Video · Engineering judgment
How a lack of judgment turns an upgrade into a downgrade
A rate-limiting ticket, an implementation that is technically correct, and a corporate customer behind a NAT. The problem was never the code.
“Understand the system, have good judgment, think critically.” It gets repeated so often that it stops meaning anything, so in this video I take it down to a concrete case: a ticket asks for rate limiting on the payment endpoint, 10 requests per minute per user; an AI agent implements it flawlessly at the WAF, with its 60-second window and its Terraform; three days later a corporate customer with hundreds of employees, all behind the same NAT, shares a single bucket and starts getting throttled, while whoever is genuinely abusing the endpoint walks around the limit by rotating IPs. The implementation was coherent, the infrastructure made sense and the tests passed. What nobody decided was what “user” actually means, and that was the whole ticket. Eight minutes, in English, on where engineering judgment ends up once writing the code stops being the bottleneck.
Highlights
- The WAF and API Gateway protect the edge and absorb traffic spikes. Neither of them should be the one deciding which authenticated user is behind a request.
- “Per user” and “per IP” look like the same thing until a corporate NAT shows up: then you throttle together the people you shouldn’t, and whoever rotates IPs walks straight through.
- The idempotency key opens the question the ticket never carried: if the first attempt failed halfway through, does the retry consume quota? There is no universal answer, but somebody has to own the one you pick.