In recent years, I got gaslit into believing AI is great for Software Development multiple times, only to snap out of it every time as soon as I discovered issues in AI generated code, which rendered typing the solution code by hand in a concise and expressive language like Ruby just as many keystrokes as instructing the AI to fix the code if not less (like when effectively capitalizing on internal DSLs).
As a top-level Senior Software Engineer with a perfect 4.0 GPA Master's degree in Software Engineering with Distinction, award-winning open-source projects, and many conference talks under my belt, whenever I tried using AI in the last few years, I was generally disappointed. It seemed to produce results that satisfy mediocre/subpar developers, but not strong top-level Software Engineers.
That said, I just got gaslit into believing AI can be great for Software Development yet again in Claude Opus 5.5, and for the first time, it seems to be sticking for longer than a week. Could this be it? Will I become an AI fan in Software Development finally? The reason I was impressed by Claude Opus 5.5 in recent Ruby on Rails full-stack development work is because for the first time ever, it seems to speak the same language as Senior Software Engineers' instead of behaving like a Junior Software Engineer. It actually provides multiple solutions with an analysis of pros, cons, and trade-offs, and does unbiased reviews of the code in parallel. As a result, Claude Opus 5.5 has been able to truly produce solutions that seem to be high quality so far. And, it is clear that AI will produce higher quality solutions the stronger the Software Engineering skills of the AI's user are.
If that is the case, I might finally become an AI-driven Software Engineer. But, I have no regrets over pushing against AI in the past as it was truly bad for top-level Software Engineering work, and had bad effects on delivering business value to customers as a result, which escaped the attention of subpar/mediocre gullible developers. In fact, because many strong Software Engineers pushed against bad AI, AI companies came up with better AI, so it was a good thing to critique AI in the past, and many of the valid soft-skill points that were raised in the past still matter for doing high quality Software Development work with AI. Of course, there is a chance this gaslighting phase might be a bubble that will suddenly end too one day. We'll see!
If I end up sticking with AI this time, I do have future blog post ideas about how to use AI in the most effective ways. Time will tell if such blog posts will see the light of day.
No comments:
Post a Comment