Thesis

Google turned data and information into a commodity. AI is doing the same thing to reasoning.

It doesn't worry me. The calculator saved us from doing arithmetic by hand, but if you put a sum on paper in front of me, I can still work it out.

What worries me is that AI ends up doing our reasoning for us while we lose the ability to reason without it.

There's only one thing we can do about it: keep learning, keep reading.

Learning is a good long game. It compounds. Steve Jobs said you can only connect the dots looking backward. But once you start connecting them, you realize the superpower you've quietly been building.

But, there is a problem:

Estimated Amazon ebook releases

Monthly releases, 2020-2025

Estimated content authorship share

Human-written vs. AI-generated content, 2020-2025

Principles

1.01^n exists for a reason: to surface all the quality content that's been buried under so much slop.

Good content is evergreen

Good content holds up on re-reads. Paul Graham wrote Taste for Makers in 2002, and in 2026 the word "taste" is more relevant than ever.

Knowledge, truth (if we're even capable of getting anywhere near it) doesn't care about years or decades.

Classic content does not need to be rewritten

Darwin wrote On the Origin of Species in the 19th century. If you want to learn about biology and evolution, read Darwin. Reading the people who invented something gives you a completely different vantage point. You can trace their thinking, follow the steps that led them to their conclusions, and understand the foundations that everything built on top of them is standing on.

A lot of the books we read today are repackaged versions of ideas published in the 19th and 20th centuries, with a few new brushstrokes added. There are exceptions, but the pattern holds: most cognitive psychology books published after Thinking, Fast and Slow are basically extended commentary on Kahneman's original work.

Rule of thumb: if a book has been in print for 200 years, it'll probably still be in print 200 more years. That's the Lindy Effect.

Good content follows power laws

Not everything follows a normal distribution. Some things follow power laws, where a small minority captures most of the value.

Wealth distribution works this way. Most people have very little and a handful have an obscene amount.

Content works the same way. Most of it is low quality or worse, and only a small number of genuinely exceptional pieces manage to float to the surface of the information ocean we're all swimming in.

Systems don't help

Good content tends to be long and needs multiple reads, which is pretty much the opposite of what our reward circuits are looking for.

The real problem isn't the algorithms penalize good content, it's that they've inverted the success metric for ideas. Historically, an idea's staying power depended on whether it survived scrutiny over time. Now it depends on whether it generates sharing behavior within 48 hours.

This means the selection pressure on public ideas has fundamentally changed: not toward truth, not toward nuance, but toward immediate emotional response. Disruptive 19th century ideas polarized expert communities over years; disruptive ideas today polarize mass audiences over hours, before expertise even enters the conversation.

The amplifiers don't just accelerate spread, they change what kind of content is evolutionarily fit.

---

So here I am, trying to dust some of that off.

Books and essays that have helped me think more clearly. Most of them are about software and product, but the ideas travel well beyond that.

You'll want to read some of them more than once. Others won't click at all. That's fine. That's kind of the point.

Read slowly. Let things sink in. Take notes. Some of them have highlights with thoughts of mine attached. Push back on them. This isn't about ego. It's about trying to understand the world a little better so we can make better decisions.

Start here