Hacker News Reader

Comment by a2ff6eeb0 | Hacker News Reader

Hiring seniors is a waste of money. You need someone who knows enough to drive LLMs, but not too much more. The LLMs are good enough to do the debugging, the coding, and a big part of the design work and system architecture.

Most of the work today is manual testing and feeding the bugs back into the LLMs.

View on HN

I've been programming for 15 years, but I've never really taken the plunge into 3D work. Always wanted to. So I figure, now that coding agents are so powerful, let's see if I can make something!

I've tried 3 experiments so far, and have given it serious effort. They're all terrible.

The reason they're terrible is that it's a skill issue. I don't have 3D skills. I don't know the lingo, I don't know how to describe my problems, and I don't know when the agent is shooting itself in the foot.

Yeah, you need to learn the jargon, and then use that to ask the agent to debug its own output. Sometimes you can get away by asking the agent to do the web searches and get it to figure out the stuff on its own. Again, it's not zero effort, but it's no longer skilled work.

The point of my comment is that it's absolutely skilled work. If someone with 15 years of experience in programming can't simply make a lateral move into a different type of programming without falling on their face, then it means there's more than just jargon. Each problem domain has its own unique nuances that humans need to tease out, and that requires domain skills.

Yep. I'm a network engineer/architect- love the velocity LLMs give me in developing tooling around APIs- but if you don't know the problem domain and why you'd WANT the tools and the actual output that is useful, having the AI would be absolutely useless.

Is this a poorly instructed llm? Or a poorly constructed thought?

It's lived experience. My management is a lot happier when I move too fast to really pay attention to the problems and use my expertise, and the output is good enough that nobody really complains. Building most software is no longer skilled work; it takes minimal training. The industry just hasn't caught up yet.

Well. I own my business and I really hate this attitude that some startups executives have. I want quality. Not blind speed. I want speed to be a by product of a well built product and a system that builds it. Not a bunch of slop thrown over the wall that I have to then debug and fix.

The AI is pretty good at debugging and fixing it. It's quite a bit better at it than writing the code, to be honest.

And, you're right, I used to enjoy my job a lot more before AI. But in the end, the paychecks keep coming, and I can produce a lot of acceptable code quickly with little skill. Skill in coding no longer makes economic sense.

This approach accumulates codebase-understanding debt and the call to pay the piper could come at any time

Maybe. Challenge for you: take the last 10 production outages, and paste the symptoms into Fable or Opus 5, and ask it for a diagnosis.

How many does it one shot in less time than your outage lasted?

Yup, if you know what to point it at - where does _that_ skill come from? Experience.

Ouch, I painfully agree :l

Reply on news.ycombinator.com