Maybe.
So yesterday I helped my brother setup an OpenClaw agent to help him job hunt. He’s been looking for a while, and it’s demoralizing. So I recommended he do that, which of course, turned into me sitting with him. And I used Claude to tell me how to do it. Claude was wrong about multiple steps, but it was still faster to feed it an error and find a solution.
As we were doing it, he and I were talking about how much harder it would have been without me, but in the end, I think it was mostly about motivation. The LLM, despite wrong roads, got us there. And now it is going to look for and apply to up to 10 jobs a day for him. And I realized that any non-computer person with enough desire and stick-to-it-iveness could accomplish the same thing.
So while I agree that for mission-critical software, human programmers will still be used - the blow up of software engineering as a profession was because of the rise of the Internet, and every company of every type needing to be on the web. But WYSIWYG editors like Squarespace and Wix have made huge inroads for small businesses.
think that LLMs are going to make those inroads for apps, webpages, etc soon. And I think it’s going to be “affordable” because companies like Squarespace and Wix are going to figure out how to ameliorate the costs and offer a cheap enough package that it makes sense for businesses to keep using them. Because those customers are already ok with “good enough”. As long as it works - they don’t have some perfect vision about what it’s supposed to look like. The car is shiny, and it drives.
And the thing is, yes, LLMs are expensive, but at scale, they are more affordable than a human being. Take a developer that makes 150k/year. They actually cost about 240k/year with benefits. Plus lots of intangibles, like having office space for them, supplying hardware, paying HR and accounting to handle them. The list goes on.
Now look at cloud computing. AWS contributes anywhere from 50-70% of Amazon’s operating profit. In numbers that’s $10B to $15B a year. They offer cheap web hosting on demand. You get charged for what you use. Amazon pays for the IT engineers, rack space, redundancy, deals with keeping hardware up to date, electricity costs, infrastructure, etc. People using it just pay fractions of pennies per minute of usage.
That is the future that all the big companies are betting on for LLMs. They invest and then over time, people using it makes up the investment - and they have control of a piece of the market. Like the DotCom Bubble burst, I expect at some point, some companies will lose their shirts. But not all of them.
Right now, we are in the “first taste is free” phase of LLMs. But once it’s ubiquitous enough, people will need to use it to stay competitive. Then we take that 240K income and that’s 12k/month to spend on LLM tokens. And if a company needs to ratchet that cost down, they can. No firing/layoffs necessary, and they can change their mind next month without having to hire someone.
Sure, they won’t get the same output quality they would from a human. But if they can get a level of good-enough consistency then it’s worth it. And that’s the danger - the danger of scale combined with the danger of average being good enough.
As for inscrutable code, if the LLM is the only one that has to deal with it, it doesn’t matter. And the only people who will pay for it are the people who will pay through the nose for those of us who are skilled enough to fix those problems. But we will be highly-paid, bespoke troubleshooters. And the field will become much smaller. Those people becoming unemployed from software now will end up reskilling - most likely into labor jobs that LLMs aren’t good enough to do.
(It wasn’t, but…)