personal_asset

The Indie Hackers Who Saved $25K a Month Are Also First in Line to Be Replaced by AI

He cancelled every SaaS subscription and rebuilt them himself with AI. Four days later he said indie hackers like him might be the first developers AI makes extinct. Judgment you can trust comes from testing, not arguing.

那批省下2.5万美元的独立开发者,也是最先被AI淘汰的一批

Lately, on any tech forum thread that gets more than three comments, eight times out of ten it circles back to the same question: is AI actually going to take programmers' jobs.

It's easy to see why this question keeps getting reignited every few months. Describe what you want in plain language, and AI writes the code and runs it — you just glance it over, tweak the details. There's a name for this workflow now: vibe coding. The stronger the tools get, the more punctually the job anxiety flares up again.

I only made sense of it after stumbling on two pieces of news from July, four days apart.

The person at the center of it is Pieter Levels, an old hand in the indie hacking world — Nomad List, Remote OK, Photo AI, all still alive and running, all his. On July 26, 2026, he cancelled every SaaS subscription he was paying for and rebuilt those tools himself with AI. His monthly costs now come down to just domains, servers, storage, and AI API calls. He later worked out the number in a follow-up post: cancelling those subscriptions alone saved him about $25,000 a month.

Four days later, on July 30, he came out and said something that caught a lot of people off guard: indie hackers like him might be the first type of developer AI makes extinct. The gist of what he wrote: coding used to be hard and expensive, so execution mattered more than the idea itself. That equation has flipped — execution now costs roughly $20 a month, so anyone can build their own software instead of paying for someone else's SaaS.

On the surface these two moves contradict each other. Look closer and they're the same story from two sides. The very mechanism that saved him $25,000 — the collapse in execution cost — is the same thing that just handed indie hackers' old moat, the ability to write code and ship it, to everyone else overnight. He put that new capability to work on himself first, got a real number out of it, and only then made the call. That judgment grew out of a real change in his own bank account, not out of picking a side first and hunting for evidence to back it up. That's exactly what neither side of the online argument bothers to do — nobody tests anything, they're just stockpiling ammunition for the team they already joined.

But a lower execution bar doesn't automatically mean everyone wins. Around the same time, another crowd of indie hackers was spotted making the same mistake: cranking out feature after feature, dashboard after dashboard, auto-publishing pipeline after pipeline with AI — what Levels himself called "AI factories." Peek behind the curtain and most of them have no traffic and no revenue. One writeup nailed the real bottleneck: it was never about whether you could build it — AI already tore that wall down — it's whether anyone sees it or pays for it once it's built. That's a distribution and acquisition problem, a completely different animal from cheap execution. Arguing online and building alone in the dark might be skipping the exact same step: testing against a real market, instead of just talking yourself into your own conclusion.

Bring it back down to the argument ordinary technical people have every day, and it's the same lesson: how far AI can actually go isn't decided by who's loudest or most certain in the comments — it's decided by whoever actually hands it a real piece of work first. Even just handing AI a document or a small script and watching how far it gets you tells you more than ten essays on whether AI will replace programmers. Judgment was never something you figure out before you act. It's something you only get to after you've already acted.

I'm not clean on this one either. Last Wednesday night, curled up on the couch scrolling my phone, I hit this exact debate, thumb already hovering over the comment box ready to jump in — and then I swiped past, remembering that little script of mine that's been sitting there half-finished, untested. If this piece is written for anyone, it's probably for the guy who swiped past that night.

Sources