How to Stay Focused When Everything Wants Your Attention (Including the AI You Built)

TLDR

In the 30 days before I wrote this, I touched 42 different project folders, on top of a full-time sales job. Everything in my life is currently asking for my attention, including several AI systems I built specifically to help me focus. So I went back to two teachers: Matt Larsen’s free six-hour agency course and Nick Saraev’s “9 Years of Business Advice.” Then I compared what they teach to the focus rules I’ve had to hard-code into my own tools over the last seven months. The single most useful number I found wasn’t in either course. It was in my own history: work I start the same day gets done 100% of the time. Work I put on a list gets done about 5% of the time.

42 Folders in 30 Days

Let me show you the problem with receipts, because I think a lot of builders will recognize it.

In the 30 days before I sat down to write this, 42 top-level folders in my build directory got touched. Group the ones that belong to the same product and it’s still around 20 separate client and business threads in one month. The running handoff log my AI sessions share has 11 different projects tagged in it. In late March, in one eleven-day stretch, I shipped work across 8 different projects, including one sitting that produced 32 commits across 16 pull requests.

All of that happens around my 9-5 in sales.

Here’s the uncomfortable part. From the inside, that volume feels like productivity. It feels like momentum. And some of it is. But when I actually audited my own history, a lot of it turned out to be the most comfortable way to avoid the one boring thing that would have moved the business forward.

The Weakest Link Is You

Matt Larsen’s six-hour course is sold as agency training, not a productivity course, and that’s what makes his focus advice land. He treats focus as part of the business math, not as self-help.

His frame: your results come from three things multiplied together, your offer, your funnel, and yourself. And you’re only as strong as the weakest one. From the transcript, lightly cleaned up: even if your offer and your funnel are a ten out of ten, “if your ability to focus, your skills, your habits, your knowledge is only a two out of ten, that means your entire agency is going to be a two out of ten.”

That’s a brutal sentence if you’ve been polishing your offer and your funnel while your calendar is a mess. It was for me.

His practical tools are old-school on purpose:

• An information diet. “Other than oxygen, content is the thing that we consume the most.” He says YouTube and news have probably cost him twenty times more money than every bad hire or business mistake he’s made. He blocks sites, and uses a physical tag he has to walk across the room to get before he can unblock anything.
• A kitchen timer, not a phone. Sixty-minute sprints, then a quick log of what you actually did that hour.
• A printed schedule in 15 to 30 minute blocks. How do you eat an elephant.
• Three to five things a day. He tells a story about Steve Jobs via Kevin O’Leary: there are only two to five things you need to get done in the next day, and 80 to 90% of your time should go to them. His rule is not going to bed until they’re done.
• Burnout is a progress problem. This one reframed things for me: “the loss of motivation and burnout both come from when you feel like you are no longer progressing.” Working hard at something that stopped teaching you anything is what burns you out, not the hours by themselves.

Inputs, Lag, and the Boring Lever

Nick Saraev’s “9 Years of Business Advice in 63 Minutes” hits the same themes from a different angle.

Focus on inputs, not outputs. He tells a story about writing a blog post every day for 31 days and ending up with three subscribers. It felt like failure. Years later he found out those posts had quietly ranked on the first page of Google. “Anytime that you focus on outputs you’ll only really ever get this far, but if you focus on inputs, you will massively overshoot it.”

Account for the lag. Results show up long after the work. He compares it to network lag in a video game and says you’re better off assuming the lag is longer than you think. Larsen says the same thing with a number: 90 to 180 days between actions and results, and most people quit inside that window.

Structure beats willpower. “Don’t ask yourself how can I do X, ask yourself how do I structure my life in a way that I can do the thing I want to do every day for a year.” I’ll come back to this one, because it’s basically the thesis of everything I’ve built.

Pull the boring lever. “It’s actually more powerful to say no than it is to say yes.” He describes business as a minefield of shiny objects, and says the dull, boring thing is where 90% of your effort should go. Find the lever that pays a dollar and pull it all day.

Decide at 70%. Waiting for the last 30% of information usually costs more than it’s worth.

Do it yourself once before you delegate it. “Anytime in my business that I’ve ever delegated something I haven’t done before, the outcomes have been absolute trash.” I’d apply that to AI too. Hand an agent a workflow you’ve never done yourself, and you have no way to judge what comes back.

What I Learned by Building Focus Into My Tools

Here’s where my experience adds something neither course covers: what happens when you try to give focus rules to an AI, and to yourself, through software.

Back in March I wrote about the AI agent that inherited my worst habit: I’ll start a task, notice something adjacent I could automate, spend three hours automating it, and realize the original task would have taken twenty minutes by hand. I built a check for that called /insanity. If I’m three steps into automating something that would be faster done manually, it’s supposed to stop me.

In July it failed in the most instructive way possible. In one session, my AI built and certified a full production release twice before I’d seen a single pixel, then burned through six different browser-automation workarounds when a working API route already existed. My message afterward: “this was a huge waste of tokens,” and “this would have been a good opportunity for the insanity skill to kick in.”

It didn’t kick in, and the reason is written into the file now: “the state where this skill is needed is exactly the state where judgment about needing it is offline.”

That’s true for people too. The moment you’re deepest in a loop is the moment you’re least able to notice it. So the check stopped relying on judgment and started counting. It fires automatically when:

• SPEND: two or more full build-and-verify cycles, or three or more helpers launched, since I last weighed in
• REDO: it’s about to rebuild something it already built this session with nothing new from me
• FIDDLE: it’s reaching for a third workaround toward the same goal
• AUDIENCE: it’s polishing something I haven’t even looked at yet

When one fires, it asks one plain question about the cost and whether to keep going, then respects the answer.

Earlier, in June, I’d said a version of this to my AI out loud after it spent an hour on four failed attempts at an official login flow when a simpler path already worked: “you can tell me if I’m wrong. I don’t want to go down any rabbit holes. I don’t need to.” The rule from that one is my favorite for humans: if the manual version takes less time right now, do it manually and don’t build anything.

The 100% vs 5% Number

In June I had my AI system audit my own notes, plans, and history. It came back with a sentence I didn’t love: “building is the dopamine, the boring revenue move is the deferral.”

The evidence was hard to argue with. At one point I had 267 content pieces queued and zero posted. There was a month where I was averaging seven and a half builds a day while a finished client site sat undelivered.

But the most useful number in the whole audit was this one:

• When I made the first move on something in the same session I decided to do it, it got done 7 out of 7 times.
• When I put something on a list to do later, it got done about 6 out of 122 times. About 5%.

That’s not a productivity tip. That’s a law, at least for me. So I turned it into three tripwires:

1. A plan that ends with no move, no date, and no kill note is decided against by default. If I’m not willing to start it, schedule it, or kill it, I’ve already decided. I just haven’t admitted it.
2. The third time I say “I need to” about the same thing, it’s a stall, not a plan. Across more than 30 tracked “I need to” statements, fewer than 20% got done within a week.
3. No new build starts while a client or revenue action is waiting on me. This is Saraev’s boring lever, made mandatory.

When Your Own Tools Start Nagging

Here’s a problem neither course prepared me for: the tools I built to keep me focused started competing for my attention.

My AI used to end every completed task with a list of “carried items,” loose ends it thought I should know about. Every single time. In September I said: “For carried items, I think let’s just do those once a day, or maybe once in the morning, once in the evening, and that’s it. Not every single time.” Now it’s enforced by code. The list can physically only show up twice a day.

Before that I’d noticed a pattern in its reports where every one ended with some leftover problem to worry about: “it almost feels like you feel obligated to find a few things that you couldn’t do, and that’s what you end with.” And earlier, the line I think about most: “It should be more like extra credit than extra homework.” Fix the small thing and tell me you fixed it. Don’t hand me another item to carry.

It kept happening in other places. A paid AI checker I built started running on its own and cost $52 in ten days, so now it only runs when I type its name. My always-on agents were sending me the same two suggestions over and over, zero of ten accepted, so I turned their scheduled updates off.

The lesson: every tool that can talk to you is a tool that can interrupt you. Rate-limit your tools the same way you’d rate-limit your inbox.

Where the Experts Disagree

I don’t want to hand you a neat summary where everyone agrees, because they don’t.

Larsen disagrees with Larsen. One of his videos prescribes working until “the only time you’re happy is when you are doing this,” five one-hour sessions a day, six days a week. Another says burnout comes from a lack of progress, not from hours. Those don’t fully fit together, and I’d rather name that than pick whichever one is convenient.

More tracking isn’t always the fix. Larsen’s input tracker is excellent. But my own experience says tracking and reporting can grow into their own attention tax, big enough to need a limit. The cure for a scattered mind usually isn’t a better dashboard.

“Always focus on one thing” is not what Saraev says. In another video he tells a newer freelancer to say yes to almost everything early, because experience compounds, and to narrow down only once revenue is real, because taking everything later makes you plateau. Sequence matters. Focus at the wrong stage can be its own mistake.

My Focus System, As It Actually Runs

This is what survived seven months of testing it on myself:

1. Make the first move in the same session. If it matters, do one real piece of it now. The 100% vs 5% number is the whole reason.
2. Every parked item gets a date or a kill note. No date, no kill, no move means it’s already decided against.
3. Third mention means stall. When I catch myself saying “I need to” a third time, I do it that day or delete it.
4. Revenue before new builds. If money is waiting on me, nothing new starts.
5. Rate-limit every input, from news to notifications to the AI tools that report to me. Twice a day is plenty for most of them.
6. Let a machine count the loops. I’m bad at noticing when I’m stuck, so the counting is automatic now.
7. Decide at 70%. The market finishes teaching you for free.
8. Track inputs daily, judge progress weekly. Daily outputs lie. Weekly progress tells the truth.
9. Do it by hand once before you delegate it, to a person or to an AI.
10. Say no more than yes. Every unnecessary yes is tomorrow’s focus, spent today.

Saraev’s line about structure is the one I keep coming back to. Don’t ask how to be more disciplined. Ask how to structure your life so the thing you want to do is the thing that happens. For me, a lot of that structure lives in software now. The discipline isn’t that I’m better at resisting. It’s that I built things that resist for me.

I appreciate you reading this one. It’s the post I most needed to write for myself. If you’re a service business owner being pulled in ten directions and you want systems doing some of the resisting for you, book a call from the home page, or start with the free tools and see where the leaks are.

Build first, learn fast.