My whole career I yearned for green field projects but somehow predominantly worked on existing grown code bases and legacy projects.

That naturally meant reading and understanding more code than writing. Sometimes my LOC count was even negative, and I was proud of that accomplishment.

Now with AI I write even less and I've given up on the dream to gain fulfillment that way. The ability to quickly understand large amounts of code from questionable sources, be them machine or human, should hopefully stay valuable until my retirement, especially when supported by AI? What do you think?

▲

pkulak 25 days ago | parent | next [-]

Personally, and this is just my wild guess; I think almost no one is _really_ going to learn to code again. That's not a slight on the next generation or anything; I wouldn't know how to code right now if LLMs existed 25 years ago. I love coding, but it's the _result_ that's the real accomplishment. The journey a bit... but having the result always just sitting there, ready to grab with no effort, is too much temptation to resist.

And as this goes on, folks who can run an LLM _and_ understand/criticize/rework/re-prompt are just going to get more and more scarce. Even using an LLM in my preferred style, where you guide the model through a long series of small steps, will fade away.

	▲	autoexec 25 days ago \| parent [-]
		> And as this goes on, folks who can run an LLM _and_ understand/criticize/rework/re-prompt are just going to get more and more scarce. As long as LLMs keep constantly making mistakes and introducing bugs and humans keep having to verify their output and clean up after them it should mean plenty of work for the few humans alive who can actually understand the code. Future AI models being trained on an increasingly large body of vibe coded bug-filled slop will only make the problem worse. A small number of people with skills that are in demand will tend to make good money, and jobs that make good money will attract more people into learning how to code.

▲

ethanrutherford 25 days ago | parent | prev [-]

Statistically speaking, most codebases are brownfield. Without joining a startup, working on a greenfield project is actually a pretty rare treat. And deleting code is wonderful. I'd go as far as to say lines removed is more often than not more valuable than lines added; every line that's no longer in the codebase is code you no longer have to understand or care about. So long as functionality remains intact, of course.

The problem, as I see it, with prolific use of AI to generate code is that it goes in the exact opposite direction. More and more code is bolted on top of existing code, more and more edge-cases, patch-ups, workarounds, etc. accumulate, the codebase grows and grows. In the end, no matter how good you are at understanding "code from questionable sources", you're still a human being. The AI can generate new code at rates several orders of magnitude faster than you can injest and understand it, and when your meat brain becomes exhausted, the machine does not tire. From a business perspective, your employer will weigh their options: they can wait for you to interpret the code and generate good code (whether by hand or by machine + human review).. Or they can just keep pulling the lever on the slot machine until it works well enough to sell. And for the business exec just looking for the fastest path to paydirt, I'm afraid the latter option is going to look way more appealing.