Ultralearning Author Revisits His Method in the Age of AI

Scott Young has spent since 2019 building an audience around the idea that intensive, self-directed learning—what he calls ultralearning—can substitute for or complement formal education. His book sold hundreds of thousands of copies and spawned an entire subgenre of personal productivity literature. Now, in an article published in late April on his blog, Young himself outlines what he would change about the method if he were writing it today, with AI as a new variable.

It's an unusual exercise: an author publicly deconstructing parts of his own conceptual framework. The fact that he does so methodically and without dramatics makes it worth pausing to consider his arguments.

What Changes According to Young

The article, which appeared on Hacker News though with modest initial traction, identifies several pillars of the original method that AI calls into question or at least reframes:

The value of direct memorization. Young argued in his book that memorizing concrete facts was part of the process: without a factual foundation, there is no deep understanding. With language models capable of retrieving information immediately and contextually, the question is no longer whether to memorize but what deserves to be memorized. His updated position suggests memory remains valuable, but the selection criteria shifts: it matters more to internalize reasoning structures than individual data points.

The principle of directness. One of the original ultralearning pillars is to practice something as close as possible to real application—if you want to write, write; if you want to code, code. Young now refines this by noting that AI can serve as a high-fidelity practice environment in domains where access was previously expensive or slow. A code learner can get instant, granular feedback without waiting for a mentor. This accelerates the deliberate practice cycle, but also introduces the risk of becoming dependent on external scaffolding without consolidating your own mental model.

Feedback as a bottleneck. The original method recognized that obtaining quality feedback was one of the biggest obstacles to self-directed learning. Young acknowledges this is where AI changes the rules most dramatically: the cost of feedback has plummeted. The consequence isn't that learning becomes abstractly easier, but rather the bottleneck shifts: it's no longer access to feedback, but the ability to distinguish between useful feedback and noise.

Why This Revision Matters

Self-taught learning has a structural tension that the article touches on in passing but deserves explicit naming: most personal productivity frameworks are written at a particular moment and age with the technological context around them. Getting Things Done, for example, assumes a certain friction in information management that barely exists today.

That Young explicitly updates his premises rather than ignoring the shift in context is methodologically honest. He doesn't rewrite the book—or discredit it—but rather points out where conditions have changed enough that the original recommendations need qualification.

Who finds this article useful: primarily readers already familiar with the method seeking an updated roadmap. It's also a reasonably worthwhile read for anyone designing training programs who needs to reconsider where to focus when access to information and feedback is no longer the central problem.

What the Article Doesn't Resolve

There's a question the article touches on but doesn't fully answer: if AI makes instrumental learning more efficient—acquiring concrete skills to solve concrete tasks—what happens to learning as a process of developing your own judgment? Young mentions the risk of cognitive dependence, but doesn't develop how ultralearning should protect against it. It's probably the missing chapter.

---

At ClaudeWave we remain somewhat skeptical of productivity frameworks that update every time a new tool emerges. But Young's ordered self-critique is more useful than most articles proclaiming that AI "changes everything" without specifying what or for whom.