So I disagree. I’ve been using AI for development for two years. It’s all I use. I just completed a three month hyper focus and the product produced at the end of those three months would’ve taken a year pre-AI.
Would be interested in a healthy debate on both sides of the article maybe with supporting metrics?
Honestly, ever since AI showed up, instead of needing a team of 10 people, you just need 3 ppl with Cursor running at full speed and the job even gets done much faster.
kind of. my team is 3 people and we need to think a lot and use a lot of our expertise while using cursor. ai is still giving out a lot of errors. but indeed, the development proccess is way faster.
99% of the projects below were written via the Agent in the June-July (despite the fact that I have other things to do besides programming). Even the writing and design of the Hub and the Compass repos were done by Claude 4, based on recommendations from my conversation with Gemini, while I acted as the supervising editor.
I would’ve easily spent another two full months just to learn how to write in Python, PowerShell, Shell, Go, and YAML at the level these projects require. Sure, I had some theoretical understanding of Software Engineering, but there’s no way I could’ve written tens of thousands of lines of code in that time.
Here’s another thought: to effectively develop through Agents, you need to be a strong prompt engineer and context engineer. If you’re a senior developer, your experience helps you avoid certain architectural pitfalls — the kind I’m still running into as I learn (both in open and private projects) — but you might actually be worse at coding through Agents, simply because you’re feeding them the wrong prompts.