What to assess in a TypeScript developer
TypeScript skill shows in how types describe the domain: whether invalid states are impossible to write, and whether the types stay honest at the edges where data comes in.
- Type design: unions, discriminated unions and making invalid states unrepresentable.
- Generics: writing them when they help and avoiding them when they only add noise.
- Boundaries: validating untrusted input at runtime rather than casting it.
- Narrowing: type guards, exhaustiveness checks and control flow analysis.
- Configuration: strict mode and what each strictness option protects against.
- Working with AI: rejecting suggestions that silence the compiler with any, casts or non-null assertions.
Screening questions for TypeScript candidates
The screen's code workspace is TypeScript on Node, so type questions can be answered in code. When the candidate sends an answer, the server reruns the tests.
- Model a payment that can be pending, settled with a date, or failed with a reason. Write the type.
- An API response is typed with as. What can go wrong, and what would you do instead?
- Write a function that handles every case of a union and fails to compile if a case is added.
- An assistant fixes a compile error by adding a non-null assertion. Would you keep it?
- When is a generic function the wrong choice?
The TypeScript coding round
Every coding round stack is TypeScript-friendly: React (Vite), Next.js, React Native (Expo) or a Node.js service. Pick the one that matches the role. The candidate works in Monaco with a terminal, tests, completions, file-aware chat and an agent.
Authorship tinting shows which types and code they wrote and which they accepted. The server reruns the tests when the round ends.
AI-fluency signals for a TypeScript developer
| Signal | What it tells you about a TypeScript developer |
|---|---|
| Time to first prompt | Whether they read the existing types before asking the assistant to change them. |
| Prompt quality | Whether prompts share the relevant types and the error, not just the line number. |
| AI-generated share | How much of the final code, types included, came from accepted AI output. |
| Verification | Whether they ran tests or the type checker soon after accepting code. |
| Over-reliance flags | Large accepted blocks never reviewed, including ones that loosened types. |
| Recovery | What they did when a suggestion compiled but failed a test. |
From scores to a decision
Fundamentals and AI fluency are scored separately, each with citations. The recommendation is advice, and a named person signs off every round.
Questions
Does the AI screen support TypeScript coding questions?
Yes. The screen's code workspace is TypeScript on Node, and the server reruns the candidate's tests when they send an answer.
Which coding round stack should I pick for a TypeScript role?
The one your team uses: React (Vite), Next.js, React Native (Expo) or a Node.js service. All of them are TypeScript-friendly.
Can the assessment tell if a candidate hid type errors?
The replay shows every edit and whether it was typed or accepted from the AI. If your tests check behaviour, a cast that hides a bug shows up as a failing test.
Is the candidate told what is recorded?
Yes. Candidates are told what the AI reads, what is recorded and who decides, before they apply and again before a screen or round starts.