How to spot blind trust in AI coding tools
An engineering manager says their team rolled out Claude and Cursor three months ago. They do not know whether engineers are thinking through AI output or simply accepting it. Problems show up after code is pushed and PRs are opened, including CI failures and branch conflicts.
Key points
- The team started using Claude and Cursor three months ago.
- The poster worries that engineers may be accepting AI output without enough thinking.
- Issues appear later when PRs are opened and CI fails.
- One recent problem required redoing PRs because work started from the wrong branch.
- The poster sees both AI use and task communication as possible causes.
Quick term guide
- engine
- The core software that provides the basic functions for a game to run.
- Thinking
- A ChatGPT mode where the AI reasons through a problem step by step before giving an answer
- Pro
- A paid plan that usually offers more features or higher usage than a free plan.
- branch conflicts
- A situation where two sets of code changes clash and must be fixed by a person.
- branch
- A separate line of work inside Git.
- Solo makers
- People who build and launch their own products or services entirely on their own.
- AGI
- AI that could handle many kinds of work at a human-like level.
- draft
- A first version of a piece of writing.