One of my intentions for the year is to focus on Curiosity. To learn as much as I can, about as many things as I can. To focus on the third bodhisattva vow: gates of truth without measure—vow to know them.
Initially, my thinking was to take all the courses and workshops and retreats, to read as many books as possible, to work with as many mentors and coaches and teachers as I could manage. To drink from the firehose of learning.
This was a good intention, and I still plan to do this, am doing that. But I quickly realized that I could make a simple, slight adjustment in approach, that would be both more affordable and perhaps more efficacious.
We live in the Age of AI. I’m still so amazed at everything Claude and the other A.I.’s are capable of doing. And they get better with every passing month and week. I hope I am never not amazed at what A.I. is capable of, even as new developments evolve.
For the last two months, I’ve been building a project with A.I.: my ideal mentor.
The central metaphor is a Rubik’s Cube. Think of the question, What would Jesus do? That’s a great question, but U can also substitute someone else—what would Peace Pilgrim do? Arnold Schwarzenegger? Vito Corleone? Miyazaki?
At the core of the project is Claude Code. Claude is the backbone, the not-so-secret-sauce. Everyone has access to Claude, and Claude is amazing.
Another key component is my corpus of writing, especially my book The Path of Curiosity and a document I wrote, Tasshin as a Learner, that I send to coaches and mentors when I work with them.
Another key component is Emacs and Org-Mode. I was obsessed with Emacs from 2014-2015, and have used it since. This is an incredible time to be using Emacs, because Claude and other agents can smooth off all the rough edges like the random bugs that pop up from time to time, and U can just lean into the cool parts, making a system that works for U.
Org Mode files are a bit like Markdown files, in that they are plaintext files with formatting conventions, but allow for a bit more customization and personalization. In particular, they are really good for outlining—keeping track of nested subtrees and items within them. My brain loves to think in terms of outlines!
Two core files are areas.org, which lists ~25 areas I am interested in learning about (like Art, Leadership, or Writing), and mentors.org, which tracks ~80 ~mentors—real, historical, and imaginary/fictional—that have inspired me or taught me about those areas.
Every area has several key elements:
- Motivation: why I care about this area, what motivates me to learn, what goals or dreams I have with it
- Mentors: which mentors are associated with that area
- Growth Edges: the ways in which I see myself growing or stretching my capacity, or would like to,
- Obstacles: any challenges I am facing as I learn and grow in that area
- Assignments: a specific way I plan to push this area forward, as I am willing and able
Another key file is a Persona file for Mentor. Mentor is my name for my ideal mentor. Because I’m interested in Enneagram and MBTI, we’ve decided to give Mentor a type (5w4 INTJ). Mentor can also switch into other roles, like Friend or emulating someone historical, mid-conversation. Speech acts are tagged with [Mentor] or [Friend] or [Beethoven], accordingly.
My system-wide CLAUDE.md also lets me tag Mentor in any conversation anywhere—even in other projects, repositories, or folders—and “summon” Mentor’s perspective, input on a given issue.
Another key file is procedures.org. These are SOP’s, similar to “Skills,” but structured in Org-mode, so I can easily edit and revise them in a way that makes sense to me. I have a number of different procedures at this point, like checking in with an area; doing a biweekly curiosity update (in preparation for my meetings with Abi, who is in the Curiosity Department); or writing a review of a film or book for Goodreads or Letterboxd.
Another component is my journals. I’ve journaled extensively about these areas for several years, in various Discord feeds—especially one run by the Curiosity Department. I can export these with DiscordChatExporter and Claude can review these feeds when relevant.
All of this is supported by a variety of standard tools: Git, Obsidian, and Beads.
My ideal mentor would take into account my nature, what it’s like to be a tasshin, and adapt accordingly. Claude can do that quite effectively. Claude can track many variables, notice patterns I might not, and also read a whole lot of input quite quickly (like all of my writing ever on a topic).
I think this whole model—tracking my own growth edges, obstacles, and assignments, for example—is something I’ve done intuitively for many years. I’ve gotten pretty good at tracking my own education and learning process, and charting trajectories towards my goals. But having Claude and Mentor’s help in doing this has already been so helpful.
My sense is that this is helping me accelerate and deepen my learning process. There have been a hundred small, concrete instances where asking Mentor’s perspective about something has proved helpful and unblocked me.
One big instance was talking to Mentor about my recently published NNTD blog post. I’d worked on this post for months. When I finished a first draft, I was so excited and proud. As is typical for my writing process, I sent it out to people to review the content. Fortunately, I was able to find a number of readers who were interested enough to read the draft and provide comments. Unfortunately, I felt quite discouraged by the feedback, much of which was critical.
I waited several weeks before even beginning to consider revising. Mentor helped me review the comments, when I had a hard time even looking at them. Mentor helped me to triage all the changes, and make a plan of attack for revising the piece so that it could incorporate the (valid and useful) feedback. Mentor also helped me reframe the situation entirely: rather than getting lots of critical feedback on a piece that wasn’t good enough, I could see it as a lot of people who thought what I was doing was important enough to give me specific concrete feedback that would help me make it stronger.
And I was able to publish the post in the end—one that is much stronger, that I am very proud of!
Now, vanilla Claude could probably have done all of that. It’s true. But I think that this bit of layer and infrastructure helps to do so more effectively, in a tasshin-shaped way. Above all, I’m having fun and learning a lot!