Did I solve my own doom scrolling?

TL;DR

I have a doom scrolling problem. X is my poison — specifically the morning reflex where I pick up my phone and spend 20 minutes going nowhere. So I built a tool called ScrollProxy that replaces the feed with a structured morning summary. Here's what I learned building it, and whether it actually worked.

The Full Take

Let me be honest about something: I have a doom scrolling problem.

It's X, specifically. And it's not even the evening wind-down scroll — it's the morning one. Phone off the charger before I'm even fully awake, and the next 20 minutes just sort of disappear. No specific goal. No actual information consumed. Just friction disguised as staying informed.

I've tried deleting the app. It comes back. I've tried grayscale mode. It lasts a week. I've tried the "put your phone across the room" thing. I just walk farther to pick it up.

So I did what I do when I'm annoyed by a recurring problem: I built something.

ScrollProxy is a command-line tool I put together over a weekend. The idea is simple: instead of the X algorithm deciding what I see, I run a script in the morning that pulls signals from a curated set of sources — people I actually want to hear from, topics I care about — formats it into a clean summary, and saves it to a file I can read in about 4 minutes. No feed. No engagement loops. No quote-tweeting hot takes from accounts I don't even follow.

The build itself was a good weekend project. Reasonably scoped, actually solvable, interesting enough to hold my attention. Claude Code handled most of the heavy lifting — I described what I wanted, it built the skeleton, I iterated on the sources and the output format. Four hours, start to finish, first working version. Exactly the kind of project I'd been telling people to try.

But did it work?

Mostly. And "mostly" is more honest than "yes."

The morning reflex is real and it's stubborn. The first three days, I ran ScrollProxy and then opened X anyway, just to check. The pull is deeper than a habit — it's a physiological thing, I think. The algorithm has been training my dopamine response for years. A CLI tool isn't going to undo that in a week.

But here's what did happen: the actual information quality of my mornings went up. The things I was seeing in my summary were the things I'd told the tool to surface — not what the algorithm decided would make me angry or entertained enough to stay. I started my first real piece of work about 15 minutes earlier than usual. I wasn't in a reactive state by 8am.

Small wins, but real ones.

The lesson isn't that I built a magic cure for doom scrolling. The lesson is that you can build a tool for a problem that's specific to your life, in a weekend, without a team or a budget or a production timeline. The problem I had, the tool I built — nobody else was going to build exactly that. It was mine to build because I was the one who understood the problem.

That's the part that stays with me. Not the tool, exactly. The loop.

I had a problem. I described it. I built something. The thing exists. The problem got smaller.

That loop is available to you too. You just have to start somewhere annoying enough to matter.