I was skeptical until 3-4 months ago, but my recent experience has been entirely different

I'm seeing it too, but there's a distinction I think matters: AI isn't replacing the thinking, it's shifting where the bottleneck is. You mention systems are self-improving and code quality has increased dramatically. But the constraint isn't execution anymore. It's judgment at scale. When AI collapses build time from weeks to hours, the new bottleneck becomes staying current with what's actually changing. You need to know what competitors shipped, what research dropped, what patterns are emerging across 50+ sources continuously. Generic ChatGPT can't do that. It doesn't know what YOU care about. It starts from scratch every time. The real question is how do you build personal AI that learns YOUR priorities and filters the noise? That's where the leverage is now. Excited for the future :)

Agreed. I don’t know if it will create or eliminate jobs but this is certainly another level from what we’ve seen before. Since last 2 months, calling LLMs even internet-level invention is underserving. You can see the sentiment shift happening last months from all prominent experienced devs to.

My guess: projects "learn" every time we improve documentation, add static analysis, write tests, make the API's clearer, and so on. Once newly started agents onboard by reading AGENTS.md, they're a bit "smarter" than before. Maybe there's a threshold where improvements become easy, depending on the LLM and the project? As a hobbyist programmer, I feel like I've been promoted to pointy-haired boss.

But still the hypothesis of the article holds stance. If you want a new feature, you still have to think it through and explain the AI how to implement it, and validate the result. You might be able to do more with less, but that is with every technological advancement. Regarding your experience, it sounds like your codebase is such good quality that it acts as a very clear prompt to the AI for it to understand the system and improve it. But I imagine your codebase didn't get into this state all by itself.

If these agent are so great was isn't ChatBotKit a highly successfully public company worth hundreds of billions and not just a glorified chatgpt wrapper? If you're able to do so much with so little why isn't that actually bearing out in becoming a profitable company? What's the excuse? Do people really need to know that a bunch of code at a company that won't exist in 10 years is something worth caring about?