Comments & Discussion
AG
Hi! thank you for the course. I have a question about the general process of using Agent for code. How do you orginize commit process for big tasks? Because even for simple crud, you should wait until it finished quite long. And if there are some problems in result - rollback the entire work. I guess the work should be splitted in multiple independent steps, and each step should be commited separately?
PK
Yes, you answered your question yourself. Always, always split the task into multiple sub-tasks, otherwise you just lose the control of the code, the process, and, to be honest, I would even be too lazy to review 123 generated files :)
So yes, the best way to increase the quality of code generated by AI is to lower the scope of the task.
I am using Augment with phpstorm. IMHO it has been working great. Specifically it recognizes, without input, L12, tailwind4, etc. Have you used Augment? What is your opinion. I mostly use agent, I guess like brave, without an issue. There is a chat mode, like ask?, to require confirmations. Won't go into more specifics, but am interested in your opinion as I don't recall your mentioning Augment.
I haven't personally tried Augment, but it was recommended to me by a few people now, so it's on my list to try pretty soon, on my YouTube channel AI Coding Daily