| This is what I call the "how hard could it be?" fallacy. It gets devs in trouble alllll the time. > we've all gotten this rigid view that software we work with is fixed and unchangeable and the LLM boom is going to change that by making ALL of the software we use "any shape we want" What? Literally nobody in software engineering has this view lol. We take open source code and libraries and adapt them all the time. And make new ones. Your steel analogy is bad, because you're missing what's complicated about both manufacturing and coding. I've taken welding and shop classes, I could make a motorcycle. Turning a part on a lathe isn't that hard. Bending steel just needs the right tools. So should I build instead of buying, if I want the motorcycle itself and not a hobby project? Haha absolutely not. I'm not buying from Honda because I think vehicles are immutable and unchanging things, I'm buying from Honda because it'll be quicker, cheaper, safer, and far more reliable than anything I could do. If I want a hobby project, sure, but otherwise it's a bad idea. [0] Same thing with code and for the same reason. > The old advice about the time spend writing your tools is tempered by the fact that LLMs make it very very much easier for a focused smart team to build things. Yeah, you misunderstand the problem lol. Building is the easy part. It was never the gate. You can't prompt your way out of understanding what to build. It's so much harder than you think it is. You also can't prompt your way out of the hassle of running a biz critical system, dealing with outages, supporting users, etc. [0] After taking a welding class, you'll instantly understood why it's a trade. Making consistent, quality welds is not easy. |
Uh, I'm not guessing what it will be like, I'm doing it. Both reflecting on a better past when organizations did much more of their computing in house and advocating for modern organization use the new tools we have at our disposal to return to building our own tools.