LLMs don't write tests for correctness of behavior but fixedness of implementation.
- LLMs don't write tests for correctness of behavior but fixedness of implementation.
- When tests are expensive to produce, they are nearly always real.
- The creator only bothered to create them if they were useful.
- But when they're cheap, they just become random fixedness of implementation pins.
- The bot just sprayed them in to hold everything in place.