After writing my latest year-in-review, I realized that I had already done eight of them! It feels like yesterday when I started writing them. Given that this body of writing is probably the most personal and raw prose about me on the Internet, I thought why not put it into an AI and start asking questions? So I copied all 8 blog posts (31,000 words in total) into ChatGPT (The “o1” model was the only one that would let me copy that many words in), and started chatting with it.
My first question was “Given the above context about Brian, summarize his personality in one paragraph.” Here’s what I got:
Brian’s personality shines through as a thoughtful, introspective, and growth-oriented individual who consistently balances ambition with humility. He values lifelong learning, embracing challenges like learning to sing, speak Mandarin, and deepen his technical expertise. At work, he’s driven to build effective teams and deliver meaningful results, all while reflecting on how to improve both himself and his organization. His commitment to family, the joy he finds in being a parent, and his awareness of the trade-offs between work and personal life reveal a caring nature that integrates professional success with personal fulfillment. He is analytical yet empathetic, methodically pursuing goals and viewing setbacks as opportunities to refine his understanding of the world.
Not bad! Hits some of the key points that I would say about myself like being a lifelong learner, thoughtful about trade-offs, and analytical. Given what I have written, it’s all pretty obvious, but it was a good synthesis of it. Interestingly, I don’t really characterize myself as empathetic, but maybe I am? I think I often have an awareness of how people might be feeling, which I feel comes more from analytical reasoning than intuition so maybe that’s what the AI picked up on.
Next question: “Tell me something about him that he did not explicitly say.”
He never explicitly mentions it, but he consistently adopts structured, almost systematic ways of tackling every challenge he faces. Whether it’s learning Chinese through daily flashcards, investing with a probability-based framework, or iterating on team processes, he shows a natural inclination toward methodical, data-driven progress rather than relying on flashes of inspiration or ad-hoc efforts. This tendency suggests he gains as much satisfaction from refining the process of learning or building solutions as he does from the outcome itself.
Also pretty good because I make big efforts to be systematic in my way of approaching things. If you read what I wrote, it’s probably not a big insight but still pretty accurate. The part about gaining satisfaction from process of refining instead of the outcome is pretty good too. That I feel is a bit deeper of an insight where you may have had to reflect a bit to get it out of my writing.
I started asking a whole host of random questions, more curious about how the AI would respond rather than actually looking for insight, here are a few interesting ones with some commentary:
- “Tell me 5 things he can improve on”: I got back social engagement, stress management, clear boundaries, delegation and oversight, and leisure and creativity. Some of them are a bit generic, but the one on social engagement was spot on. I think that’s one area I’m not actively prioritizing but probably could have the highest leverage for me.
- “What kind of life partner would best suit Brian?”: It hit on a lot of themes from my writing on someone who appreciates my need for quiet time, supports my values, and someone who will nudge me toward balance. Good advice, even if pretty generic.
- “Use a popular personality test framework and evaluate him based on the context above.”: Using Myers-Briggs, I got INTJ (Introversion, Intuition, Thinking, Judging). I don’t think I’ve ever taken this test but the overall explanation makes sense to me.
- “What investing advice do you think you could give Brian given his style?”: It took some of my lessons learned and thought up some things to help manage it (e.g. investing checklists). It also had some generic advice like diversification. Not great but impressed it got some specifics in there.
- “What’s 5 phrases that can be used to describe Brian.”: These kinds of prompts are great for brainstorming. Most of them are close but off. I usually use it to brainstorm and then mix and match. My favourite one from this response was “analytical yet introspective thinker“.
- “Which philosophy does Brian match best with?”: Stoicism, no surprise there, but I don’t think I’ve written about it explicitly. Pretty good for the AI to understand it.
- “Which books would you recommend him to read?”: All popular books on topics that I write about with suggestions like Meditations by Marcus Aurelius, Principles by Ray Dalio, Deep Work by Cal Newport, Range by David Epstein (haven’t heard of this one), and Atomic Habits by James Clear. Not bad for a superficial list but probably misses the range of things I’m interested in.
- “What about fiction?”: Mostly came back with a bunch of sci-fi books. I’m not a big fiction reader so I’m sure some of them are good, but at the same time I do have more diverse interests. Again, the AI is spitting out the ones that are most associated with someone of my personality without consideration other data. If I gave more context, I’m sure it could give better recommendations.
- “What about 5 TV and 5 movie recommendations”: Again mostly sci-fi / computing. It definitely has bucketed me into a certain segment. It’s based on the writing from my year-in-review, so I can’t fault it too much.
My first conclusion from this exercise is that it is way cool! To be able to even get responses like this probably would require someone knowing me for at least a year. It’s so cool that I can just plop in a bunch of my personal writing and it gets to know me, even if the insights aren’t deep.
The other obvious conclusion is that it is solely basing the responses on the writing that I gave it. And while the answers were generally in the right ballpark, my writings are not sufficient to define me. I’d like to believe I’m more multi-dimensional than the public prose that I put on the internet. The answers I would say were not deeply insightful but were directly reasoned from my writing. It’s not going to replace a therapist, but it’s still useful as a Socratic tool to ask questions about yourself.
One last thing I’ve been thinking about: is the AI just telling me what I want to hear? Is there is a horoscope effect going on? On the one hand, it did give specific references to things I said. On the other hand, the AI is explicitly trained to give human preferable responses. Given how agreeable it is, and how scalable has become, there are a lot of dark use-cases with this technology. However, I’m an optimist and I think the good will outweigh the bad (I explicitly don’t think we’re not even close to the Terminator scenario).
There is so much boring, low-priority work that needs to get done but doesn’t. I want to live in a world where I don’t need to fill in PDF forms, where I don’t have to figure out how to use vlookups in Excel every time I need them, and where my mom doesn’t need to ask for my help with computers. These are modest but each one of these things adds small amounts of friction to the world and make it objectively worse. And while these problems don’t matter that much, there are millions of problems like these that can be solved with AI, it’s just a matter of time. The future is going to be awesome! (Although this doesn’t quite address the source of our biggest problems (i.e., people) but that’s a topic for another day.)