I do things that don't scale. AI does them for me.

Why Scale Matters

AI will replace you if you stop getting better, and to stay ahead you have to start by doing things that don’t scale. I’ve been a freelance developer for four years, and I spent a lot of time procrastinating on how to scale before I actually tried to scale. My mistake was trying to learn scaling first instead of scaling first.

Do Things That Don’t Scale

A month ago I heard a second‑hand summary of Paul Graham’s article “Do Things That Don’t Scale.” The article isn’t new, it’s more than ten years old, but it became one of the best business lessons I’ve learned recently. One line that stuck with me is: you can’t wait for users to come to you; you have to go out and get them.

As a developer I could build a complete SaaS product and then try to sell it, but that approach doesn’t work as well as talking to real clients, learning their problems, and then building a solution for them. Even if the resulting software doesn’t scale massively, it can still make me a lot more money than spending all my time studying scaling without delivering a solution.

AI‑Driven Outreach

The twist today is that AI lets us do the unscalable work at machine speed. I’m testing a process where I look at a prospect’s website, identify problems, and write a personalized email that includes a proof‑of‑concept redesign of their site. I send the email with the new version attached and ask, “Do you want to use this?” It isn’t a sales pitch; it’s a finished project delivered before the conversation even starts.

The work is still guided by me. I set the standards each task must meet, and the AI asks for my approval before taking any action. My four years of experience lets me make the right decisions, while AI handles the repetitive parts. This combination lets me adapt to each client in minutes, turning my knowledge into a scalable service.

Learning Before Automation

Doing things that don’t scale teaches you. When you perform a task manually, you learn the problem and the solution. Later, when you automate, you already have the muscle memory to know exactly what to do. With AI, it’s still essential to understand the fundamentals behind each task and decision; that knowledge keeps you irreplaceable.

So the core message is: keep getting better. Learn how to learn, study, and solve problems with your own mind. Let AI handle the execution, giving you free time that you can spend on deeper learning instead of scrolling through meaningless content.

Building a Tool

While I was reading Graham’s article, I built a desktop‑app MVP with AI in a single day. The app aggregates my Claude Code workflow into one view: a folder tree, a markdown reader/editor, and a real terminal on the side. It’s not an SDK wrapper; it requires the user to have their own Claude subscription, which avoids platform‑specific issues.

If you have an Anthropic subscription and want to test it, leave a comment below and I’ll send you a copy. No charge, just a chance to be an early tester.

That’s it for today. See you in the next video.