Experimental Apps
A running set of small experiments — most never ship, and that is the point.
- Status
- EXPERIMENT
- Date
- Ongoing
- Stack
- Next.js / React / AI / TypeScript / Cloudflare
The problem
Every technique used in a client project has to be learned somewhere first, and a paying project is a bad place to find out that an approach does not work.
The idea
Keep a permanent scratch space. Small, disposable builds with one question each, finished in a weekend or abandoned without ceremony. The ones that survive graduate into real builds.
Building it
How it came together, step by step.
One question per experiment
Each build answers a single question — can this interaction feel instant, will a model handle this reliably, is this pattern worth the complexity.
Editor
A weekend, then a decision
Hard time limit. At the end it either graduates, gets archived, or leaves behind one technique worth keeping.
Interface
Recording what was learned
The write-up matters more than the code. Most of these exist as a paragraph and a screenshot.
Terminal
Shipping the good ones anyway
A few end up as small public tools, even without a plan for them.
Mobile
Features
What it actually does.
Interface prototypes
Interaction patterns tested in isolation before they reach a real product.
AI experiments
Probing where models are genuinely reliable and where they quietly are not.
Performance tests
How fast a given approach can actually be made, measured rather than assumed.
Throwaway tooling
Small scripts and utilities built to solve exactly one recurring annoyance.
Public write-ups
Each experiment documented on Instagram, including the ones that failed.
Graduation path
Experiments that hold up become numbered builds.
Tech stack
What it is built with.
- Next.js
- React
- AI
- TypeScript
- Cloudflare
The result
Where it ended up.
- Experiments run
- 20+
- Graduated to builds
- 4
- Time budget each
- 1 wknd
- Documented
- All
Build in public