I am a father, and there are tons of things I want to be better at. Getting better requires self-reflection, and self-reflection requires tools. Some people prefer to think silently, some record voice memos, some need physical notebooks, and some use phones or computers to capture their thoughts.

None of these approaches is right or wrong; it's personal. After trying many ways, I found a routine that works—at least for now.

I like having notes, even if I don't always enjoy writing them, and I like structure. For years I used Notion to store my information and organize it with checkboxes, links, reminders, calendar views, and tags. I could easily check what I was thinking on the exact same day a year earlier, or get a reminder from the past to see how I was progressing.

Now we have AI. It's a tool that is genuinely good at spotting patterns. It hallucinates and makes mistakes, but it works if you know how to direct it. In a recent post (Further thoughts on how to use your knowledge with AI), I shared how I use AI with data. Here, I want to focus on a more personal workflow: using AI for self-reflection and self-improvement.

When I write down what's on my mind, it's a messy mix of thoughts, emotions, and in-the-moment feelings. By tomorrow, I might feel completely different. That's normal—life moves, and so do we.

For me, separating feelings from facts is essential. I'm an IT person after all. I need logic, facts, and clarity to make decisions and stay present.

AI can help here, up to a point. I tried using AI as a personal psychologist. It worked poorly because it isn't human. It's a machine—essentially a probabilistic magic eight-ball that predicts the next likely token. It doesn't actually understand anything. But it excels at connecting pieces and finding relationships across messy data.

I store my notes in Obsidian. I used to type them out, but eventually switched to voice dictation to get raw thoughts down without friction. The dictation itself is what helps me think; the raw transcript isn't meant for me to re-read cover-to-cover later.

AI, on the other hand, is the reader. I first had AI build a psychological baseline of me using years of exported Notion notes. The resulting profile was surprisingly accurate.

From there, my ongoing routine is a two-step session.

Step one: I set aside an hour on a quiet day (say, Tuesday), close the door and windows for privacy, and dictate whatever is on my mind without filtering. Things I did, things I care about, worries, loose to-dos. Everything. Every so often, I pause to add links: if I remember touching a topic a few days earlier, I search for that note and link it.

Step two: I ask AI to analyze the day's note—either the entire file or a specific section—against my baseline profile. This is where it shines: it strips away the immediate emotional charge and isolates the underlying facts, giving me an objective view of where things stand.

You still have to make your own decisions and take responsibility for any changes. But it saves huge amounts of mental energy, highlights recurring patterns, and gives you an external perspective to test your assumptions against.

To run this without constantly switching apps, I built Sidecar AI, an Obsidian plugin that lets you query AI models directly inside your notes.

This isn't the only way to reflect, nor is it a universal formula. But as a parent trying to stay intentional, it's the system that works for me.