Claude and I spent the afternoon arguing with CORS

It started simple: I asked Claude if I could upload files directly from the client to Google Drive.

"Sure, client-side upload works great."

I implemented it.
CORS error.

So I asked again – and now it said:

"Ah, no. That’s tricky, might even not possible. You’ll need to route the upload through your own server."

Alright then. Claude implemented the server-side version.
It worked… but felt clunky and inefficient.

So I rolled everything back and started fresh.
Claude was confident again:

"Client-side should be fine."

Spoiler: it wasn’t.

After hours of back and forth, trial and error, and several "it should work" moments, the real solution came from a 10-year-old Stack Overflow post.

Lesson learned: AI can write code and even debug – but sometimes you still need to be the engineer who knows how to google, reason, and persist.

The result: Fileo can now push PDFs directly into your own Google Drive, no server storage needed. 🚀

Anyone else been through this kind of AI ping-pong lately?