dontfeartheringo wrote: Mon Aug 11, 2025 6:31 am
AI was supposed to be writing all the code by now.
I can tell you as a professional software engineer that I don't think anyone thought AI would be writing all the code by now. We like to joke around that AI is a senior-level engineer with no real-world experience. It's an assistant, really. It's good at writing code... in a sense that an apprentice is good at measuring the wood for the carpenter to cut it.
It might know how code can work, but it doesn't really know how your system works. It can look at your system and give you guidance for writing something, but it doesn't know how to problem solve without your guidance. It'll give you ideas for solving problems, and explain the reasons if you ask it. But it won't know the problem exists unless you tell it and explain what it is.
I use AI on a daily basis to help me speed up my productivity. But if you let it do your job, you'll immediately regret it. I work with a trust-but-verify method. I have it scaffold my tests, sketch out functions, give me a complicated algorithm to try out, and explain what certain things do. Essentially, it's removed the "google it" method that every good engineer is used to. It still comes with the same pitfalls of needing to verify if an answer is the right one or not.
There's definitely a lot AI can do, but it needs the expertise of the engineer, who has gone to all the meetings, understands the business goals, can handle when things aren't perfect, and has the experience to navigate the complicated environment of a software engineer and all the moving parts of an application, which might include more than one code base working harmoniously with each other.
The other thing is--this being kind of an arm's race with jobs--you need to be using AI and know how to use it effectively because it's now just part of the job. It's like having a good editor, or understanding the latest frameworks or languages. It's now another tool in the toolbox that you are expected to know. And it can improve your productivity, if used effectively. Otherwise it's garbage.
Anyway, I say this because I can't imagine any engineer who worked on AI thinking this would replace actual engineers. I suppose you could say that this could help startups who don't have a lot of money use less engineers, but that's kind of a stretch.
I think what you're saying is hyperbole that is often used from the outside looking in. I can't even imagine actual managers who know how things work considering that AI would write all the code. Maybe one day? I dunno, it depends on how intelligent AI can get. Once it can problem solve, understand business needs, and think on it's own, then we're all in trouble.
So, hopefully divestment will slow it down. I can seriously imagine a world where AI can problem solve and do all the things. It's in its infancy, and right now everyone is trying to understand how to use it and how it's actually useful. It can be useful, just not for everything. It's about as useful as the internet is, just faster and giving you a right or wrong answer.