One of my GitHub experiments was more about attention than about a new software product. I built a system that found people whose profiles matched certain criteria, followed them, and counted on some of them following me back.

After following close to 2,000 people, I had roughly 800 followers. Then I used AI to write a program that unfollowed those accounts.

My own summary of the result is that I got about 800 followers without doing anything impressive in software. That's the center of the story. The number was real, but it doesn't say what someone might assume it says about my projects.

Automating a social action

The system had a selection step before the follow. It wasn't picking accounts at random. It looked for people whose profiles matched criteria I chose because those people seemed more likely to respond.

So it was a software workflow with a social result. Profile information drove an action, and people's responses changed a number on my account.

The implementation details matter less than that relationship. The goal was to raise the chance of a follow-back. Follower count was the thing I was watching. It was not a measure of how many people installed, used, or reviewed anything I'd built.

I also don't have a controlled measurement that ties each follower to a specific outgoing follow. I remember the scale of the experiment, and that's why every figure here is approximate.

The cleanup was another programming task

Once the following phase was done, I had AI write the program that unfollowed those accounts. It's another small example of how I used AI: I had a repetitive account task and asked for software to do it.

The cleanup changed the outgoing relationships. It didn't change how the follower count grew, and it didn't turn that count into proof of an audience around a particular project.

Those distinctions keep the outcome in proportion. I automated a workflow, got the approximate result I described, and then automated the cleanup.

What the number didn't tell anyone

A GitHub profile is a compressed view of someone's activity. A visitor sees repositories, descriptions, contributions, and follower counts before they've looked at any of the actual work.

A count can give an impression that goes further than what it measures. Following an account is an action someone took. It doesn't tell you whether that person read the code, understood a project, tried a demo, or found an app useful.

The same count can come from very different reasons. Some people follow because they like a project. Others respond to a social interaction. The number alone doesn't tell those apart.

That doesn't mean the followers were fake. They were real accounts connected to my profile. The point is that the way you read the metric has to stay close to what actually happened.

Putting the experiment next to the rest of the portfolio

My project history includes games, photo tools, Raspberry Pi experiments, a moderation bot, business websites, and AI media pipelines. Those projects can be described by the problems they solved and the things I got working.

The follow experiment sits alongside them as a different kind of build. Its main result was a change in my account's social numbers. It didn't show the same things as syncing player positions, running a game across a phone and a TV, or assembling an AI video workflow.

That difference is useful. It adds context to the public profile instead of asking a visitor to guess how good the software is from a follower count.

Keeping the story honest

I won't turn these approximate numbers into a precise conversion rate. I also can't say all of those followers became engaged readers, users, or customers. The experiment didn't establish any of that.

What I can say is narrower: criteria-based following at a scale close to 2,000 accounts, about 800 followers afterward, and an AI-written program to unfollow.

It's a software story, a social experiment, and an example of a metric that needs explaining. In a collection about things I've built, explaining where that number came from is more useful than letting it stand in for the work.