diff`, `gate-manifest`); Distill (`lesson-foundry`, `skill-rot`, `model-ratchet`, `rejection-mining`).
- **Frontier-free doctrine**: hitting precision targets with mid-tier models (Opus/Sonnet/Haiku class) rather than frontier ones. Five substitutions replace frontier capabilities: (1) the **generator-verifier gap** keeps the engine running — verifying is cheaper than generating, "you never need a model smarter than the gate it must pass"; (2) **search replaces insight** — N diverse attempts selected by mid-tier judgment beat a single frontier pass; (3) **decomposition replaces horizon** — smaller tickets keep models below the degradation threshold; (4) **banking replaces presence** — expensive models strike permanent structures (contracts, skills, lints) then exit, mid-tier models operate inside them; (5) **measurement replaces metacognition** — parallax divergence, consensus stats, and enumerated gaps replace confidence queries.
- **Sixth substitution**: the human is the frontier tier, occupying the two specification gates where intent verification is ground truth.
- **Honest loss account**: the doctrine sacrifices single-pass architectural elegance, subtle cross-cutting intuition, latency, and long-horizon refactors resistant to decomposition. Mitigations: judge panels, premortems, human gates; the residue (~5% of the work) runs under maximum supervision.
- **Adoption sequencing**: (1) prosecuting existing PRs (zero workflow change, builds trust); (2) rails and test generation (installs the trust anchor); (3) interrogation (spec questioning after agents miss requirements); (4) full parallelism and distillation (once habits are established). **Anti-pattern**: mandating the full cycle from day 1 creates ceremony before compounding materializes.
- **Throughline of the seven installments**: "replace trust with structure, and structure with measurement" — the mechanism amplifies with stronger models, making it a lifecycle rather than a workaround.
- **Delivery**: repo `github.com/voodootikigod/adlc`, package `@adlc/core` (npm), zero-dependency npx tools with deterministic exit codes (0 = pass, 2 = gate failed) for CI integration.

## RésuméDe400mots

The seventh installment closes the series with proof: an open-source toolkit of eighteen tools, built with the ADLC cycle itself. Parallel agents followed the build-prosecute-fix loop around a frozen core, `@adlc/core`, which centralizes LLM calls, git operations, CLI conventions, and findings ledgers. The structural principle — "pinned means merged" — holds that this shared core must be merged before any fan-out, to avoid the dependency hell of parallel development. The tools are organized across five phases: Specify (spec-lint, premortem, parallax, coldstart), Rail+Build (rails-guard, hollow-test, preflight, merge-forecast, model-router, flail-detector, consensus-fix), Prosecute (review-calibration), Integrate (behavior-diff, gate-manifest), and Distill (lesson-foundry, skill-rot, model-ratchet, rejection-mining).

The doctrinal core is "frontier-free": targeting precision goals with mid-tier models (Opus, Sonnet, Haiku class) rather than frontier ones. Five substitutions replace frontier capabilities. First, the generator-verifier gap keeps the engine running: verifying costs less than generating, and "you never need a model smarter than the gate it must pass." Next, search replaces insight: N diverse attempts, sorted by mid-tier judgment, outperform a single frontier pass — and measurement proves the stack's capability. Decomposition replaces horizon: smaller tickets keep models below their degradation threshold. Banking replaces presence: expensive models strike permanent structures (contracts, skills, lints) once and exit, and mid-tier models operate within them. Finally, measurement replaces metacognition: parallax divergence, consensus statistics, and enumerated gaps replace confidence queries. A sixth substitution places the human as the frontier tier, at the two specification gates where intent verification is ground truth.

Williams keeps an "honest loss account": the doctrine sacrifices single-pass architectural elegance, subtle cross-cutting intuition, latency, and long-horizon refactors resistant to decomposition. Mitigations are judge panels, premortems, and human gates; the residue, about 5% of the work, runs under maximum supervision.

On adoption, he rejects the big-bang approach: starting with prosecuting existing PRs (zero workflow change), then rails and test generation, then interrogation, and only then full parallelism and distillation. Mandating the entire cycle from day one is the anti-pattern, since it imposes ceremony before compounding pays off. The throughline of the seven installments distills into one formula: "replace trust with structure, and structure with measurement" — a mechanism that amplifies with stronger models, making it a lifecycle rather than a workaround. The code ships at github.com/voodootikigod/adlc, as zero-dependency npx tools with deterministic exit codes for CI.

## GrapheDeConnaissance

- Chris Williams —publie→ The ADLC Toolkit (DOCUMENT, 0.97)
- Chris Williams —a_créé→ ADLC Toolkit (TECHNOLOGIE, 0.95)
- ADLC Toolkit —est_basé_sur→ ADLC (METHODOLOGIE, 0.92)
- ADLC Toolkit —utilise→ @adlc/core (TECHNOLOGIE, 0.92)
- Chris Williams —affirme_que→ « pinned means merged » : merger le core avant le fan-out (CITATION, 0.9)
- doctrine frontier-free —recommande→ atteindre la précision cible avec des modèles mid-tier (AFFIRMATION, 0.92)
- Chris Williams —affirme_que→ on n'a jamais besoin d'un modèle plus intelligent que le gate qu'il doit passer (CITATION, 0.93)
- generator-verifier gap —permet→ fonctionnement frontier-free du cycle (vérifier moins cher que générer) (CONCEPT, 0.9)
- search —remplace→ insight d'un modèle frontier (CONCEPT, 0.88)
- décomposition —réduit→ dégradation des modèles sur long horizon (CONCEPT, 0.89)
- humains —est_instance_de→ tier frontier sur les deux portes de spécification (CONCEPT, 0.88)
- ADLC —affirme_que→ « replace trust with structure, and structure with measurement » (CITATION, 0.94)
- Chris Williams —recommande→ adopter le cycle par séquencement de la douleur, pas en big-bang (AFFIRMATION, 0.9)
- ADLC Toolkit —utilise→ outils npx à exit codes déterministes (0 pass / 2 gate échoué) (TECHNOLOGIE, 0.89)

---
Canonical: https://www.thekb.eu/en/fiches/williams-adlc-7-built-with-the-lifecycle-2026-06-12/