First instance of the periodic deep-eval per methodology/12_milestone_evaluation.md, introduced in v1.15.0. Project-tailored rubric applied at project-wide scope. Cadence: every 3rd loop (Phase 1).
The methodology project is docs-only, so the standard product-shaped rubric (UX, frontend, backend, security, etc.) does not apply. Project-tailored areas reflect this project’s actual surfaces:
| Area | What 10/10 means for the methodology |
|---|---|
| Doc completeness | Every load-bearing concept has a doc; no gaps where an adopter would have to derive a missing rule. |
| Doc clarity | A fresh contributor can apply each rule after one read. Cold-read findings on clarity are sparse. |
| Doc currency | No stale claims; examples match current canonical forms; cross-doc rules consistent. |
| Cross-doc consistency | Same rule stated in one place; restatements explicitly cross-linked; no drift between docs. |
| Adopter discoverability | A new adopter finds the project, knows what it does, knows what to read first. |
| Tool compatibility | Works as-shipped with at least 3 AI tools natively; documented adaptation path for others. |
| Self-improvement velocity | Cycle runs end-to-end without maintainer authorship per fix; aggregate quality rises across cycles. |
| Governance integrity | Hard rules enforced; escalate-on-doubt works; cross-AI gates catch what implementing sessions miss. |
| Examples + reference material | Adopter-facing examples + cheatsheet present + accurate to current methodology version. |
Thresholds (Closed beta target): min 7/10 per area, average 8/10 across all (one notch below the standard 8/9 — closed beta is “trustworthy enough for ~50 strangers,” not GA).
| Area | Score | Evidence (paragraph) |
|---|---|---|
| Doc completeness | 9 | 13 numbered docs covering planning (00–05), disciplines (06–08), operations (09–10), human roles (11), milestone evaluation (12). The v1.15.0 additions (doc 12 + foundational diagram) closed the most visible gap. No load-bearing concept currently missing. -1 for no doc explicitly covering the brief/discovery process before strategy starts — currently embedded in 00’s “Why foundational work matters” section but could merit its own doc as Step 0. |
| Doc clarity | 8 | First eval cycle surfaced 14 unclear findings (BL-0007/0008 cold-reads). v1.14.0 fixed ~30 patches, escalated 5 T1→T2; 23 T2 findings remain for maintainer authorship. The methodology is internally coherent but still has clarity drift the loop continues to surface. -2 because of the 23 deferred T2 clarity items still pending. |
| Doc currency | 9 | Zero stale findings in Run 3 cold-read (docs 06–11). The one Run-2 stale finding (grep examples in doc 04) was fixed in v1.13.0. Examples match canonical forms post-v1.14.0. -1 because the v1.13.0 demo patch’s “no new stale findings introduced” claim isn’t testable against future drift; need next eval cycle to confirm. |
| Cross-doc consistency | 7 | 15 inconsistent findings + 5 cross-batch in the first eval. v1.14.0 fixed ~12 (cross-refs added; em-dash normalized; subagent-orchestrator vocabulary aligned). Remaining: ~8 T2 cross-doc items (authority chain ordering; six-gates count framing; lock-override rule omitted from hard-rules tables). Below threshold but improving; trajectory positive. |
| Adopter discoverability | 6 | Below threshold. README is honest + structured; CHEATSHEET (v1.15.0) helps. But: zero external adopter signal yet. No Show HN / blog / awesome-list submission has happened. The methodology is unfindable by anyone who doesn’t already know about it. This is the dominant blocker to closed-beta readiness. |
| Tool compatibility | 8 | 5 templates support 6 tools (3 natively — CLAUDE.md, AGENTS.md × 2 — and 3 via adaptation from AGENTS.md: Cursor, Aider, Continue.dev). E04 charters native templates for the latter three but is not yet active. -2 because no real test of an adopter using a non-Claude tool exists; tool compatibility is documentation, not validation. |
| Self-improvement velocity | 9 | 4 cycles ran in one day producing 30 patches at ~30 sec maintainer-review-per-fix. The compounding promise of v1.13.0’s tier matrix is operationally demonstrated. -1 because volume so far is one full cycle; sustained-cadence claim requires the November 2026 pass to confirm. |
| Governance integrity | 9 | Escalate-on-doubt fired correctly (5/10 T1→T2 sample rate). Diff-verification caught 3 mechanical issues in the v1.14.0 batch. Maintainer-merge gate preserved across all loop-authored changes. Hard rules (no force-push; never override locks; tier-matrix limits) all held. -1 because the rules have been tested by one maintainer with strong methodology context; adopter-facing robustness is untested. |
| Examples + reference material | 8 | CHEATSHEET.md (v1.15.0) + examples/ folder with fictional tinker project (v1.15.0). Both shipped today. -2 because: only one worked example, narrow domain (CLI utility); no test that adopters actually use these to bootstrap. |
Average: (9 + 8 + 9 + 7 + 6 + 8 + 9 + 9 + 8) / 9 = 8.11 / 10 Minimum area: 6 (Adopter discoverability)
Threshold (Closed beta target): min 7, avg 8. Actual: min 6, avg 8.11.
NOT READY for closed beta — one area (Adopter discoverability) below the minimum-area threshold. The average passes, but per the rubric rule “no area averaged away,” the project does not yet qualify for the milestone declaration.
Superseded 2026-08-19 (appended; the finding below is left as written). The four staged drafts this section proposes filing items against were deleted by maintainer decision — a good project sells itself — so
self-development/distribution/no longer exists and none of BL-E03-D01/D02/D04 will be filed. The 6/10 score and the NOT READY verdict stand; the prepared path to changing them is what was removed. Two corrections to the finding itself: “zero external promotion has happened” was already inaccurate when written — see the inventory in P5 — and D03 (GitHub Discussions) was never dependent on the drafts. Current position: P5.
Root cause: zero external promotion has happened. The distribution plan from self-development/brief/06_distribution.md was authored but not activated.
Action: activate distribution plan as a P0 item in a new epic or in E03 (which is now active). See self-development/distribution/ for prepared materials (drafts staged for maintainer-led publication). Once ≥ 1 adopter signal lands, re-score this area.
New items proposed (to be filed if maintainer accepts):
self-development/distribution/SHOW_HN_DRAFT.md).awesome-lists (PR draft staged in self-development/distribution/AWESOME_LISTS_PR.md).self-development/distribution/BLOG_POST_DRAFT.md).Trajectory is positive but the 8 remaining T2 cross-doc items in loop-notes/2026-05-25.md should be partially shipped before the next eval to keep the score climbing. Recommend maintainer ships 2–3 highest-impact T2s as a v1.16.x patch release between now and Run 6.
manual-verified test value not exercised yet. This pass introduced the value in doc 04; no item has used it yet. Watch the next cycle to see if it gets adopted appropriately or abused.None identified this cycle. All 23 deferred T2 findings + this eval’s surfaced gaps have known solutions — they’re queued for maintainer authorship, not marked unsolvable.
Self-assessment caveat. Per doc 12’s rule “Use cross-AI for scoring, not the implementing session,” this eval was conducted by the same session that authored v1.16.0. Scoring is biased toward charity. A fresh cross-AI session should re-score the same rubric as a check before the closed-beta milestone is declared. Surfacing as a follow-up item.
Single eval is a snapshot. Trajectory matters more than point-in-time score. Re-eval at Run 6 (next cadence point) to see whether the discoverability score moves after the distribution-plan activation.
Reviewed by: _________________________
Date: _________________________
Verdict: [ ] Acknowledged — discoverability gap accepted; distribution items filed.
[ ] Conditional — distribution scope adjustments before items file.
[ ] Reject — re-score with cross-AI before next action.
Notes: