Async Code & Retries

Start from a crawler that fetches pages on a thread pool, claims each URL once under a lock and finishes cleanly. That version is worked in Concurrency & Systems. In most rounds the hour doesn't end there. The interviewer keeps the same code on screen and changes one assumption at a time, and each change hits a different part of what you wrote.

Three of those changes get their own problem here. First the whole crawl moves onto one asyncio event loop. Then the frontier, the queue of URLs waiting to be fetched, has to know when the crawl is really over. The last one assumes the network fails, so every fetch needs retries, a timeout and a backoff schedule that won't hammer a host that's already struggling.

The Anthropic angle

Asyncio usually comes first, with a question about the GIL close behind. The frontier comes up too. A stop condition that only checks for an empty queue loses pages under eight workers and still returns without an error. For failure handling, it helps to know that Anthropic's own API answers overload with a 529 and its client SDKs retry with backoff. Each step should be a small edit to code you already trust. If you rewrite from scratch at the second follow-up, the interviewer will take it as a sign your first version had no room to grow.

AnthropicPremium lesson · Coding round

The 3 Anthropic problems are below.

1. 1A Premium questionCoding · Medium · Frequently asked
2. 2A Premium questionCoding · Medium
3. 3A Premium questionCoding · Medium

Each with hints, the follow-up and a tested solution.

Get the Anthropic questions

$49 a year · all 100 companies · cancel anytime

From 20,000+ candidate reports, forum threads behind logins and write-ups in Chinese.

Read a free Anthropic lesson firstAlready a member? Sign in