Notes on Claude Code

Six weeks ago, Terminal (that weird 80s-looking app on your computer) was maybe 1% of my working life. Now it's 80%. Because I've been using Claude Code on just about everything.

Haven't figured out the full potential or the limits, but I've thrown a lot at it. A few notes:

What it's helped with
Cleaning my hard drive, rewriting scripts, building games, planning projects, writing code, automations. Some limits for sure. But I've built a whole SaaS app with it. More on that shortly.

The real unlock: MCP
Multi-context protocol. I've wired up Notion, Figma, image generation, my blog, and more. Write a couple of prompts, get a whole content set. I've also started using Claude as a project manager — everything becomes a scoped ticket, then I ask it to pick one off the shelf and work on it.

Agents
I’ve built about 50. Maybe 10 are actually useful globally, another 20 scattered across projects. The rest - learning.

The effect
It's like ADHD x10. I'm more productive than I've ever been. Also context-switching constantly. I think the trade-off is worth it. Time will tell.

The catch
I have to work harder at being creative. The tool doesn't hand that part to you — and it's easy to let it become a crutch. And I’m doing a lot of typing.

On "game over for Job X"
Yes. And no. And yes. This thing can do a high percentage of a sizeable chunk of knowledge work. What that means, I'm still figuring out.

Very curious to hear what people have been doing (or choosing not to)?

I keep wondering if I'm overinvested or if everyone else is underinvested. Probably both.