Using AI to analyze life patterns

Link post

If you’re anything like me, you may have a lot of notes from various introspective activities – annual / monthly reviews, worksheets, therapy notes, etc. Often these notes are just gathering dust in a paper folder or some forgotten corner of Google Drive. There may be some insights there – how personal problems change over time, what things you tried and what worked and didn’t, and so on. One use case for capable AI models is to sift through this data and extract patterns, insights and interesting tidbits that you may have forgotten or not noticed.

Here are some things I got AI to do for this analysis:

Transcribe handwritten worksheets and diagrams (I did have to take photos of the sheets – this was the most laborious part). The transcription was occasionally wrong and needed correction, but overall saved a lot of time.
Summarize my answers to “Hamming questions” over the years. This was originally a CFAR worksheet named after Richard Hamming’s question “What are the most important problems in your field?” but applied to your life.
In the past I have found it useful to make a visual representation of my problems as a graph showing how they contribute to each other. I had only a few of these from past years, so I asked AI to draw diagrams for each year based on my notes and worksheets.
Compare worksheets and diagrams from different years, identify persistent feedback loops, look how loops and key nodes change over time.
Identify interventions that worked in the past that I could try again (therapy and coaching notes are particularly good to mine for this).
Do deep dives to analyze specific patterns.

Claude Opus 5 and Fable did a pretty good job at this. Fable had a tendency to over-compress its findings, and I had to ask it to use simpler language. I also tried this 1-2 years ago, and I think the models are getting better at picking up non-obvious patterns.

Learnings about problems and patterns

Here are some patterns I found in my answers to Hamming questions over the past 10 years:

Rate-limiting step / bottleneck: sleep, tiredness, focus, anxiety, and confidence were recurring themes over the past 10 years. On the one hand I made progress on these issues, and on the other hand they appeared from new sources, so overall there is some improvement but they are still there.
Things I don’t feel “allowed to care about”: earlier years listed various things like having kids and mentoring others, which I eventually ended up doing. Creative pursuits also appeared on this list a lot (e.g. drawing or writing poetry) and I still don’t do much of this.
Things to do to “advance the plot” in my life: more regular meditation, try different kinds of therapy, getting help / advice sooner, more delegation / babysitting.
Things I tend to do as a poor substitute for what I really need (“pica”): general theme of procrastination and being reactive (scrolling social media, gravitating to logistical tasks over more open-ended tasks or hard decisions, joining projects run by others). Most of these have improved with better reflection and prioritization systems, but the core issues remain.
“My life is fine except X”: what came up as X has generally shifted from mostly mental problems (anxiety / overwhelm) to external / object level things (kids waking up at night, uncertain AI future etc).

One thing I learned from this is that some of my persistent problems appeared in different guises or with different causes. These days I mostly attribute tiredness, lack of sleep and feeling overwhelmed to the demands of parenting, but looking back at my notes from 2015-17 this was just as much of an issue (if not more) before I had kids. There are internal and external factors affecting sleep, and the external factors have increased after having kids, but the internal factors are pretty stable and the same habits are still useful (e.g. put away screens and wind down after 10pm).

Overwhelm was a recurring feeling triggered by different things (finishing my thesis, balancing research and outreach, taking care of a baby, etc) which seemed like a mental pattern that attached itself to life circumstances rather than a fact about those circumstances. I had a default response to having a lot of stuff going on with a sense of shame about not doing everything or comparing myself to others, and this made me feel overwhelmed (while the same circumstances did not feel as overwhelming when I gave myself permission to drop things).

There were several persistent feedback loops between problems (which became weaker over time).

The judgment loop: underperform → judge myself → feel anxious → perform worse. This comes up most clearly at work, but there is also a parenting version: feel anxious → be triggered / impatient with kids → shame / guilt → anxiety.
The pressure loop: feel behind → do lots of stuff compulsively → focus on small things and create more pressure → feel more behind. This has gotten better with explicit prioritization structures, but the compulsiveness pattern is still there, particularly when working on the computer (e.g. sometimes I find myself unnecessarily switching tabs).
The disconnection loop: feel stressed → ignore internal signals → lose touch with what I want → more stress. This seems mostly fixed and didn’t show up in recent years.

Comparing my bottleneck diagram from 2015 with one from 2025, there are some nodes in common (“should” mindset, confidence, anxiety) and some things that did not persist (neglecting internal signals, working memory, excessive effort, insomnia), which mostly map to solved problems. There are some new kid-related nodes though (“stress with kids” and “reduced alone time”).

Learnings about approaches and solutions

There were various things that worked in the past that would be good to try again or do more often:

Reflection periods in a different physical location (library, park, coffee shop, etc).
Aiming for a crappy draft (to counteract writers block).
Noticing when I’m stuck with something and asking for help sooner.
Aversion factoring: noticing if I’m avoiding doing something I would like to do, writing down what is good about doing it, what feels bad / aversive about it, and any counterpoints to those objections. I often ended up doing the thing after doing some aversion factoring on it. (Aversion factoring is traditionally taught with diagrams, but I found it easier to do in list form.)
Using anger / annoyance as an unblocker: noticing when I’m feeling stuck or tense and writing down what I feel annoyed about, which helps me to process emotions in the background and get into a more agentic mindset.

There were also things that I tried several times that didn’t work:

Putting time blocks on the calendar for things I intend to do, without supporting systems to make it happen. For example, reading blocks on my work calendar often got ignored in favour of project work that felt more urgent, and “alone time” calendar blocks on weekends kept not happening. If there are mental or situational obstacles to doing the thing, a calendar reminder is not sufficient.
Closing comms channels (email / slack) to avoid distraction, e.g. during focus blocks. This didn’t alleviate the emotional reasons for getting distracted, and I often ended up going back to the comms channel to get some information I needed. What did work was to do focused work entirely on paper without a computer (reflecting in a notebook, reading a printed paper, etc). Deleting social media apps was also ineffective because I can open them in the browser (while using an app that prompts me to be intentional when I unlock the phone worked better).

I identified some meta-principles on things that tend to work for me:

Without chain of thought, I am not a general intelligence. I need to write everything down: todo items (before doing them), my mental state before starting work, thoughts about research directions or papers I read, loopy / obsessive thoughts, decisions, factoring aversions, etc.
Habits that stick are usually attached to an existing routine: meditation after breakfast at work, leg stretches while using mouthwash before bed, doing regular exercise classes. One takeaway is that if I want creative pursuits to happen, I should sign up for a class in drawing or improv or something, and then I would actually do it.
Habits that rely on willpower don’t work. It’s good to look for physical solutions to problems (medications / supplements, nasal spray for colds, weighted blankets, focused work in a different place without electronics, etc). Habits that necessarily have a mental component (like avoiding sugar / snacks) only work if I get into a mental frame where I just don’t do this thing, rather than a willpower-sapping frame of trying to do less of it.

Overall, the benefits I got from doing this kind of analysis can be broken down in two ways: looking at problems or approaches, and whether what I did was successful or not. I think each of these categories can be worth a deep dive with some AI help.

	Successful	Unsuccessful
Problems	Noticing problems that I successfully fixed or improved and appreciating that.	Noticing problems that have recurred despite various interventions, and reflecting on why.
Approaches	Being reminded of things that worked before that I forgot about.	Identifying things that repeatedly didn’t work in the past and probably need a different approach.

Appendix: prompts I used for this analysis

I did this in two separate conversations, since the amount of context was too much for one conversation.

Conversation 1: analyzing problems and patterns based on life reviews

Inputs

Photos of Hamming worksheets from a range of years (2015-2026)
Doc with annual / quarterly reviews for a range of years (2019-2026)
Doc listing significant events in my life during those years (optional but seemed helpful)

Stage 1 — transcription

I will scan some documents, please transcribe and put in a doc. Flag particularly illegible words as [unclear] rather than guessing.

(Then uploading photos in batches, one worksheet at a time.)

[I would recommend reviewing the transcriptions to correct things the AI failed to parse before moving on.]

Stage 2 — identifying patterns in worksheets

Look over these worksheets over the years and identify key patterns.

Can you make some tables of how these patterns show up over time? (rows = years)

For each question on the worksheet, summarize the answers to that question each year, and put them in a table.

I find your summaries a bit hard to follow, they are dense and overcomplicated. Focus on key insights and explain them in simpler language.

Make tables with verbatim answers to each question from each year (not summarized), and put into a doc.

Stage 3 — adding other sources

Here is a file with notes from my annual / quarterly reviews. Look over these, summarize different exercises / reflections and how they evolve year by year. Connect with the worksheets in that year.

I added a doc with a timeline of significant events in my life. Please analyze how these events relate to my answers over time.

Can you regenerate the per-question tables with some commentary on connection to these events?

Stage 4 — bottleneck diagrams

I also have these diagrams of the relationships between my problems over the years. Think about patterns in how the problem dynamics changed, connecting this to patterns in the worksheets and timeline. Explain in simple language.

Please redraw them digitally, like the first diagram. Also make a table with rows for years, columns for nodes, recording how many in and out edges the node has.

Some years I filled out the sheet but didn’t make a diagram. Can you make a diagram based on sheets and annual review for each year? Make these with weighted edges (light / medium / heavy for how strong the relationship is between those nodes). Then give me an updated analysis for how the loops and key nodes evolved over time.

Stage 5 — deep dives

Go into [specific findings] in more detail.

How do the [X] and [Y] loops relate / connect to each other? Did this change over time?

Regarding levers, were there specific ideas / techniques / intentions that came up in these reviews that I may have dropped or forgotten? E.g. things that worked in the past.

I want to understand the [X] loop better. I can see how [mechanism A] will increase [B], but why does this result in [C]?

Conversation 2: learning about what approaches worked and didn’t based on therapy / coaching notes

Inputs

Doc with therapy notes from a range of years (2014-2025)
Doc with coaching notes from a range of years (2022-2026)
Doc with notes from accountability buddy meetings (2024-2026)

Prompts

Review my therapy and coaching notes, and identify patterns in suggestions and ideas, especially things that have worked before that I may want to try again.

Here is a meta principle I identified from these insights about what kind of things work for me: habits that stick are usually attached to an existing routine. Any others you would suggest?

For the principles and ideas that worked before but lapsed, identify the difficulties I had implementing these, and how I could make it more likely to work in future.

Zoom into ‘trying the physical fix first’. Give me an extensive list of physical fixes I tried or considered.

Make a list of stuff I tried that didn’t work, especially if it came up several times.