Yes, I do mentioned about using AI as part my current journey as I found or rather rediscovering SWE passion, AI could do so much as my brain could understand and design some system or service that I wanted to build, using AI code assistant of course.
And by using AI Code Assistant, I have quite enormous code output, I say enormous was compared with my previous life obviously, you can see the commit and push in my Github repo.
One push can have multiple files.
This invites speed traps, I felt it, I can see that this not “normal”, how can I track all changes underneath, the changes done by my AI agent, it’s too fast, too soon, afraid but happy at the same time.
I can set some guardrails, rules, souls, or any kind of approach that industry currently use, and deeply I still have doubt about how can I trust everything the AI tell me or write me to, the codes, the decisions.
Yes, I do keep my skepticism, a healthy one, I need to keep the output in check, how it honor my BDD, PRD and task that I delegate to AI. But the traps still there, it always there, so be careful, don’t too dependent and blindly trust anything from AI, they can spit out anything believable for their users.