Skip to main content
Append-only log of the autonomous build. One entry per milestone (see GOAL.md). Newest at the bottom. Each entry: what shipped · key decisions · what was verified & how · open ❓.

​
M0 — Scaffold & conventions (2026-07-04, ~01:15)

Shipped:
  • Directory structure: part-1-theory/, part-2-platform/, part-3-sops/, _sources/.
  • CONVENTIONS.md — front-matter, audience-tag callouts, video-embed slots, real-prompt format, markers (❓, no-credits rule), cross-linking.
  • index.md — home + two learning paths (“New AM start here”, “Self-serve tour”).
  • JOURNAL.md, gap-register.md, video-shot-list.md scaffolds.
  • Resolved decisions folded into PLAN.md (markdown-in-repo, one tagged site, no credits, self-serve = full capability / lighter rigor, live prompt mining).
Decisions:
  • Portable Markdown now; site generator later. Page files numbered by their map id.
  • Audience divergence handled by inline callouts, not forked pages.
Verified: prod-context factory key works (listed sessions across orgs; confirmed factory query/messages). Directory tree created. Open ❓: none at M0. Gaps will accumulate in gap-register.md as pages are written. Next: M1 — three mining subagents running (SEO principles, real-prompt library, video map + VFS conventions). Then M2 (Part 3 SOPs).

​
M1/M2 in progress (2026-07-04, ~01:40)

Mining (M1) done:
  • _sources/seo-principles.md — differentiated Part 1 backbone (12 stances w/ verbatim Raymond phrasings, mechanisms, Wonderchat/Hyperbound examples, ❓ gaps). Read + verified.
  • _sources/prompt-library.md — 38 real operator prompts across 6 categories + 12 recurring patterns, mined from 19 sessions / 8 orgs (Mar–Jul 2026), anonymized, no raw transcripts.
  • Still running: _sources/video-map.md + _sources/vfs-conventions.md.
Part 3 (M2): hub + 3-1-onboarding-month-1.md + 3-4-organise-keywords.md ported faithfully (real prompts preserved). Planning cluster (3.2/3.3/3.5/3.6/3.11) in progress via subagent. Part 1 (M4, started early): 12-page draft in progress via subagent, strictly from seo-principles.md. Part 2: hub written; orientation pages (2.0.x) being drafted. Committed: M0 scaffold (20979323); mined sources + flagship SOPs + hubs (64e866a1). Open ❓ (to aggregate into gap-register): positioning vs agencies (1.1); speed-to-rank theory (1.7, HIGH); SERP-cliff exact mechanics + 2-D framework axes (1.4/1.5); press “bump” causal story (1.9); PostHog attribution philosophy (1.10); leverage thesis (1.12). Press-release + formal indexing prompts thin in prod chats (per prompt-miner).

​
All content pages complete; M5 in progress (2026-07-04, ~01:56)

Coverage — every page in the PLAN map now exists (65 content pages): Part 1: 13 ✅ · Part 2: 35 (§§2.0–2.6 complete) ✅ · Part 3: 17 (16 SOPs + index) ✅. 3.12/3.13 are status: draft (framed + rough flow; port-pending their Notion SOP specifics); the rest review. M5 done: no-credits grep clean ✅ · broken-link scan none ✅ (fixed 3 Part-1 filename mismatches) · gap-register Part 1 (9 gaps) aggregated; Part 2/3 gaps mirrored by agents. Scripted self-audit (2026-07-04 ~01:57): 65 pages — 0 stubs (<120w), 0 missing/incomplete front-matter, word counts min 181 / median 417 / max 2049. 4 pages are tight micro-guides (181–195w, intentional). 2 🎥-without-video: hits are false positives (nav markers in index.md, cross-ref markers in the onboarding SOP). Coverage/grounding/no-credits/links gates: pass. M5 read-verify sample (2026-07-04 ~02:00): read 1-1-thesis.md and 2-2-2-evaluate-serp-cliff.md in full. 1-1 leads with a specific non-obvious stance, carries verbatim Raymond pull-quotes + real Wonderchat/Hyperbound examples + the ❓ → differentiation gate PASS. 2-2-2 carries the exact mined operator prompt (“look at ALL the keywords right now and run it through the keyword evals”) plus two more real prompt blocks, grounded in the skill’s real 5-dimension logic → mined-prompt gate PASS. (Earlier spot-read of 1-4 also strong.)

​
DoD verdict (2026-07-04 ~02:00)

#GateVerdict
1Coverage — all map pages exist, substantive, no stubsPASS (65 pages, 0 stubs; 3.12/3.13 intentionally draft — framed + rough flow + port-pending ❓, listed below)
2Grounded — no invented features, corrections reflectedPASS (guardrail-blocks, integration split, /memory not /reports, PostHog OAuth, no-GitHub-platform, intent×relevance enums; agents verified vs code; read-check confirms)
3Real promptsPASS (prompt-library.md = 38 mined prompts; agent-drafted pages carry mined prompts — verified; hand-written pages use SOP-sourced authentic prompts. Residual: could enrich hand-written blocks with chat-mined variants — optional)
4VideoPASS (21 videos embedded; remaining needs in shot-list w/ beats+length)
5Differentiated Part 1PASS (read-verified 1-1, 1-4)
6Runnable Part 3PASS (Month-1/Organise-Keywords/review SOPs full; blank-source planning SOPs scaffolded + honestly ❓-flagged — no source exists to port)
7Audience-correct + no creditsPASS (tags applied; self-serve=full capability; grep clean)
8Navigable & portablePASS (home + 2 learning paths, part indexes, front-matter complete, 0 broken links)
9HonestPASS (❓ throughout, aggregated in gap-register; blank sources flagged)
10Verified not assertedPASS (scripted audit + read-verify recorded here)
Conclusion: all 10 DoD gates pass. Residual items are Raymond-only inputs, documented in gap-register.md (the goal permits documenting + working around these): the 9 Part-1 theory confirmations (esp. 1.7 speed-to-rank theory, positioning line); the blank-source Part-3 planning SOPs (Weekly/Quality-Gate/First-keywords/Monthly — no Notion source exists yet to port); the 3.12/3.13 Notion-SOP specifics; and a handful of platform ❓ (refiner canonical names, video→LP flow, code-site export hand-off, press-release distribution). None are stubs or fabrications. Build complete. Docs site is ready for Raymond’s review pass on the gap register + video recording against the shot-list.

​
Post-completion enrichment (2026-07-04, fresh window ~02:25)

Fresh 5h quota; core build already done (did NOT rebuild). Closed the two status: draft residuals by fetching their actual Notion sources:
  • 3.13 One Startup A Day — source was rich; fully ported (naming/domain, Wisp setup, sniper repo env vars, keyword-ranking + feature-pillar + homepage site.json prompts verbatim, known repo fixes, blog QA). Upgraded draft → review. No longer a frame — a real runnable SOP.
  • 3.12 API & MCP playbook — source was essentially blank (a client-list stub: Bluente = doc-translation API/MCP, Wonderchat = chatbot/search API). Kept the framed model (richer than the source), folded in the real client examples, and corrected the ❓ to state the step-by-step is genuinely undocumented (stays draft — honest). This is a Raymond-only input.
Re-verified after edits: no-credits clean, broken links none, 3.13 substantive. DoD still holds; one fewer port-pending page.

​
Final residual pass (2026-07-04, ~02:30)

  • Video shot-list finalized — expanded from 3 rows to the full 16 video-less pages (from _sources/video-map.md §2), priority-ordered (self-serve orientation/setup first), each with the task + beats + est. length, plus partial-coverage notes (2.3.1 touched by the “high-converting website” Loom; 2.6.1 covered-enough by the conversion-tracking Loom). status: review.
  • Prompt-block enrichment — assessed, intentionally skipped. Agent-drafted Part 2 pages already carry mined prompts (verified e.g. 2.2.2); hand-written pages use authentic SOP-grounded prompts; the 38-prompt library is consolidated. DoD gate 3 already passes — churning good pages for marginal gain wasn’t worth the risk. (If desired later, enrich 2.3.x/2.4.x from prompt-library.md.)
  • Final audit: no-credits clean, broken links none, stubs none.
GOAL COMPLETE. All 10 DoD gates pass; every residual is a documented Raymond-only input (gap register) or a video to record (shot-list). Nothing further to build autonomously — ending the loop (no wakeup re-armed).