Post

Talked to an AI-pilled software engineer friend last night. One of those "I haven't opened my code editor in months" guys. Something I landed on during the conversation: engineers and companies are really underestimating the amount of leverage they're giving to AI providers by going all-in on it

Used to be that engineers had the leverage of being the person who could turn a client's requests into software. The more they offload to the AI, the more they're becoming clients, not engineers. OpenAI/Anthropic are banking on a world where they have a monopoly on the ability to create software.

It's the software equivalent of being a business based entirely around selling products on Amazon. You might be able to make that work, but you're basically a serf on Amazon's land. They can change their rates, fuck with recommendations, whatever, because you're locked in

Right now AI is being used to write code, but if you're already outsourcing code generation, why even own your own codebase? Just let Anthropic do that for you. And suddenly all your "software company" does is pay rent to Anthropic for the ability to make requests to their AI.

It's really sad how engineers have embraced this and raced to find ways to outsource their own jobs, rather than recognized it for the threat it is and started unionizing. What lack of class consciousness does to a motherfucker

5 reposts33 likes

I do understand the productivity argument. But that's not free productivity. You're getting it by giving up ownership of the process, just like you do when you pay for a service instead of building it yourself. Maybe it's cheaper than hiring an engineer, but now that capacity belongs to someone else

The productivity is also entirely illusory. "We commit more lines of code, so we ship more!" Yeah & it's all worse such that every company that goes all in has a temporary run of wins, as it's an excuse to normalize repos & clear out tech debt, only to introduce a new class of problems it can't fix.

For which, the only response nearly always becomes, "spend less tokens, use your own reasoning more & add additional processes & validation around shipping" such that agile teams that go through that adoption lifecycle effectively become waterfall ones to get the constant issues introduced in check.

& everyone, barring those who did keep their editor open & did at least read every thinking block before each & every line went in, such that they were keeping their skills sharp by trying to never let a bad line get through, generated or not, is so atrophied productivity craters in the LLM's wake.

“I don’t want to get left behind,” he said as he enthusiastically funneled company funds into AI, vaguely wondering why he hasn’t gotten a raise in 3 years