Sunday night, ten minutes
Minute 1-2: Brain dump.
Open Claude. Type everything rattling around: what I finished, what I dropped, what I'm worried about, what's due at the day job this week. No order. Just get it out of my head.
Minute 3-4: The honest calendar.
Tell it the truth about the week. "Tuesday is shot, late meeting. Thursday I have a full hour. Weekend is family." AI cannot plan a week it doesn't know about.
Minute 5-7: The prompt.
"Using everything above, give me ONE goal for the week for my side business, then a plan that fits ONLY the time I actually listed. One task per available day. Smallest task on the worst day. Tell me what to skip."
Minute 8-9: Push back.
Read it. If it's too ambitious, say so. "Cut it in half." It will. The plan should feel almost too easy. Easy plans get done.
Minute 10: Park it.
Copy the plan into a note on my phone. Monday morning I open the note, not my brain.
That's it. No app, no system, no color coding. Ten minutes on Sunday buys me a week of not deciding what to do next, and deciding was always the part that stalled me.
Save this for tonight or next Sunday. Then come tell me what your one goal was.
Before work, four minutes
Copy everything between the lines.
Why it works: it makes AI fit the plan to my real day instead of a fantasy day. On a bad day it hands me a 15-minute task. On a good day it hands me a bigger one. Either way, I start.
The "one thing I should NOT do" line is the secret. It's saved me from more shiny-object detours than anything else.
Save this and try it tomorrow morning. Then tell me what it handed you.
The session itself
Here's the whole method. It's four steps and it fits in half an hour.
1. Pick ONE task. Not a project, a task. "Write the welcome email," not "work on onboarding." If you can't finish it tonight, it's too big. Cut it smaller.
2. Hand AI the boring half. The first draft, the formatting, the research, the "what am I forgetting" list. You keep the judgment, it takes the typing.
3. Set a timer for 30 and put the phone in another room. Not for discipline points. Because the task will expand to fill whatever you give it, and you're giving it thirty.
4. Finish it. Then actually stop. This is the one everybody skips. Stopping while it still feels good is what makes you willing to come back tomorrow. Grinding until you're empty is why people quit in week three.
The thing nobody tells you: thirty minutes a day is about 180 hours a year. That is more than four full work weeks. Nobody builds anything in a weekend sprint. They build it in Tuesdays.
The note that feeds all of it
This is the "Build After Bedtime" series. Day 1, and it's deliberately the smallest possible thing.
Most guys use their first free hour to buy a course or research a niche. Both feel like progress. Neither produces anything, and by night three the momentum is gone.
Do this instead. One hour, and you're done.
Set up a capture file.
That's it. One note, one doc, one app, whatever you'll actually open. It has four headings:
- ASKED ME: things people asked me for help with today - ANNOYED ME: anything clunky, slow, or stupid that I had to do - LIKED IT: a post, product, or email that made me stop scrolling - IDEAS: raw, unfiltered, no judging
Then for the next two weeks, you add one line a day. Sixty seconds, usually on your phone, usually in a parking lot.
Why this and not something that feels more productive:
You cannot pick a business from an empty head at 9pm on a Tuesday. Every guy who tries ends up choosing whatever he saw on YouTube that week, which is why so many of us start three things and finish none.
But two weeks of that file and you're not guessing anymore. You have your own data. The same problem shows up four times. The same question gets asked by three different people. That repetition IS the business, and you can't see it in a single sitting.
Then AI becomes genuinely useful, because you can paste the whole file in and say: "here are two weeks of notes from my actual life, what's the pattern and what could I build from it?" That's a real answer. "Give me business ideas" is not.
One hour tonight. Four headings. Then one line a day.
Day 2 is what to do with the file once it has something in it.