How to AI: The Builder
Before Claude Code, I had created experts. I had experimented with "generative" AI without much success (turns out that I am not a text file).
At best, these were time and brain-savers. Claude could write something poorly or spit out some slides. It could be my thinking partner. But couldn't really DO anything.
If this were construction, Claude could help me plan and talk me through steps. But I still had to get out the saw and hammer and nails. I still had to build the house.
Then I used Claude Code.
I still remember, the first time I hit “enter” in Code. I remember marvelling at the thinking, installing, failing, retrying... and then the finished product. Claude gave me something I never could have built on my own.
That’s when I wondered if the world might be changing.
With AI, the limiting factor is no longer your skills, but your imagination. Here are some Claude Code projects from my own life to get your imagination going.
Personal Website
Difficulty: 2/10
Skeptically, I decided to try and build a personal website.
I expected Claude to behave like my advisor. Chat discussions. Pointers on building it. Code snippets, step-by-step instructions, etc. What happened instead blew me away.
I switched over to Claude Code and just like before, I 'context dumped'.
"I want the site to have two halves: personal and professional. Check out this Google Drive folder and this resume for my professional stuff. The main things I want to communicate on the personal page are x, y, z. Here are some pictures. Ask me questions if you’re not sure of something. I'm not that technical so speak to me like a newbie. Go."
It asked me a few questions and then... it started building. It built a site and showed me a preview. I suggested tweaks. Claude implemented them. It looked good. I told Claude to snag me a domain name and publish it.
Claude prompted me to download and sign up for a few nerdy services (Cloudflare, Git, Node.js) along the way. To be clear, I did not need to understand these services. I just needed to download, sign up, etc. so that Claude could use them.
Within an hour and a half, I had a running website on a real domain. The project was silly, but it got my feet wet. I asked myself: “What could I build next?”
Capture App
Difficulty: 4/10
As a GTD guy, my email inbox is more than just messages. It's where I capture my best, most important thoughts throughout the day.
Whether it’s reflections on Kierkegaard or a reminder to grab beans at Costco, I normally lean on a simple voice notes → email app to catch important thoughts in real-time so I don’t forget them. However, about a year ago, my favorite voice notes app suddenly disappeared.
I was mad. But I wondered: Could I recreate it? Could I better it?
I went into Claude Code and did my big context dump. I explained the goal of the app, how it should work, and shared some mockups of what I thought it should look like. I let it know which phone I had, what types of inputs it should take, what each note should look like in my inbox.
It asked me a few questions. I had to download/sign up for a few more apps for the backend (e.g. Android Studio, Resend email service).
A few hours later I had a real app installed on my phone.
The crazy thing? It worked!
Today, after multiple iterations, it sends tasks straight to my task manager, files big thoughts in my second brain, reads text from photos, auto-tags and more.
“Capture” is now my favorite (and most used app) on my phone.
A New Problem
After a few weeks, though, I noticed that with all these thoughts coming through, my inbox was really starting to fill up. My greatest joy (capturing thoughts on the fly) was amplifying the modern thorn in my flesh (hours worth of email to process, every single week).
If only I had an app to process my inbox for me.
Inbox Zero App
Difficulty: 7/10
Processing an email is complex. There are a lot of potential 'actions' for any given email. But while the number of actions is large, it's finite. While pulling out the actions (and decisions) is hard (even for humans), it's divisible and comprehensible.
What if Claude could process my email for me?
My first attempt failed. I tried to do too much at once. But I learned a lot about the problem, and in my second attempt I knew that I’d have to use the experience of fifteen years at a software company to build this more-ambitious app systematically.
My second time: I still can't believe I(?) built this .
With one click, Inbox Zero sorts my entire inbox by action needed (events, tasks, projects, decisions, replies, read-later, junk). It creates a proposal for every email, a detailed explanation for what it’s going to do. I review, tweak items if needed, and click "go".
Inbox Zero then creates calendar events, creates projects and enters tasks with context tags, drafts emails, logs thoughts into my second brain, archives junk, organizes a reading list and more.
Whether I have ten items in my inbox, or three hundred, Inbox Zero quickly crunches them until they are gone.
And afterward? My inbox is empty (and my heart is full).
Conclusion
I don't know if the Terminator's coming for us or not. I don't know what AI will or won't grow into 'someday'.
I do know that I can capture my thoughts on the fly and route them to where they need to go. This very post came from a previously “Captured” thought.
I do know that I have nearly eliminated email from my life. I know that I’ve replaced three brain and soul-sucking hours in front of a computer screen with three hours in the sunshine.
I do know that I am surrounded by people who need experts, but they either can’t afford them or don’t even know where to start.
Whether and how you use AI is up to you. But whatever you decide, AI is coming.
Make the best of it.