I kept losing 20–30 minutes fixing broken API responses

Not formatting JSON — actually fixing it.

Things like:

Nothing major, but enough to break parsing and force you to stop what you’re doing.

What surprised me wasn’t that this happens — it’s that it happens in places where you don’t control the source:

In theory, the answer is simple: validate early, fail fast.

In practice, you still end up needing to fix the data just to keep moving.

That small gap between “invalid” and “usable” kept showing up in my workflow more than I expected.

So I ended up automating that part for myself.

What I’m curious about:

Do you usually fix this kind of data manually, or just fail and push it upstream?

Feels like one of those problems that isn’t big enough to get its own category, but happens often enough to be annoying.