I vibe coded (๐คฎ) unit tests yesterday for the feature Iโve been working on for the last couple of weeks. It mostly worked OK but enforced my view of not using the AI tools for more than fancy autocomplete, naming and code review. The stuff it wrote runs and does what it should but lacks an elegance I aspire to in shipping code. Also doubt itโs maintainable by anything other than LLMs.
Feel bad for future development on projects that use this stuff for actual code.