2 ms·
Auto Playwright
- verdverm 3y agoThis requires calls to an LLM every time I run the tests 1. How reliable is this given what we know about LLMs? 2. Will this make the tests unbearably long? 3. Will this also make testing much more expensive? 4. Why not have the LLM use the playwright UI to create the code, then run this generation step when the tests need updating? Or maybe just generate the code directly? In the end, I think LLMs are best used as assistants, not replacements. Here, we are turning too much over to the AI, with no human in the loop 5. How would you know the LLM correctly executed the operation and didn't do something else that still passes during testing?