Grab the 9 prompts I use to turn docs into practice reps + a 90-day pilot you can steal

Your ten years of experience might be one year repeated ten times.

We tend to repeat the same habits, and that limits our career growth.

But what if AI could change that?

In 2019, David Perell wrote an essay called “Learn Like an Athlete.” His observation was sharp: athletes train, musicians train, performers train—but knowledge workers don’t. We just show up and do the work, hoping we get better through exposure.

Perell’s original thesis focused on “learning plans”—the idea that knowledge workers should structure their skill development the way athletes structure off-season training. Set clear objectives, break them into daily increments, measure progress. He used LeBron James as his central example: a player who treats his off-season with the same rigor he brings to games, systematically building specific capabilities rather than just playing more basketball.

That essay circulated widely, and Tyler Cowen picked it up on Marginal Revolution with a characteristically pointed addition. He didn’t disagree with Perell, but he sharpened the challenge into a question I haven’t been able to shake: What is it you do to train that is comparable to a pianist practicing scales?

The shift from “learning plan” to “scales” matters more than it might seem. A learning plan is still vague—you could satisfy it by reading books, attending conferences, collecting experiences. Scales are brutally specific. A pianist doesn’t just “work on their technique.” They practice a defined sequence, get immediate feedback from the sound, and repeat until the pattern becomes automatic. The skill is isolated, measured, drilled.

Most of us knowledge workers don’t have an answer to Cowen’s question. We read. We react. We accumulate years of experience. But experience isn’t the same as practice. Practice means isolating a skill, getting feedback, and repeating until patterns become automatic. Knowledge workers have never had a way to do that.

We didn’t until 2025. That’s changed now. AI makes useful feedback cheap enough to actually practice. For the first time, you can define what “good” looks like for fuzzy work like a decision doc or a spec, get consistent critique on every attempt, and see patterns in your weaknesses over time. You can get incredibly personalized feedback on everything you write, if you simply setup the system.

That’s what this post is for.

This piece makes the case for deliberate practice in knowledge work—where the athlete analogy holds, where it breaks, and how to build something like a practice gym for your team. It’s written for the skeptical operator who’s tired of theory and wants something concrete to try.

Here’s what’s inside:

• The fundamental question. Should knowledge workers even try to train like athletes, or does the analogy collapse on contact with reality?

• What the research actually says. Ericsson’s findings weren’t just about physical skills—plus the counterarguments you should know about.

• Where the analogy breaks hard. The honest limits of drilling your way to better thinking.

• Five skills that keep showing up. Judgment, orchestration, coordination, taste, and updating—and one popular “skill” I’d argue against.

• What a practice gym looks like. A concrete approach you can pilot with one team over 90 days.

• The emotional reality. What this looks like from inside the person doing the practicing.

• The dark version. How this goes wrong, so you can avoid building something people hate.

Plus nine prompts you can use immediately:

• Decision Document Evaluator. Score decision docs against a rubric with quoted evidence and suggested improvements.

• Spec/Orchestration Evaluator. Evaluate specs, PRDs, and process docs that turn strategy into executable work.

• Executive Update Evaluator. Critique communications to busy decision-makers against a focused rubric.

• Rubric Builder. Facilitate the really messy human conversation about “what good looks like” for any artifact your team produces.

• Example Annotator. Mark up real documents with explicit scoring rationale to calibrate your rubrics.

• Messy-to-Drill Converter. Turn real situations (Slack threads, vague requests) into focused 15-minute practice exercises.

• Stronger Version Generator. See what a better version looks like and understand specifically what you missed.

• Weakness Pattern Analyzer. Identify systematic gaps across multiple evaluated documents.

• Interview Exercise Generator. Create artifact-based hiring exercises aligned with how your team actually works.

Let’s go figure out how to actually level up our skills.