Perplexity coding interview prep
Multi-step questions reported at Perplexity, where follow-ups build on code you already wrote and every earlier test still has to pass.
Grow a plain dictionary into a store — then keep extending it while the old tests still have to pass.
set(key, value) and get(key). get returns null for a missing key.
delete(key) and size(). Step 2 runs all Step 1 tests again, so breaking get to make size pass still fails.
How it works
This Rate Limiter question has four steps and 45 minutes on the clock. Each step adds a requirement to the solution you have already built.
“LeetCode teaches you to solve a puzzle once. Anthropic and OpenAI make you build a small system, then keep extending it while the old tests still have to pass. That is exactly what LoopPrep drills.”
The bank
Start with three questions now. The rest cover the same kind of evolving, practical coding problem.
Pricing
No credits, per-question purchases, or daily run quota.
A complete way to experience the multi-step format.
More full-length questions for continued interview practice.
Cancel any time. Your saved progress stays with you.
LoopPrep vs LeetCode Premium
LeetCode is better for breadth. LoopPrep is narrower on purpose: AI-lab style follow-ups, where the system you built has to survive the next requirement.
| LoopPrep | LeetCode Premium | |
|---|---|---|
| Monthly price | $12 / month | $35 / month |
| What you practice | Multi-step systems: build, then extend while old tests still pass | Mostly single-function puzzles, solved once |
| Questions | 33 multi-step questions | Thousands of problems, across many formats |
| Steps / timed practice | 163 steps · 26.7h | Per-problem, not built around follow-up steps |
| Test cases | 1461, visible in a multi-file workspace | Hidden tests, as in a judge |
| Free tier | 3 full questions, every step, unlimited runs | Many free problems; Premium adds company tags and more |
| Languages | Python and TypeScript, in your browser | Many languages, in a judge |
| Best for | Anthropic, OpenAI, and other AI-lab style loops | Breadth, company-specific tags, and algorithm recall |
| If you cancel | Progress stays; premium steps re-lock | Premium features stop at period end |
LoopPrep counts on this page are not typed in: they are filled from the question bank at build time and corrected from the live catalogue on load, so they cannot go stale when a question ships. No annual equivalent is shown — LoopPrep is monthly only, and LeetCode's annual price is a different product choice, not a like-for-like monthly saving.
Questions about the questions
No. Pyodide and the TypeScript compiler both run in a web worker on your machine; your code and results never reach a server.
Python and TypeScript. 30 of the 33 questions ship in both, with the description and the tests written separately for each rather than translated. Pick the language in the question; your work is kept per language.
Yes. Code, step progress and timings persist locally per question. Reset clears one question without touching the rest.
They are interview-style questions: one premise followed by requirements that build on working code as time runs down.
Each company tag comes from candidate interview reports on public prep sites. The bank has questions reported at OpenAI, Anthropic, Perplexity, xAI, Scale AI, and Cursor. A tag means candidates described a question like it, not that the company wrote it.
They are designed to leave room for follow-ups, tradeoffs, and deeper implementation work within a timed session.
You keep every free question and everything you've already written. Premium steps re-lock; nothing you solved is deleted.
Start with a full multi-step question and practice the follow-ups that make interviews challenging.
Start coding now