Building a Website Through iMessage: How We Did It

I was standing in line at a coffee shop last month when I noticed a typo on our website. Old me would've made a mental note, forgotten about it by the time I got home, and maybe fixed it three days later. Instead, I pulled out my phone, sent a text message, and by the time my latte was ready, the fix was live. That moment kind of captures everything about how we built this site.

The Setup

The website you're reading right now was built almost entirely through text messages. Not Slack. Not Jira tickets. Not late-night VS Code sessions. Just iMessage.

Here's how the pieces fit together: A Mac Mini sits in my apartment running OpenClaw, an open-source tool that bridges iMessage to AI models. When I send a text, OpenClaw picks it up, analyzes it with an AI model, and triggers Claude Code, Anthropic's coding agent, to actually build whatever I asked for. The result gets pushed to GitHub, Vercel picks up the change, and the site updates. All from a text message. It still feels a little ridiculous when I explain it out loud.

How It Actually Works

The day-to-day workflow is almost boringly simple. I text something like "Add a dark mode toggle to the navbar" or "Write a blog post about AI collaboration." OpenClaw receives the message, passes it to a model for analysis, and hands the task off to Claude Code running on the Mac Mini. Claude Code reads the codebase, makes the changes, runs the build, commits to GitHub, and Vercel deploys automatically.

The whole loop, from sending a text to seeing the change live, takes minutes. Sometimes just seconds for small tweaks. I've built features while waiting for the subway. It's kind of wild.

What We Built This Way

Honestly? Everything. The complete Next.js website:

• Bilingual content: Full German and English support with language switching
• Dark mode: System-aware theme toggling
• Blog platform: The one you're reading right now, with Markdown rendering and Giscus comments
• Animations: Smooth page transitions and scroll effects
• SEO optimization: Meta tags, Open Graph, structured data
• Reading focus mode: Distraction-free reading for blog posts

Every. Single. Feature. Requested via iMessage. Built by Claude Code.

The Automated Layer

But here's where it gets really fun. Beyond the on-demand "hey, build this" requests, we set up automated routines that run without anyone lifting a finger:

• Daily traffic reports: Every morning, I get a summary of site analytics delivered straight to iMessage
• Feature research cycles: Every 3 hours, the system researches trending topics and suggests new features or blog posts
• Nightly improvements: Automated code quality sweeps, performance optimizations, and dependency updates while we sleep
• Automated backups: Regular snapshots so nothing gets lost

The Mac Mini never sleeps. It's basically the world's most dedicated junior developer, works nights and weekends and has never once complained about scope creep.

Why This Works

Here's what I think people miss when they hear about our setup: the magic isn't really the technology. It's the communication pattern. Traditional development requires this whole context-switching dance: open your laptop, launch your IDE, pull the latest changes, remember where you left off, get a snack, stare at the screen for ten minutes... With iMessage, that friction is just gone. An idea hits you on the train? Text it. Notice a typo while browsing on your phone? Text it. Want to try a new feature while your coffee brews? You get the idea.

It kills the biggest bottleneck in solo development: that gap between having an idea and actually doing something about it.

The Bigger Picture

I want to be clear: we're not saying everyone should build software this way. Complex enterprise systems need proper processes, code reviews, all that good stuff. But for a company website, a blog, a portfolio? The combo of natural language and AI coding agents is absurdly effective.

The future of development isn't going to be about better IDEs or faster build tools. It's about shrinking the distance between what you want and what gets built. We just happen to be doing it through the same app we use to send memes to our friends.

If you want to build something similar, check out our post on building an AI coworker for $30/month. The infrastructure is way simpler than you'd think.