flâneur

LMCanvas — Linear chat is a dead end.

lmcanvas.ai · 119 words · saved by 1 readers

Linear chat is a dead end. Branch every conversation.

Reproduced. react-query keeps `['avatar']` cached across the logout because the key never changes — B's session reads A's cached body until staleTime expires. Interceptor reads `getToken()` fresh on every request from a zustand store that's synchronously updated on login/logout. No closure capture, no shared singleton. Not the bug here — whatever the cache branch finds is the real cause. Good call — `ureq` drops tokio entirely and shrinks the binary by ~40%. The whole `main` becomes synchronous, no `#[tokio::main]` attr, and error handling stays the same with `anyhow`. Should work as-is — `req

saved by

related reading