On Sun, Dec 7, 2025 at 10:49 AM Joshua O'Keefe <[email protected]> wrote:
> I've mentioned before I have some level of professional experience on the > back end of LLMs. This keeps me pretty wary of the pitfalls of using them > thoughtlessly but they're great tools within their limitations. The best > way I've been able to use these things in a software development context is > to treat them as junior developers: assume they know very little (they > do!), are worse at the job than you are, and are prone to jumping the gun > (they are!). I don't have as much practical experience as Joshua, but I do have a computer science background and I agree completely. Generative AI can make an expert programmer more productive and even replace the legions of "code monkey" programmers who are tasked with writing the tedious "KLOCs" (thousands of lines of code) needed to get things done on modern systems. But, there's a flip side, when people use Generative AI unaware of its limitations. One of the big ones I've seen is where people expect AI to do things they do not themselves know how to do. It too often leads to results which are worse than simply wrong; they are wrong and deceptively appealing. In the extreme, some people fool themselves into thinking they've accomplished something great when they simply don't have the expertise to see the errors in what the AI is spitting out. (See “vibe physics <https://www.youtube.com/watch?v=TMoz3gSXBcY>”.) TL;DR: Generative AI is only as smart as the user. It is not magic. Not yet, anyhow. —b9
