Rung
Practice React the way an interviewer asks: twelve exercises, a real editor in the browser, feedback you can trust.
- Role
- Product, design and engineering
- Stack
- Next.js 16 · React 19 · TypeScript · Sandpack · Groq · Tailwind 4 · Vitest · Playwright
The problem
Going from a junior to a mid-level React developer is mostly about the details interviews test: state, effects, memoisation, data fetching. Reading about them is not the same as getting them wrong and seeing why.
The solution
Rung gives twelve exercises, each in an editor that runs React in the page. When you submit, deterministic checks decide whether it passes, and then a model explains what went right or wrong without handing over the answer.
What it does
- Twelve exercises with a live preview
- Pass or fail decided by code, never by the model
- AI feedback with explicit criteria, so it cannot invent a rule
- 81 automated checks across the twelve exercises; a rate limiter that falls back to memory if its store is down
- Tagged v1.0.0, with a test suite
Decisions
The model explains, the code decides
An AI that grades is inconsistent. Here the tests decide, and the model is only asked to explain a result that is already known.
Checked against the real service
The original model had been retired by the provider and answers ran longer than asked. I tested against the real service, not only mocks, and made the validation tolerant of what models actually return.