Putting the Personal in Personal Software, Again

I built a media server and an Android app for it. Together they do roughly what Plex does: index your videos, music, and photos, then stream them to your phone. I called the server portal and the app portal-app. They're both on GitHub now, MIT licensed, free to anyone.

Should Plex be worried? No. Not even a little.

Plex is a business. Their objective is to generate revenue, grow a user base, retain subscribers, and satisfy investors. Those are legitimate objectives. They just aren't mine. My objective was to watch my own videos on my phone without paying a monthly fee to a company that also collects data about what I watch. That's it. That's the whole requirements document.

These apps may not work as published for even a single other person. The server assumes a specific directory structure. The app assumes a specific server. I haven't abstracted anything I didn't need to abstract. There's no onboarding flow because I'm the only one being onboarded. That's not a bug. That's the entire design philosophy.

I wrote about this before, the idea of putting the personal back into personal software. Portal and portal-app are the next examples in that series. Plex has to serve millions of different users with millions of different setups and millions of different expectations. I only have to serve me. That asymmetry is enormous, and it gets more pronounced every time I reach for an AI to help me build something.

That last part is what I keep turning over in my head.

We're still early. The tools are good but rough. The workflows are faster than anything I had two years ago but still require someone who knows what they're doing. I built portal and portal-app in an evening with Claude handling most of the scaffolding. A year ago that would have taken me a week. In two years it might take an hour. At some point the cost of building custom software drops far enough that the calculus changes. Not only for individuals avoiding subscription fees, but for organizations avoiding per-seat SaaS contracts.

That's the concept I'm trying to wrap my head around lately. The economics of it still look cloudy to me. What happens when a mid-size company decides it's cheaper to build a custom CRM with AI assistance than to pay Salesforce? What happens when that custom tool is actually better fit for their specific workflow? What happens to the massive portion of our economy currently focused on the development of software, from students to professors and from junior developers to senior technical fellows?

I can see some stars through those clouds. I can't call it a clear sky yet. But I notice I'm looking up more than I used to.

For now, portal works. My phone plays my videos. Nothing is collected. No subscription renews. And for that specific objective, I'm done.