Building myself a boss

I’ve been writing code for more than fifteen years. Built many side projects. Finished zero.

That sentence does a lot of lying if you leave it alone. The truer version: plenty built, none crossed the last line – listed, marketed, launched, used by anyone but me.

The pattern is always the same. I start with what fascinates – architecture, the hardest technical bit, the design. Hours, days, months disappear. Then, eventually, the interesting part thins out and what’s left is distribution and validation. The boring and… the scary part.

I bail at the last 10%. Every time.

Not a waste

Before this turns into an “I wasted fifteen years” post – it isn’t that. I thought it was, for a while. It isn’t. Side projects were never a failure. They were the input that built the career.

Every one of them sat on the cutting edge of whatever I found interesting at the time. My first job came directly out of a side project. Later jobs came the same way. The stack I learned at night for fun became the stack work paid for six months later. That kept happening. MRR was never the goal; the goal was curiosity, and the career was the compound interest.

The shape repeats across eras. A link-saving app for devs I used myself for years, nobody else knew it existed – marketing wasn’t on my radar, not as something I was avoiding, as something that didn’t register. A game from the Flash/AIR era, fully playable, basically a submit-and-wait step from the App Store – a new project pulled my attention before I got there, and the game just… stopped. A post-ChatGPT tool that started as a joke (church-sermon generator) and grew into a real app. Stuck on marketing, tried to come back, gave up again.

Three projects. Three eras. Same shape.

If I’d pushed back then, I’d have been early at something. I wasn’t.

Cannot not-build

Somewhere in the ADHD diagnosis process earlier this year something structural surfaced. Not in the “productivity hacks for ADHD” sense. There has to be a project on the table. Always. Something being designed, something being built, something being made. Not preference. Not hobby. If there isn’t one, I start one.

This extends well past code. I designed my own house. I designed every piece of custom furniture in it. Not a budget move. Not a weekend hobby. Because not designing them wasn’t an option. That’s my drug.

Which reframes the “why can’t you ship” question. The honest answer isn’t “because I’m lazy” or “because I can’t commit.” It’s that the thing I actually need is functionally unlimited, and the last boring part is finite. Appetite wins every time.

Building the missing input

So what I’m actually building, with pocket.ceo, is the missing input.

An AI boss that lives in Telegram, remembers what I said I’d do, pulls me back when I drift to the next interesting thing, and pushes on the boring part until it’s done. Not another productivity tool. Productivity tools optimize the parts that already work. What’s broken isn’t planning. What’s broken is the lack of someone on the other side of the conversation expecting me to finish.

Three reasons I think it might be different this time:

1. My approach is shifting. I see how marketing actually gets made now, as a thing in the world, not a mystery block to hit at the end. I stopped treating it as someone else’s problem.

2. AI tools changed the boring part specifically. The exact work that used to kill these projects – landing pages, marketing materials, SEO content – is now the work these tools are best at.

3. The product itself is external accountability. Leon, the first persona, exists to do the thing I couldn’t do alone – keep reminding me what’s important, and what’s noise. I’m fully aware this is partial placebo. You have to want it. Leon just makes it easier to keep wanting it.

Still might bail

The last fear – the honest one – is that the harness was the only truly interesting part, and when only marketing and launch are left, I’ll bail like always.

That one I’m afraid of.

But there are reasons it should be different this time.

And this time I have help 😉