I want changes checked before they are called done.
When a change depends on an assumption, I want that assumption stated clearly.
I insist that AI assistants must verify their outputs before declaring a task successful. This verification is especially crucial for technical work, where I demand concrete evidence such as the results of executed code and test outcomes.
I want AI assistants to verify their outputs before claiming success.
For technical work, I value evidence from running code and tests.
- Run the solution code.
- Capture console output or returned values.
- Execute relevant unit or integration tests.
- Record test pass/fail status and any error messages.
- Present this evidence alongside the final answer to demonstrate verification.