APIs
APIs: free-recall review
Retrieval beats re-reading. For each prompt, reconstruct a full answer from memory before you open the model answer — the effort of pulling the cascade back together is what makes it stick.
Reconstruct the track’s spine without looking back: how modeling enables honest status codes, why the status code is a retry contract, when cursor pagination is mandatory, what the OpenAPI break diff protects, where protocol sits, and how rate limits cap the blast radius.
- 01Walk the cascade: how does one resource-modeling mistake become a production outage, link by link?
- 02Why is an HTTP status code a contract rather than decoration, and what does each class commit the client to?
- 03When is cursor (keyset) pagination mandatory over OFFSET/LIMIT, and what is the tradeoff?
- 04What is a breaking change versus an additive one, and what does the spec-first OpenAPI + CI break diff actually buy you?
- 05A teammate wants the public API to be gRPC because the benchmarks are far faster. How do you reason about it?
- 06How do rate limiting and idempotency together contain a retry storm, and what must the throttle response include?
If you can reconstruct each answer from memory, you hold the track’s spine: modeling nouns is what lets status codes be honest; status codes are a machine-readable retry contract backed by Idempotency-Key and Retry-After; cursor pagination keeps large/live reads flat and stable; spec-first OpenAPI with a CI break diff stops silent drift; protocol is an orthogonal axis (REST at the edge, gRPC internal); and rate limits plus idempotency together cap the blast radius of a retry storm. It is one cascade, reviewed as a chain.