FlexMesh didn't come from lived courier experience. SuperRoute's founder approached the owner: "let's build something for fun." The domain knowledge (route optimization, the multi-carrier courier niche) came with the partner; the partnership is visible in the repo (partner section commit, 2025-12-25).
The owner's own reason for saying yes: he had long disliked his own state of having insight in only one dimension — growth, SEO, talking to users were blank axes. FlexMesh was accepted, in part, as a deliberate training ground for the missing dimensions.
This reframes the growth story (growth): the obsessive measurement stack, the phased SEO campaign, the GEO interrogation loop weren't only means to FlexMesh's success — acquiring that 见地 was itself a first-class goal, and that curve did lift: from zero to professional-grade SEO campaigns, a self-invented GEO method (geo-by-interrogation), and audit-grade funnel diagnosis, in ~15 months. It also explains why the project survives a flat user curve — its truest KPI was never only DAU (daily active users) — and why a "for fun" project carries 107 typed analytics events (measurement-stack): play gets no rigor discount.
Interview archetype: "why did you build this?" — a friend invited me to play, and I took it because I refused to stay one-dimensional.
The axes it actually trained (evidence in the repo)
The division of labor made the training structural: the partner brought the routing engine and the domain; the owner brought everything else — and "everything else" was precisely his blank-axis list. The one axis he already owned (backend) got spent sparingly and surgically: a 12-commit BFF (bff-thin-adapter).
| Axis (new to him) | Evidence |
|---|---|
| Frontend | fe repo 123/123 his commits — frontend-from-zero |
| Mobile | the entire Flutter app: bloc, DI, go_router, camera/permission ladder, universal links, iOS build hell (Podfile, TestFlight, APNS — Apple's push service, AASA — the universal-links config file); from 2026-08 the deepest axis of all — the label-detection pipeline and its laws (label-detection-pipeline) with a learned locator baked in (learned-locator, model-locates-laws-decide) |
| Compliance | consent gated in the SDK, E2E-tested, policy-as-lint — compliance-from-zero |
| Legal / corporate | FlexMesh Inc. (Ontario), self-authored 21-section T&C, per-channel tax posture — legal-and-corporate-from-zero |
| Release / DevOps | Docker (2025-05), GitHub Actions → Vercel (2026-02), self-hosted Appwrite, releases through 0.7.5 across both store reviews, a written release SOP (release-sop, one-release-three-builds) |
| Localization | 6 locales, hreflang, DeepL pipeline, a key checker for ARB (Flutter's localization resource files), ESLint-enforced i18n (2025-12-24 — compliance as a lint rule) |
| Content ops | Google-Sheets-as-CMS + a full upload toolchain; blog/city/comparison production lines; 15 feature demo videos |
| Growth / SEO / GEO | growth — the phased campaign, the interrogation loop, the plateau |
| BD (business development) / partnership | the SuperRoute collaboration itself |
| Community / support | Discord integration (2025-09) |
Deepened so far: frontend, compliance, legal/corporate, the BFF, and since 2026-08 mobile (the scan pipeline) and release. The rest are rows until they earn notes.