dandi8

joined 2 years ago
[–] dandi8@fedia.io 1 points 3 days ago

Unit testing with LLMs is just asking an AI to hallucinate requirements.

Tests are what documents expected behavior and are therefore the worst candidate for code gen.