I use books to program and deprogram myself

My Goodreads read shelf contains 743 unique titles marked read. My Readwise archive contains 327 book files and 5,690 highlights. The public, sanitized library on this site contains 318 pages.

Those numbers describe different things. They should never have been collapsed into one claim.

Dataset	Count	What it supports	Limitation
Goodreads read shelf	743 unique titles	I marked these books as read	Self-reported; not independent proof of completion
Readwise archive	327 files; 5,690 highlights	I engaged with or imported passages from these books	Incomplete subset with four exact duplicate-title pairs
Separately dated notes	29 books	A dated note independently mentions the book or author	Very conservative and incomplete
Public library	318 pages	Books safe and useful to publish here	Deduplicated and filtered through a publication denylist

The Goodreads shelf is the correct denominator for questions about the full reading record. Readwise is the stronger signal for engagement with specific passages, but it covers only 44.0% of the 743-title shelf before accounting for duplicates. A highlight can mean agreement, disagreement, surprise, or simply “come back to this.” None of these datasets proves that I finished or endorsed every book.

Did I really read 34 books about relationships and sex?

The exact 34 I previously published was not defensible. It came from an automated classification of 34 Readwise files, not the full 743-title Goodreads shelf.

After auditing all 743 Goodreads titles under a stricter definition—books primarily about romantic relationships, attachment, dating, mating, or sexuality—the current count is:

• 35 of 743 titles, or 4.7% of the read shelf;
• 30 of the 35 have an explicit Goodreads read date;
• 26 of the 35 have a Goodreads rating;
• only 13 are explicitly placed on my custom love or sex shelves, showing that those tags are incomplete rather than a usable taxonomy.

The original Readwise number also hid two errors. Its 34 files represented only 30 unique titles because four titles were duplicated. Six of those 30 were broad false positives such as books about mathematical love, professional relationships, or difficult conversations. Under the same strict definition, the Readwise subset contains 24 unique relationship or sexuality titles. Thirteen have at least ten highlights and seven have at least thirty.

So 34 and 35 are numerically close by coincidence. The better claim is: 35 of the 743 books I marked read—4.7%—are clearly centered on relationships, attachment, dating, mating, or sexuality.

Show the audited 35-title set

1. A General Theory of Love
2. Aphorisms on Love and Hate
3. Attached
4. Attachment Disturbances in Adults
5. Codependency For Dummies
6. Come as You Are
7. Eight Dates
8. F*ck Like a Goddess
9. How to Not Die Alone
10. Love Sense
11. Mating in Captivity
12. Models
13. No More Mr. Nice Guy
14. On Love and Loneliness
15. Out of the Fog
16. Polysecure
17. She Comes First
18. Superflirt
19. The 5 Love Languages
20. The All-or-Nothing Marriage
21. The Art of Seduction
22. The Course of Love
23. The Evolution of Desire
24. The Five Love Languages Singles Edition
25. The Flirting Bible
26. The Game
27. The Mathematics of Love
28. The Mating Mind
29. The Psychology of Romantic Love
30. The Red Queen
31. The Science of Happily Ever After
32. The Seven Principles for Making Marriage Work
33. The Way of the Superior Man
34. Why Is Sex Fun?
35. Women's Anatomy of Arousal

What the Readwise subset suggests

I gave each of the 327 Readwise files one automated primary-domain vote. Highlight count measures marking depth, not importance. This table describes the highlighted subset, not all 743 Goodreads titles.

I expected the highlighted archive of a technical founder to be dominated by software, AI, and business.

It is not.

Domain	Books
Relationships, attachment, and sexuality	34
Entrepreneurship, product, and strategy	33
Epistemology and rationality	26
History and biography	24
Civilization, politics, and conflict	20
Wealth, finance, and power	20
Mathematics and physics	17
Sales, marketing, and persuasion	17
Spirituality and contemplation	17
Consciousness and direct observation	16
Biology and evolution	15
Leadership and organizations	14
Meaning, mortality, and existence	13
Self-knowledge and psychology	10
AI and computation	9
Learning, memory, and knowledge	9
Resilience, courage, and adventure	9
Software, systems, and reliability	9
Creativity, writing, and design	8
Health, longevity, and embodiment	7

The surprising result is not that I read widely. It is that relationships and human connection form the largest automated primary-domain bucket in the Readwise subset, while AI and software account for only 18 primary files combined. This does not establish the same ranking across all 743 titles.

The highlighted books look less like a technical curriculum than a long attempt to understand reality, other people, power, and myself.

I rarely read only to know

In one of my notes, I explicitly grouped books by what I wanted them to change in me.

I used Napoleon Hill to increase desire and motivation. Brian Tracy and Naval were instruments for commercial agency. Biographies of Rockefeller, Rothschild, and Morgan offered models of capital, power, and long time horizons. Paul Graham offered startup judgment.

Other clusters served different interventions:

• Popper, Yudkowsky, Pearl, Taleb, and Galef trained belief correction.
• Feynman, mathematics, physics, biology, and evolution kept my worldview below the level of startup fashion.
• Gottman, Perel, Buss, Johnson, Manson, and attachment literature trained attention toward human motives and relationships.
• Krishnamurti, Taoism, Goenka, Stoicism, and mortality writing challenged the assumption that more achievement or control would create freedom.

I was not building one consistent worldview. I was collecting programs that could modify different parts of me.

Accelerators and brakes

The reading record repeatedly contains an idea and its antidote.

Accelerator	Brake or corrective
Wealth, power, biographies, sales	Mortality, meaning, attachment, ethics
Rational models and prediction	Direct observation and anti-authority traditions
Productivity and optimization	Presence, love, embodiment, rest
AI and automation	Human agency, privacy, relational complexity
Independence and self-reliance	Belonging, interdependence, gratitude

Business books increase agency, but books about mortality question what the agency is for. Rationality improves models, while contemplative traditions warn that the model can become another escape from reality. Relationship books teach concepts, but an actual relationship immediately reveals how little the concepts guarantee.

The most useful books did not simply add a belief. They destabilized a belief that had become too dominant.

My reading algorithm

The archive and my notes suggest three rules:

1. Wide funnel, tight filter

I start many books and stop quickly when the fit is poor. Finishing is not the objective. Finding a model that changes perception or action is.

2. Prefer compression

I value short books, low filler, and stories capable of carrying dense models. This is one reason biographies matter: they compress decades of decisions and consequences into something a human mind can simulate.

3. Read for application, not reverence

A book is useful when it changes behavior, improves a decision, gives language to an experience, or reveals that one of my current models is wrong. Saving highlights is not the final step. Reality has to answer back.

What books could not teach me

Reading can become sophisticated avoidance.

You can read about courage instead of taking a risk, communication instead of having a difficult conversation, startups instead of talking to customers, or meditation instead of observing your own behavior.

One of the strongest corrections in my dated notes was that making mistakes through real interaction can teach faster than endlessly simulating the interaction internally.

Books prepare models. Experience supplies error signals.

This is why I now think of reading as both programming and deprogramming. Some books install ambition, agency, and technique. Others remove certainty, ego, and the need to control. Neither operation is sufficient by itself.

The complete public, sanitized archive currently contains 318 book pages. You can browse the library here. It is a deliberately filtered publication surface, not the full 743-title Goodreads shelf and not a completion ledger.