And once LLM companies start charging more money than they spend, the “we’ll just have Claude maintain the code” types and vibe coding crowd could either be stuck with extremely heavy subscription fees, or a totally unmanageable code base.

Go for it. I got out of that ridiculous ouroboro of an industry years ago. That might be great for you, but surely you can’t imagine VC funding enough startups to save the software labor market? The tech workforce is giant. Do you think there are tens or hundreds of thousands of ways to make unique new software products where even a single digit percentage are commercially viable? Software isn’t fungible: it needs to solve specific problems that people have, and do it well enough to deal with the hassle of switching software.

This isn’t an individual problem— it’s an industry-wide problem.

(I pulled the 2/3 number out of a hat to illustrate the point. I put exactly zero analysis into that.)

The problem isn’t whether or not companies need us— it’s about how many of us they need (demand), and how many of us there are (supply), because that determines our value. Companies pay people what they’ll work for, not based on how much they contribute to the bottom line; in economics, paying more than you have to for anything, including labor, is the irrational path. A steady, high demand for software developers has kept salaries high because that’s the only way they could get capable people to work for them.

The higher-end of markets aren’t immune to this. As the demand for lower-level workers drops, people will upskill trying to move up rather than get lopped off. Since there are fewer positions the further up the hierarchy you get, you don’t need a huge increase in supply to affect demand. That’s when you start seeing the most experienced, highest-earning people getting shit-canned because someone is willing to do a good-enough version of their role for 2/3 their sizable salary.

This can all happen without a single entire role being completely automated out of existence.

Low effort drive-bys were easy to spot because the amount of code was minimal, documentation was nonexistent, they didn’t use the idioms and existing code effectively, etc. Low-skill drive-bys were easy to spot because the structure was a mess, the docs explain language features while ignoring important structural information, and other newbie gaffes.

One latent effect of LLMs in general is multiplying the damage of low-effort contributions. They not only swell the ranks of unknowingly under-qualified contributors, but dramatically increase the effort of filtering them out. And though I see people argue against this assertion all the time, they make more verbose code. Regardless of whether it’s the fault of the software or the people using it, at the end of the day, the effect is more code in front of people that have to revise code, nonetheless. Additionally, by design, it makes these things plausible looking enough to require significantly more investigation.

Now, someone with little experience or little interest in the wellbeing of the code base can spit out 10 modules with hundreds of tests and thousands of words of documentation that all sorta look reasonable at first blush.