Has AI Already Eaten the How-To? Writing, Data, and What Comes Next

A publisher friend showed me a chart last month I haven't stopped thinking about. Sales of generic how-to titles — the “Ultimate Guide to X” shelf — are down by a third in three years. Meanwhile, memoir-driven teaching (“how I did X, scars included”) is up by half. The market has already voted: information is free; testimony is precious.

I ran my own numbers across this site and the newsletter. Pure instructional posts (“How to set up Y”) now get a third of the engagement they earned in 2021. But posts with personal data — my sleep logs, my revenue splits, my failed experiments — outperform everything else by multiples. Readers aren't avoiding learning. They're avoiding unowned learning.

Why the generic guide dies first

A how-to is a claim about the world, and machines now make average claims instantly. What they can't do is warrant the claim: show the scar, stake reputation, answer the follow-up question at 11pm. Teaching was never really about information transfer — it was about trust transfer. The wrapper (steps, frameworks) was just the delivery van.

What survives: three durable formats

• The lab notebook: dated entries, real numbers, honest failures. My slow-carb logs still outrank every recipe I ever posted.
• The annotated interview: not the transcript — the marginalia. What I tried, what worked, what I'd skip (see Field Notes from Giants).
• The live teardown: coaching a real person through a real problem, like the reading episode (#246) — unscripted beats comprehensive.

What this means if you teach for a living

Stop competing on completeness — you will lose to the machine on breadth every time. Compete on stakes: put your name, your numbers, and your weird opinions into everything. Charge for transformation (cohorts, feedback, community), and give away the information with your fingerprints all over it.

The how-to isn't dead. The anonymous how-to is. Sign your work — and mean it.