Case study · 2025
Seed
One generator for web, Ionic, PWA, and desktop apps. It created 5+ apps and became the team's path for SDK and mobile platform updates.
- Runs on
- WebiOSAndroidDesktop
- Built with
- Shared templates · TypeScript · React · Ionic · PWA · CI
- Role
- Full stack. Chose the generator's defaults with the UX lead, then built it end to end, from the review step to the CI wiring.
The problem
Each new app started from a blank repo. Web, Android and iOS through Ionic, PWAs, and desktop each needed their own setup, and every team rebuilt the structure and picked its own patterns.
The cost did not stop at day one. Every SDK bump and mobile platform update had to be worked out again, app by app.
Constraints
- Several targets from one team: web, Android, iOS, PWA, and desktop.
- Mobile SDKs and platform requirements change on their own schedule, not ours.
- Standards written in a document get skipped when a deadline moves.
One generator
Decisions
Put the standards in the generator
Structure, the shared UI patterns, tests, and CI come already wired, so the first commit of any app already follows the team's standards.
Trade-off. Changing a standard means changing the generator, which is slower than editing a wiki page and far more likely to stick.
One template per target, one place for updates
Each target has a maintained template. When an SDK or a mobile platform requirement changed, the work was done once in the generator, and the team moved its apps forward from there.
Review before generating
Teams see the defaults, template, stack, and features, and confirm them before anything is written, so the output is predictable.
Make the right path the fastest one
Routes already shaped, shared patterns already imported, tests already in place. Following the standards became less work than skipping them.
Result
- 5+ apps generated across web, iOS, Android, PWA, and desktop.
- The team's standard path for bootstrapping apps and for SDK and mobile updates.
Code
The original is private. These are public rebuilds of the same ideas, written from scratch.
- scaffold
A new app starts with primitives, tests, and CI already wired.
Read deeper
- Constraints are how AI stays useful
A model will happily write a second platform inside a pull request. Typed edges, a small design-system surface, and generators are what keep that speed inside the architecture you already chose.
- The model starts from zero
A prompt has read the internet and none of your postmortems. The mistakes you already paid for have to live in the primitive, or the next generation will make them again.
The screens are rebuilt in React with sample data. Open them from the work section.
Next: Smart Scan →