Every case study at a glance

Showing 46 of 46 case studies

• The Blank-Main Bug — Catching a Production SSR Regression After Promotingappendix
  v—2026-04-20

  Production site shipped 251 bytes of HTML per page across 48 routes. Suspense + useSearchParams in the root layout suspended the entire chi…

  251 — Bytes per page across 48 routes

• Watching the Framework Build the Site That's Replaying It — The DispatchReplay Componentappendix
  v—2026-04-21

  Static blueprint diagrams tell you what a system is, not whether it's running. DispatchReplay plays a recorded trace of a real feature flow…

  2 (Sprint I + fitme-story meta) — Live traces

• The Lego Metaphor — Designing the PM-Flow Ecosystem Pageappendix
  v—2026-04-22

  How to render an 11-skill, 10-phase, 15-data-file framework on a single page without it looking like an org chart. The answer: a Lego wall …

  11 · 10 · 15 — Skills × phases × shared files

• External Validation — Did Our Numbers Hold Up?appendix
  v—2026-04-16

  Independent review of the normalization model, velocity claims, and measurement methodology — confirming what is solid, flagging what is we…

  5 (normalization, velocity, cache, complexity, baselines) — Methodology dimensions reviewed

• What If We Had Measurement From Day One? — A Retrospective ROI Analysisappendix
  v—2026-04-16

  Counterfactual experiment: retroactively applying deterministic measurement infrastructure to all 24 features, then computing the cost, the…

  24 — Features reanalysed under counterfactual

• How We Normalized Complexity Across 16 Different Featuresappendix
  v—2026-04-16

  Raw metrics like wall time and file count are meaningless without normalization. The Complexity Unit (CU) model — additive factors for task…

  16 — Features normalised under CU model

• Framework v7.8.1 — Branch Isolation + Feature-Closure Completeness from One Full PM Cyclelight
  v7.8.12026-05-07

  Two cooperating pre-commit gates shipped as one feature in advisory mode on 2026-05-07. BRANCH_ISOLATION_VIOLATION prevents agents from mut…

  FT2 #244 + #245 + #246; fitme-story #53 — PRs merged

• UCC Passkey Auth — Replacing basic-auth on the operator dashboard with WebAuthnlight
  v7.8.12026-05-07

  Two-PR cross-repo ship that replaces shared HTTP basic-auth on /control-room/* with WebAuthn passkeys. Per-operator identity, per-device cr…

  fitme-story #55 + FT2 #248 — PRs merged

• Bridge to v7.9 — How v7.8 Closed the v7.7 Silent-Passlight
  v7.82026-05-03

  v7.7 shipped a gate (CACHE_HITS_EMPTY_POST_V6) that ran on every commit but exercised data on 0 of 46 features — a textbook silent-pass. Th…

  0/46 (silent-pass) — CACHE_HITS_EMPTY_POST_V6 effective coverage at v7.7 ship

• Smart Reminders Behavioral Learning — PR-1 Shipped Across iOS + Backendlight
  v7.82026-05-04

  Sub-feature of Smart Reminders. PR-1 shipped fully on 2026-05-04 in two halves — FT2 PR #190 (iOS data layer + Settings toggle-off, squash …

  15 / 15 — PR-1 tasks: planned / complete

• Unified Control Center — Migrating an Operator Dashboard Across Two Repos in 11 Dayslight
  v7.82026-05-06

  Retired the legacy Astro operator dashboard, migrated it inside the public showcase as a basic-auth-gated /control-room route, instrumented…

  42 / 44 — Tasks done

• Import Training Plan — Resume from Audit-Flagged Partial Ship to Full Phase 1 Ship in 14 Hourslight
  v7.82026-05-06

  Resumed an audit-flagged partial-ship feature: rolled back to research mid-flight after discovering the original PRD claimed an impossible …

  4 — PRs landed

• Push Notifications v2 — From v1 Partial-Ship to Platform-Layer Rebuild in a Single Daylight
  v7.82026-05-07

  Reopened v1 push-notifications after audit UI-016 caught a substrate-built-but-never-wired partial-ship. Single-session full PM cycle (Phas…

  FT2 #239 — PR landed

• Case-Study Presentation Refactor — Locking Alt-A Across 25 Studieslight
  v7.72026-04-28

  An 18-hour serial sprint locked a uniform presentation pattern across 25 case studies — every one now leads with a SummaryCard, a "how to r…

  25 of 25 — Case studies backfilled

• Validity Closure — How v7.7 Closed the Last Closable Class B Gaplight
  v7.72026-04-28

  v7.7 closes A1–A5 + B1–B2 + C1 from the post-v7.6 gap inventory: 5 new gates (4 write-time pre-commit hooks + 1 cycle-time check + 1 adviso…

  2 of 9 — Post-v6 fully-adopted features

• stats-v2 — Resume, Reconcile, and Ship Despite a Three-Layer Bug Stacklight
  v7.72026-04-30

  stats-v2 was the first paused feature picked up after v7.7 shipped. A "small remaining tasks" job turned into a three-layer reveal: the sta…

  ~2.5 hours — Wall time (resume → ship)

• When the Gate Read One Field and the Data Lived in Anotherlight
  v7.72026-05-01

  The v7.7 case study claimed cache_hits was "100% gated on next write." A 2026-04-30 sweep found 43 of 46 state.json files used the legacy `…

  43 — state.json files migrated

• HADF Phase 2 — Cloud Fingerprinting Measurementlight
  v7.72026-05-01

  Pre-registered measurement experiment to test whether cloud inference endpoints cluster naturally by hardware class via TTFT/TPS alone. Pre…

  0.5566 — Silhouette score (max across k=2..6)

• Mechanical Enforcement — How v7.6 Closed the Class B Gap from Gemini's Auditlight
  v7.62026-04-25

  v7.6 promoted four agent-attention checks into pre-commit failures and added two recurring CI defenses, closing the remaining Class B → Cla…

  7 of 7 — Class B → A promotions

• auth-polish-v2 — Three Workstreams Bundled, 18/18 Tasks Shippedlight
  v7.62026-05-01

  auth-polish-v2 bundled three auth workstreams — forgot-password, biometric activation refinement, Google Sign-In SDK activation — into one …

  18 / 18 — PRD tasks: planned / shipped

• UI-Audit Baseline Burndown — P0 27 → 0 and a Hard Gatelight
  v7.12026-04-24

  The make ui-audit scanner had been running advisory-only since shipping — 27 P0 + 103 P1 across 44 files. This burndown migrated 12 view fi…

  27 → 0 — P0 findings

• V7.0 HADF — Teaching the Framework to Detect Chip Architectureflagship
  v6.12026-04-17

  V7.0 HADF asks whether passive hardware fingerprinting can improve dispatch routing without requiring provider cooperation. 5-layer archite…

  17 — Static chip profiles

• 185 Findings, 12 Critical — What a Full-System Audit Revealedlight
  v6.12026-04-18

  The same AI that built the framework audited its own work across 22 features and 6 framework versions, using a 4-layer methodology (paralle…

  185 — Total findings surfaced

• Building the Site That Tells the Story — A Two-Hour Meta-Buildstandard
  v6.12026-04-20

  PM framework built the website that hosts its own case studies. 37 commits over 2 hours of wall clock. 8 routes, 36 pre-rendered pages, 12 …

  2h 2m — First commit → preview deploy

• The Dual-Sync Race — Two Backends, One Last-Writer-Wins Silencelight
  v6.12026-04-18

  v7.0 audit's top backend finding — two sync paths (CloudKit + Supabase) both pull on login with no merge coordination. Last writer wins. No…

  13 / 3 — Sync findings / critical

• The Stacked-PR Misfire — When "Merged" Didn't Mean "On Main"light
  v6.12026-04-19

  3 stacked PRs (M-2a → M-2c → M-2b). All marked "Merged" by GitHub. Main only got M-2a — downstream PRs merged into their stacked parents, n…

  3 / 1 — PRs claimed merged / actually on main

• The XCTWaiter Abort — Learning to Stop, Rollback, and Retrylight
  v6.12026-04-20

  First M-series attempt-1 failure. XCTWaiter.wait(for: [a, b, c]) is wait-for-ALL, not wait-for-ANY — but the test code reads like 'any'. Ap…

  2 (attempt 1 aborted, attempt 2 passed) — Attempts to ship

• When We Stopped Estimating and Started Measuringflagship
  v6.02026-04-16

  Through 16 features, every velocity claim rested on ±15–30 min wall-time estimates and narrative-inferred cache hit rates. v6.0 instrumente…

  7 of 9 — Measurement DVs now deterministic

• What Breaks When You Run 4 Features at Once — And How to Fix Itflagship
  v5.22026-04-15

  A 4-feature parallel stress test exposed two bottlenecks (52% permission-routing denial, 23× agent variance) — context window pressure was …

  48% — Tool-use reduction with dispatch intelligence

• From "Zero Conflicts by Luck" to "Zero Conflicts by Design"light
  v5.22026-04-15

  v5.1 stress test had 0 file conflicts across 15 same-file edits — by luck, not by design. v5.2 Parallel Write Safety (snapshot/rollback + r…

  15 across 3 files — Same-file edits without conflict

• The Fastest Feature — 86% Velocity Improvement on Auth Flowflagship
  v5.12026-04-13

  Auth embedded into onboarding mid-flow (vs separate hub). Full PM lifecycle plus 3 design iterations in a single 100-min session — first fe…

  86% (2.1 min/CU) — Velocity vs baseline

• First Feature Under the New Architecture — AI Engine Adaptationflagship
  v5.12026-04-13

  First feature where 'how we build' and 'what we build' used the same architectural principles. Adapter protocol, validation gate, analytics…

  45% — Cache hit rate (framework → product transfer)

• Shipping 4 Features in 54 Minutes — The Parallel Stress Testflagship
  v5.12026-04-14

  4 features advanced through 8 lifecycle phases concurrently in 54 minutes. 0 build failures, 0 test failures, 0 merge conflicts across 31 s…

  12.4× — Parallel throughput vs baseline

• Smart Reminders — Six Reminder Types Designed and Shipped Inside a 12-Hour Stress Testlight
  v5.12026-04-20

  Six reminder types, a reusable guest-lock overlay, and a frequency-cap engine — all designed, specified, implemented, and test-covered duri…

  ~12h — Wall-clock from init to complete

• What If You Designed Software Like a Chip?flagship
  v5.02026-04-12

  7 hardware architecture principles applied to a PM framework — LoRA hot-swap, palettization, weight-stationary, UMA zero-copy, mixed precis…

  121,714 → 45,125 — Framework overhead tokens

• SettingsView v2 — 1170 → 294 Lines via Phased Decompositionlight
  v5.02026-04-19

  Closing audit finding UI-002: SettingsView.swift dropped 1170 → 294 lines (~75%) across 4 PRs (#122-#125) in a ~2-hour single-session decom…

  1170 → 294 — SettingsView.swift line count

• Training Plan v2 — Biggest Surface in the App, Stress-Test for the v4.0 Cachelight
  v5.02026-04-10

  The biggest surface in the app — 2,135 lines, 13 nested types, 32 audit findings — broken into 6 extracted views via the V2 Rule. First v2 …

  2,135 lines (largest in app) — v1 file size

• Can You Test AI Output Quality the Same Way You Test Code?flagship
  v4.42026-04-09

  AI-output quality treated as a testable property. 2-layer eval design (golden I/O + heuristic XCTest evals + monitoring schema) and a lifec…

  29 of 29 — Eval cases green on first run

• The Most Complex Feature Completed at Refactor Speedlight
  v4.42026-04-10

  First greenfield feature under v4.4 — new tab, new data model, 9 eval definitions, 5 views. Shipped in 2 hours end-to-end. Stress-tested wh…

  ~2 hours — Wall time (research → Figma screen)

• How 6 Screen Refactors Proved a 6.5x Speedupflagship
  v4.32026-04-10

  Six identical-scope screen refactors across four framework versions isolated framework improvement from practitioner learning. Wall time fe…

  6.5× — Speedup across 6 refactors

• The Pilot — Running the Full PM Lifecycle on Onboardingflagship
  v2.02026-04-05

  First feature run through the full 9-phase PM lifecycle. Retroactive UX/design-system alignment on a "finished" feature surfaced 24 finding…

  24 — Audit findings (Onboarding)

• Home Today Screen v2 — The V2 Rule Pilot and Birth of Screen-Prefixed Analyticslight
  vpre-v5.02026-04-09

  The second screen to go through a full UX Foundations alignment pass, and the first under the now-codified V2 Rule. Home v2 produced two pr…

  1029 → 703 lines (~32% reduction) — v1 → v2 line count

• Backlog Roundup — Two Pre-Rule Features That Stay Roundup-Onlylight
  vpre-v5.02026-04-20

  Roundup of two features that pre-date the 2026-04-13 "every feature gets a case study" rule and have source material too thin to support a …

  2 (development-dashboard, ai-cohort-intelligence) — Features in this housekeeping roundup

• Android Design System — A Documentation Deliverable That Skipped Six Phases on Purposelight
  vpre-v5.02026-04-04

  A 92-token iOS → Material Design 3 mapping that ships zero lines of Android code on purpose. Research + PRD execute normally; Tasks / UX / …

  92/92 — Tokens mapped (iOS → MD3)

• GDPR Compliance — Two Hours End-to-End on a Legal-Blocker Featurelight
  vpre-v5.02026-04-04

  The first feature in the project where kill criteria read "Legal requirement — cannot be killed." Shipped 8 files, +711 lines, full 10-phas…

  2 hours — Wall time (init → complete)

• Google Analytics — The Substrate Every Downstream Feature Depends Onlight
  vpre-v5.02026-04-04

  The feature that went from "11 shipped features, 40 defined metrics, zero analytics instrumentation" to a working GA4 pipeline with protoco…

  22 files / +1970 −39 — Files / lines (merge ac85c73)