docs(v0): code-vs-vision snapshot 2026-05-27 (post-wave + fallow-ts surprise) #53

Merged

pdurlej merged 1 commit from claude/patchwarden-crosscheck-snapshot-2026-05-27 into main

2026-05-27 13:42:15 +02:00

claude commented

2026-05-27 12:44:42 +02:00

Collaborator

What

One new file: docs/operations/code-vs-vision-snapshot-2026-05-27.md (+230 lines). Durable crosscheck artifact for operator + future cousins. Captures where the v1.0 strategy meets actual src/ + related repos at end of wave Ollama + consolidation sweep (#47-#52 + #22).

Per operator request ("Zrób crosschecka moja wizja patchwarden a gdzie jesteśmy w kodzie", this session, post-#52 merge) + operator decision to persist as 1-file PR rather than transcript-only.

🚨 Headline discovery

pdurlej/fallow-ts shipped alpha 0.1.0-alpha.2 on 2026-05-19 — FIVE months ahead of Q4 2026 strategy plan. Bet 2 wrongness #2 (F2: fallow-ts requires >40h operator post-Wave 5) NOT fired — inverse happened. Codex spadek covered it in days.

This recalibrates Bet 2 ratio from estimated ~50% → ~67% of 6 layers operational. M3 refresh (2026-10-15) should incorporate.

What's in the file

Section	Purpose
Per layer (L1-L6)	Each layer's vision claim vs actual repo state. fallow-ts surprise documented.
Per principle (P1-P5)	Architectural enforcement points for each, locked vs aspirational.
D20 actor mapping	Patchwarden / reviewer lanes / Iskra / platform / operator authority.
Top 5 luki	Multi-lane orchestration, webhook, audit storage, verdict `blocked`, Docker compose.
Top 5 dopasowań	D20 enforced, deterministic gate, stdlib-only, atomic PRs, cousin discipline.
Mid-positions	Operator-language CLI, install docs, multi-client policies.
Bet 2 ratio update	~50% → ~67% post fallow-ts. Wrongness #1 and #2 NOT fired.
Diff vs v1.0 strategy	Explicit list of stale claims (vision.md, strategy.md, roadmap.md) that M3 refresh should touch. NOT to be edited before M3 per D16.
How to use this snapshot	Next-cousin instructions + quarterly refresh protocol.

Why a file (not just transcript)

Transcripts decay; this artifact survives next handoff. Next cousin can read this BEFORE opening a code PR and immediately know what's load-bearing vs planned vs surprise-already-shipped. Quarterly refresh trigger captured at top.

Scope discipline

Read-only synthesis. No edits to vision.md / strategy.md / roadmap.md.
Per D16, those docs are locked through M3 (2026-10-15). The diff section identifies what should be updated at M3, not now.
No new code, no test changes. 138/138 still green.

Atomic per ADR-0017

1 file, in existing docs/operations/ directory (created in PR #51).
0 src changes, 0 test changes.
base=main, no stacking.

Token-accounting

~3-4% weekly Opus for the synthesis + fallow-ts verification + writeup. Lower than the wave Ollama PRs because the heavy lifting was reading 6 docs + 11 modules I'd already loaded in this session.

Verification

All referenced files exist on main: README.md, docs/decisions.md, docs/roadmap.md, docs/architecture.md, docs/vision.md, docs/product/strategy.md, policies/platform.v0.toml, spec/schemas/{finding-registry,issue-verdict,pr-verdict,review-artifact}.schema.json, all 11 modules in src/patchwarden/.
fallow-ts state validated against git.pdurlej.com/pdurlej/fallow-ts via Forgejo MCP (repo metadata, commit history, README, roadmap.md).
D20 + ADR-0017 + ADR-0018 references match docs/decisions.md and docs/architecture.md as of main@4b85179.

## What One new file: `docs/operations/code-vs-vision-snapshot-2026-05-27.md` (+230 lines). Durable crosscheck artifact for operator + future cousins. Captures where the v1.0 strategy meets actual `src/` + related repos at end of wave Ollama + consolidation sweep (#47-#52 + #22). **Per operator request** ("Zrób crosschecka moja wizja patchwarden a gdzie jesteśmy w kodzie", this session, post-#52 merge) + operator decision to persist as 1-file PR rather than transcript-only. ## 🚨 Headline discovery **`pdurlej/fallow-ts` shipped alpha 0.1.0-alpha.2 on 2026-05-19** — FIVE months ahead of Q4 2026 strategy plan. Bet 2 wrongness #2 (F2: fallow-ts requires >40h operator post-Wave 5) **NOT fired** — inverse happened. Codex spadek covered it in days. This recalibrates Bet 2 ratio from estimated **~50% → ~67% of 6 layers operational**. M3 refresh (2026-10-15) should incorporate. ## What's in the file | Section | Purpose | |---|---| | **Per layer** (L1-L6) | Each layer's vision claim vs actual repo state. fallow-ts surprise documented. | | **Per principle** (P1-P5) | Architectural enforcement points for each, locked vs aspirational. | | **D20 actor mapping** | Patchwarden / reviewer lanes / Iskra / platform / operator authority. | | **Top 5 luki** | Multi-lane orchestration, webhook, audit storage, verdict `blocked`, Docker compose. | | **Top 5 dopasowań** | D20 enforced, deterministic gate, stdlib-only, atomic PRs, cousin discipline. | | **Mid-positions** | Operator-language CLI, install docs, multi-client policies. | | **Bet 2 ratio update** | ~50% → ~67% post fallow-ts. Wrongness #1 and #2 NOT fired. | | **Diff vs v1.0 strategy** | Explicit list of stale claims (vision.md, strategy.md, roadmap.md) that M3 refresh should touch. **NOT to be edited before M3** per D16. | | **How to use this snapshot** | Next-cousin instructions + quarterly refresh protocol. | ## Why a file (not just transcript) Transcripts decay; this artifact survives next handoff. Next cousin can read this **BEFORE** opening a code PR and immediately know what's load-bearing vs planned vs surprise-already-shipped. Quarterly refresh trigger captured at top. ## Scope discipline - **Read-only synthesis.** No edits to `vision.md` / `strategy.md` / `roadmap.md`. - Per **D16**, those docs are locked through M3 (2026-10-15). The diff section identifies what should be updated *at M3*, not now. - **No new code, no test changes.** 138/138 still green. ## Atomic per ADR-0017 - 1 file, in existing `docs/operations/` directory (created in PR #51). - 0 src changes, 0 test changes. - `base=main`, no stacking. ## Token-accounting ~3-4% weekly Opus for the synthesis + fallow-ts verification + writeup. Lower than the wave Ollama PRs because the heavy lifting was reading 6 docs + 11 modules I'd already loaded in this session. ## Verification - All referenced files exist on `main`: `README.md`, `docs/decisions.md`, `docs/roadmap.md`, `docs/architecture.md`, `docs/vision.md`, `docs/product/strategy.md`, `policies/platform.v0.toml`, `spec/schemas/{finding-registry,issue-verdict,pr-verdict,review-artifact}.schema.json`, all 11 modules in `src/patchwarden/`. - fallow-ts state validated against `git.pdurlej.com/pdurlej/fallow-ts` via Forgejo MCP (repo metadata, commit history, README, roadmap.md). - D20 + ADR-0017 + ADR-0018 references match `docs/decisions.md` and `docs/architecture.md` as of `main@4b85179`.

Rows
Columns