What's the longest task you've ever tried to get Claude Code hrough？

I've been using Claude Code daily for months. Short tasks — bug fixes,  small features — it nails. But anything that runs for half an hour starts going sideways: it self-approves "done," patches symptoms instead of root causes, or just stops at "code written" before deploying.                 

Curious what other people are pushing it through:                         

- What's the longest task you've actually finished with Claude Code?

- Where did it fall apart, and what did you do to keep it on track?       

- Any tricks for stopping it from grading its own homework?

Building something in this space and would love to learn from what's already working (or failing) out there.