AI made progress cheap. Focus became the scarce skill.

I have a strange problem right now: I can make progress on almost anything.

Give me a free morning and an AI agent and I can improve an app, research a company, build an automation, or follow a new idea far enough that it starts to feel real.

AI made it dangerously easy for me to jump all over the place and still call the week productive.

Before AI, procrastination usually looked like procrastination. Now I can avoid the most important thing by doing five other valuable things. I finish the day with commits, research, and a cleaner operating system. None of that work is fake. It just might leave me in the same position afterward.

I saw this clearly in the first week of a publishing sprint I designed for myself.

The goal was to stop working in private and start putting useful work into the world under my own name. I built a proper system around it: one anchor essay, two X posts, one LinkedIn post, and one Instagram post each week. I made a schedule, built an AI workflow to stage the drafts, and added a Sunday review.

The first recorded week ended 0-for-5.

I wasn't idle. I overhauled ThriftWise's interface and started another build. It was a genuinely productive week. I had simply spent it on work that was easier to finish and easier to keep private.

The system worked beautifully right up to the exact threshold it was meant to help me cross.

Dan Koe's recent article *The Art of Strategic Thinking* quotes Bobby Fischer: "Tactics flow from a superior position." Koe's useful point is that the decisive chess move only exists because earlier moves put the pieces in the right places.

I had been treating completed work as the unit of progress. A better question was whether the work left me in a stronger position.

When my own ability to execute was the bottleneck, doing more helped. Now AI makes execution cheap for me, so doing more is no longer enough. I still have to decide which constraint matters, what good looks like, and which result I am willing to attach my name to.

An AI can recommend priorities and draft this essay. It cannot own the tradeoff, the reputation, or the consequence. That part is still mine.

So for the next six weeks I am narrowing the bar.

The thing I keep avoiding is publishing useful work with my name on it. Each week I will publish one substantial piece from work I actually did, then send it directly to three people who care about the problem. I can bring the five-post system back after I do that four weeks in a row.

Any time AI gives back to me has to go toward the work that still requires me: choosing, checking, persuading, talking to people, and standing behind the result. It cannot quietly fund another project.

New ideas wait unless they help the thing I already chose. That rule matters because one of my other failure modes appears after something works. Opportunities show up, I entertain all of them, and the original momentum disappears.

This is not a permanent theory of work. It is a six-week correction for my own pattern.

The recurring problem is not a lack of ideas or capacity. I keep stalling at the same mile, the one where the work encounters other people and my name stays attached to the outcome.

That mile needs a real deadline and a real witness. The deadline was Friday. It came and went. I spent that week on easier, more private work, which is the exact move this essay describes. So this is the honest version: the same essay, posted a week late, with the miss left in.

There is an obvious risk that this becomes another elegant framework I enjoy designing more than executing. The only defense is a receipt.

If you are reading this, I shipped a week late, receipt intact.

If it were still sitting in my vault, I would have built yet another system around the thing I was avoiding.