App Factory
What the factory has built so far
The whole plan is a pipeline: find an opportunity, validate it, design and build the app, test it, publish it, then market it. Andrew approves each idea and each new stage before it starts. This page lists every tool that exists today, what it does, and what's planned next. Newest first within each section.
Stage progress
| Stage | Status |
|---|---|
| Find opportunities | Working, Google Play only |
| Validate | Not started |
| Design and build | Not started |
| Test | Not started |
| Publish | Not started |
| Market | Not started |
Opportunity finder (stage 1)
- Grouping of near-duplicate ideas: search terms that are basically the same idea, such as "mma training" and "mma training app", become one entry on the list. 13 Nov 2026, PR #10.
- Fixed scoring: four problems the final review found were fixed, and a new Evidence score lowers ideas with thin data. 13 Nov 2026, PR #5.
- Keyword discovery: finds new search terms from Play's search suggestions, competitor app titles and review complaints, tries them, and keeps the ones that score well. 27 Oct 2026, PR #3.
- Google Play finder: searches Play for each term, reads the top apps and their bad reviews, scores each idea 0 to 100 and keeps one ranked list, best first. Runs by itself every Monday. 27 Oct 2026, PR #2.
AI and shared tools
- Room for the final review: the top-model final review can write longer answers, and failed AI calls now count toward the cost. 13 Nov 2026, PR #9.
- Lessons log: records what we learn, with evidence. Each Monday run adds what it can measure itself. 13 Nov 2026, PR #7.
- AI layer: every AI call goes through one switchable place, with a backup model, a log of every prompt and answer, and a $20 a month spending cap. Cheap model (Haiku) for tagging reviews, best model (Fable) for the final review. 27 Oct 2026, PR #2.
- Mission and rules: the AGENTS.md file every AI session reads first. 27 Oct 2026, PR #1.
Dashboard
- Built tab: this page. 13 Nov 2026, PR #11.
- Scoring tab: plain-language guide to how ideas are scored. 13 Nov 2026, PR #8.
- Lessons tab: the lessons log and a timeline of recent runs. 13 Nov 2026, PR #7.
- Phone dashboard: this website, rebuilt after every Monday run. 13 Nov 2026, PR #6.
- Text dashboard: a status page in the repo (reports/DASHBOARD.md). 27 Oct 2026, PR #4.
Planned next
Nothing here gets built until Andrew says go.
- Garmin Connect IQ finder: the same finder for Garmin watch apps and watch faces.
- Notion template finder: the same finder for Notion templates.
- App Store finder: the same finder for iPhone apps.
- Build stage: turns an approved idea into a working app.
- Publish stage: connects to the Play Console and fills in a store kit per app (listing, data safety, privacy policy, content rating) so publishing takes minutes.
- Knowledge base: books, articles and videos turned into principle notes the stages can use.
- Fix unhappy-user scoring: Play ratings bunch up at 4.5 to 4.9 stars, so this score is nearly always close to zero (lesson L001).