Raw LLM Responses

I’ve been using AI for coding for a few years now—basically since the very beginning. To be honest, it’s a fantastic tool if you want to write simple boilerplate, but over the last few years, the pace of improvement in the quality of AI-generated code has dropped drastically. The technology still doesn’t go beyond the basics. Assigning anything to AI always involves a deep dive into the code it generated during the PR. Having something to fix in a PR is the rule, not the exception. When it comes to clients where the domain for which the developer is coding is specific (90% of projects), AI doesn't move past basic tasks: comments and unit tests (and even those are very often in need of correction). I’ll be blunt: it’s not the way you’re describing it. The nature of the work has changed; there’s a huge saving on boilerplate, but it’s definitely not the case that AI writes high-quality, production-ready code, and there’s no indication that it will. Non-developer is certainly not capable of preparing proper context for the AI; to do it right, they’d have to be... a developer themselves. Regardless of the above, it’s sometimes sad to hear what most people think of developers: that they’re just someone who bangs away at a keyboard. In reality, it’s not that simple; the profession is much, much more complex than that. But then again, people can only see things through the lens of their own experience.