Cohort 2 of the AI Second Brain is happening: September 17 to October 8.
The Founding Cohort finished with an NPS of 70, and we've spent the time since making it even better.
If you're thinking about joining, save those dates now to your calendar.
And if you want to be the first to know when enrollment opens, join the waitlist at the link below.
Your AI Gives Generic Answers? Here's the 3-Step Fix
Most people give AI almost nothing to work with, then wonder why the answers feel generic.
In this video, I show you the system I use to fix that. I call it Personal Context Management, or PCM. It’s the next generation and evolution of PKM, or Personal Knowledge Management.
It comes down to three layers of context: persistent, project, and perishable. To show you the difference, I build a launch plan for my next book and add one layer at a time. You can watch the same prompt go from generic to useful.
By the end, you'll see how this stops being a technical trick and starts being a way of working with AI.
What I've Learned From Selling 500,000 Books
Sunday, June 14, marked the fourth anniversary of Building a Second Brain being published.
The book took two years to break even. And we’ve spent over a million dollars before a single one even hit a shelf.
This post is the most honest accounting I've given of what writing and selling books at scale entails: the finances, the mistakes, what a book is really for (it's not what you think), and why the most successful moment is also when you have the least control.
If you've ever considered writing a book, or wondered how mine led to anything beyond the books themselves, read this one.
Btw, I’ve placed every blog post I’ve ever written about the book-writing process, totaling more than 33,000 words, into this public notebook on NotebookLM. Ask it any question you have – about the writing process, proposals or acquisitions, marketing or launches – and it will use AI to draw on my lessons learned and give you an answer.
Ask AI to Interview You — Then Push for More Questions
One of the things I do early in any substantive AI session is ask it to interview me before touching the actual work.
The catch is that AI will naturally ask you two or three questions and consider that sufficient. It wants to get started. And two or three questions is almost never enough to surface what it really needs to do the job well.
So I push back. I'll say: ask me ten questions. Ask me fifteen. I want it to pull out my goals, my constraints, the history of the project, assumptions I haven't thought to mention. I'll often use voice mode just to let myself ramble more freely than I would by typing.
The reason this step matters as much as it does: the most current, most nuanced context you have is always in your head. Notes help, but there's always something more sitting just below the surface that you haven't written down yet. The interview is how you get it out before the building starts.
Try it this week: next time AI seems ready to move on after two or three questions, tell it to keep going. Ask me ten. Notice what comes up that you wouldn't have volunteered on your own.
What did it uncover? Hit reply and let me know
Notes on Having a Third Child
A few weeks ago I shared that Ángel had arrived. Since then, I've been writing down what I'm noticing.
Why three is harder than two in one specific way (and easier in another). What it looks like when a family's internal attention economy restructures in real time. And why I think raising kids is becoming a lifestyle niche rather than a default path.
Seven honest observations from the frontier of modern family life.