Why don't AI coding tools like REST?
TL;DR
A Hacker News post raises a sharp question: why do AI coding tools so often default to RPC-style POST endpoints, even when a simple GET would do, instead of following REST conventions? Is this a deliberate model choice, a token-efficiency artifact, or just a side effect of training data? Better prompting can steer the output, but the pattern is consistent enough to suggest something deeper.
Nauti's Take
Nauti finds the observation valid and useful: knowing the pattern lets you prompt explicitly for RESTful output and get cleaner APIs — that saves bandwidth and keeps endpoints readable. The catch: the post offers no real explanation, and without data it remains open whether the cause is token efficiency, training-data bias, or model preference.
Valuable for API designers working with AI tools — less relevant for teams already comfortable with RPC-style endpoints.
Sources