
100 requests per minute, and 200 got through
Where the simplest rate limiter leaks, and what real APIs use instead.
System DesignInterview questions12 min
Rate limiters, caches, queues and outages, drawn in pencil so you can watch them work.
Three Claude models, 100 answers each. Opus 5.5 picked ramen every single time.
Read the article
Where the simplest rate limiter leaks, and what real APIs use instead.

Three tiers for a small SaaS, with the server in a public subnet to skip the NAT Gateway.

From an empty folder to a live API, and what every git push creates on Vercel.
A popular key expires and ten thousand requests hit the database in the same second.
Everything that happens before the first pixel shows up.
October 2021, redrawn from Facebook's own postmortem.
I send one email when a new article is out, and nothing else.