What Ultralearning Actually Is (and Isn't)
Ultralearning: Master Hard Skills, Outsmart the Competition, and Accelerate Your Career came out in 2019. Its author, Scott Young, is a writer who became known for a stunt: he taught himself the curriculum of MIT's undergraduate computer science degree in about a year, working from the free course materials MIT publishes online, and passing the exams on his own. He followed it with a year spent learning four languages by refusing to speak English in each country he visited.
It would be easy to read the book as a highlight reel of someone with unusual time and unusual willpower. That's the trap. The interesting part isn't that Young did extreme things. It's that he went back and asked why they worked, then turned the answer into nine principles you can apply to goals that are much smaller than his.
So let's define the term plainly. Ultralearning is self-directed learning that's unusually intense and aimed at a specific, hard skill. Self-directed means you own the plan, not a school or a course. Intense means you concentrate effort rather than letting it trickle out over years. Specific means you're after a real capability you can name, like "hold a conversation in Japanese" or "build and ship a working web app," not a vague wish to "be smarter."
What ultralearning is not is a promise that everyone can do an MIT degree in a year. The principles are sound and well supported by learning science. The dramatic timelines are not the point, and treating them as the goal is how people set themselves up to quit. The point is the method, and the method scales down beautifully. This article is about running it on a normal goal, with a normal schedule, as a reader and self-teacher.
Start With Metalearning: Map Before You March
The first principle, and the one most people skip, is metalearning. It means learning about the learning before you do any of it. Young's rule of thumb is to spend roughly ten percent of your expected project time up front on research: figuring out what you actually need to learn, and how the people who are already good at it got there.
This sounds like procrastination dressed up as strategy. It isn't. Imagine deciding to learn data analysis and immediately buying a stack of statistics textbooks. Three weeks later you discover the jobs you want barely touch the math in those books and lean almost entirely on a few practical tools. You spent three weeks studying the wrong thing because you marched before you mapped.
A metalearning pass answers three questions. Why are you learning this, which tells you how deep to go. What concepts, facts, and procedures does the skill actually require, which you can sketch as a rough map. And how will you learn each piece, which means hunting down the resources and methods that experts and serious learners actually recommend.
The research itself is a reading-and-note-taking task, which is where a capture habit pays off immediately. As you read forum threads, course syllabi, and practitioner blog posts, highlight the recommendations that keep recurring and the prerequisites people warn about. Using Glasp's web highlighter here means your scattered research collapses into one searchable list of "what to learn and in what order," instead of forty open tabs you'll never reopen. The map you build in this phase is the thing that keeps the whole project pointed in the right direction.
Directness: Learn the Thing You Actually Want to Do
If you remember one principle from this book, make it directness, because it fixes the most expensive mistake in self-teaching.
Directness means practicing the skill in the context where you'll actually use it, as early as you can stand to. The opposite, which Young warns against repeatedly, is learning a comfortable proxy and assuming it'll transfer. People want to write, so they read books about writing. People want to code, so they watch tutorial after tutorial. People want to speak a language, so they tap through a vocabulary app on the train. All of that feels like progress, and almost none of it is the skill itself.
The problem is that skills are weirdly specific. Being good at multiple-choice grammar questions barely predicts whether you can order food or argue a point out loud. The gap between "knows about the thing" and "can do the thing" is enormous, and you only close it by doing the thing. So if your goal is conversation, you find a conversation partner in week one and fumble through it. If your goal is shipping software, you start building something real and broken on day three, looking up only what you need to get unstuck.
This doesn't mean theory is worthless. It means theory should be pulled in as you need it for the actual task, not stockpiled in advance for a someday that never arrives. Watch the explainer, then immediately apply it. If you're learning from a video course, you can pull a written breakdown of a lecture with YouTube Summary, grab the two or three ideas you need, and get back to building, instead of passively watching nine more hours first. The Feynman technique is directness applied to understanding: you don't know an idea until you can produce it yourself, plainly, without the source in front of you.
Drill Your Weakest Points
Directness gets you doing the real thing, which is good, but it has a downside. When you practice a whole complex skill all at once, your weakest sub-skill quietly caps your progress, and easy practice on the parts you're already decent at hides it.
That's where drilling comes in. A drill is deliberate, isolated practice on the one component that's holding everything else back. The logic mirrors a chain: the chain is only as strong as its weakest link, so you find that link and hammer it directly rather than spreading effort evenly across the whole thing.
The hard part is honest diagnosis. Say you're learning to write clearly and your drafts keep landing flat. The bottleneck might be sentence rhythm, or weak openings, or that you bury the point three paragraphs down. Until you name which one it is, you'll keep writing whole essays and improving slowly. Once you name it, you can drill it: write twenty first sentences in a row, or take one muddy paragraph and rewrite it five different ways. Isolate, attack, then fold the improved piece back into real work.
Young calls this the "Direct-Then-Drill" approach, and the order matters. You practice the whole skill first (directness), which reveals where you're weak, then you drill the weak part, then you return to the whole. Drilling without first doing the real thing risks the classic error of perfecting a sub-skill nobody actually needs. The point of a drill is always to improve the full performance, never to collect a tidy isolated skill for its own sake.
Retrieval and Feedback: The Engine Room
Two principles do most of the heavy lifting in actually building competence: retrieval and feedback. They're separate ideas, but they run together so closely it's worth treating them as the engine room.
Retrieval is the act of pulling knowledge out of your own head rather than putting it back in front of your eyes. Re-reading your notes feels like studying. It mostly builds familiarity with the page, which your brain mistakes for mastery. Trying to recall the idea from a blank page, before you check, is harder and far more effective. This is the testing effect, one of the most replicated findings in learning research, and our deep dive on active recall covers the mechanism in detail. The short version: if it feels easy, it's probably not working.
Feedback is the other half. Young's distinction here is sharp and useful. Most feedback is just outcome feedback ("that's wrong") or worse, vague praise. What accelerates learning is informational feedback that tells you specifically what to change. A language exchange partner who corrects the exact preposition you keep botching teaches you more than ten gold stars. The aim is to get feedback fast, get it specific, and not flinch from the harsh kind, because the harsh kind is usually the useful kind.
Here's how the two combine in practice. After learning something, you test yourself on it from memory (retrieval), then check against the source and note exactly where you were wrong (feedback), then drill that gap. You can build this loop right on top of your reading. Save the passages that matter with highlights, then have Glasp's AI chat quiz you on those highlights and answer from memory before you peek. You get retrieval and immediate, specific feedback in one sitting, and the gaps you find tell you what to drill next.
| Habit | Feels like | Actually does | When to use it |
|---|---|---|---|
| Re-reading notes | Solid, reassuring | Builds page familiarity, little recall | A quick refresh right before you perform |
| Retrieval (self-testing) | Hard, sometimes embarrassing | Builds durable, usable memory | The default review after any real session |
| Passive watching | Productive, efficient | Exposure without ability | Only as input feeding immediate practice |
| Direct practice | Exposing, uncomfortable | Builds the actual skill | The core of every project |
| Informational feedback | Often stings | Tells you the exact thing to fix | As fast and as specific as you can get it |
Retention: Making the Learning Last
Learning a skill quickly is one problem. Still having it six months later is a different one, and Young dedicates a whole principle to it because intense projects are weirdly prone to evaporating.
The threats to retention are simple to name. You forget through plain disuse. You confuse similar ideas that were never clearly separated. And sometimes you never encoded the thing deeply enough to begin with, because you crammed it. The fixes map onto each threat. Space your practice out over time instead of bunching it, so each return trip reloads the memory more strongly. Practice to the point of overlearning for the parts you can't afford to lose. And favor procedures you actually do over facts you merely store, because doing tends to stick better than knowing.
Spacing is the highest-leverage move for a reader, and it costs almost nothing. A little forgetting between sessions is a feature: when recall has gotten slightly harder, retrieving the idea reloads it more firmly than if it were still fresh. That's the science under spaced repetition for readers, and you don't need flashcard software to start. You need a schedule and a willingness to revisit.
This is also where your captured highlights earn their keep long after a project ends. The notes and passages you saved during an intense learning sprint become a personal corpus you can resurface on a widening schedule, whether they came from articles, courses, or Kindle highlights on the books you worked through. A skill you learned fast and never revisited will fade. The same skill, with a light review loop sitting on top of your highlights, holds. For the broader habit of not losing what you read, see how to remember what you read.
A Lightweight Ultralearning Project You Can Run This Month
Enough principles. Here's the whole method compressed into a single small project you can start this week, designed for a normal schedule rather than a year off work. Pick one concrete skill you can name, something like "write a clear one-page memo" or "read and understand basic financial statements."
Week 0, metalearn. Spend two or three short sessions researching how people actually get good at this skill. Read practitioner advice, scan a course syllabus or two, and highlight the recurring recommendations and prerequisites. Turn that into a rough map: the handful of sub-skills you need, in a sensible order. This is your plan, and it'll change, which is fine.
Week 1, go direct. Start doing the real thing immediately, badly. Write the actual memo. Read an actual annual report. Don't pre-study for a week first. The point is to produce a real, flawed attempt that shows you where you stand and what's hard.
Week 2, diagnose and drill. Look honestly at your week-1 attempts and name the single weakest link. Then drill it in isolation: short, focused, repeated reps on that one component. Pull theory in only as the drill demands it. If a concept is fuzzy, grab a quick written explainer or YouTube Summary of a lecture, take what you need, and get back to practice.
Week 3, test and get feedback. Run a retrieval loop on what you've learned: close the source and try to reproduce the key ideas and moves from memory, then check and note exactly where you missed. Get one round of outside, specific feedback if you possibly can, from a person, a community, or by having Glasp's AI chat interrogate your highlights and answer back. Drill whatever the feedback exposes.
Ongoing, retain. Schedule three or four short reviews over the following weeks at widening intervals, using your saved highlights as the prompts. That's the spacing effect on autopilot, and it's the difference between a skill you had for a month and one you keep.
Notice what's absent: no marathon study binge, no buying ten books you won't open, no waiting until you "feel ready" to do the real thing. It's plan, do, diagnose, drill, test, space. That's ultralearning at human scale.
The Honest Limits of Ultralearning
A guide that only sold you the upside would be doing exactly what the book warns against: skipping the inconvenient feedback. So here are the real limits, because knowing them is what makes the method usable.
First, it's genuinely intense, and intensity has a ceiling. Young's flagship projects were full-time efforts by someone who arranged his life around them. Most people can't sustain that, and trying to run every goal at ultralearning intensity is a fast route to burnout. The honest move is to reserve the approach for a few high-value skills that justify a concentrated push, and let most of your learning stay slow and steady. Ultralearning is a sprint tool, not a way to live.
Second, directness can tempt you to skip foundations you actually needed. "Just do the real thing" is excellent advice right up until the real thing has genuine prerequisites, and jumping in cold leaves you flailing with no idea why. Some skills, math-heavy ones especially, punish the learner who refuses to build a base first. The fix isn't to abandon directness; it's to let your metalearning research tell you honestly where a foundation is required, and to pull theory in the moment the task demands it rather than either stockpiling it or skipping it entirely.
Third, and most quietly important, there's survivorship bias baked into any book built on dramatic success stories. We hear about the person who taught themselves a degree in a year. We don't hear about the many who tried similar feats and quietly stalled, and we can't fully separate the method from the unusual drive, time, and circumstances of the people it worked for. Treat the principles as well-grounded learning science, which they are, and treat the timelines as inspiration rather than a benchmark you've failed to hit. Young himself is more measured than his stunts suggest, and his actual examples and caveats are worth reading in full. Consider that your nudge to buy the book; this is a guide to applying it, not a replacement for it.
Frequently Asked Questions
What is ultralearning, in simple terms?
Ultralearning is a term coined by Scott Young in his 2019 book for self-directed learning that's unusually intense and aimed at a specific, hard skill. The three key features are that you own the plan rather than following a school's, you concentrate your effort instead of spreading it thinly over years, and you target a concrete capability you can name. The book distills the approach into nine principles you can apply to ordinary goals, not just the dramatic projects Young is known for.
What are the nine principles of Ultralearning?
They are metalearning (research how to learn the skill before starting), focus (cultivate the ability to concentrate), directness (practice the actual skill in context), drill (isolate and attack your weakest sub-skill), retrieval (test yourself instead of re-reviewing), feedback (seek fast, specific, honest feedback), retention (use spacing and practice to make learning last), intuition (build deep understanding rather than memorizing), and experimentation (try new methods as you advance). They work together, and most beginners get the biggest gains from metalearning, directness, retrieval, and feedback.
Do I need to quit my job to do ultralearning like Scott Young?
No. Young's famous projects were full-time, but the principles scale down to a normal schedule and a single small skill. You can run a complete cycle, plan, practice directly, drill your weak spot, test yourself, and space your reviews, in a few short sessions a week over a month. The intense full-time version is one application of the method, not the method itself, and treating the dramatic timelines as the goal is a common way people set themselves up to quit.
What's the difference between directness and drilling?
Directness means practicing the whole skill in its real context, like actually writing the essay or holding the conversation, rather than studying a proxy for it. Drilling means isolating one weak sub-skill and practicing it on its own. Young's recommended order is "Direct-Then-Drill": do the real thing first so you can see where you're weak, then drill that specific weakness, then fold it back into the full skill. Drilling without first doing the real thing risks perfecting a component you didn't actually need.
How can I use ultralearning to learn from reading and online courses?
Treat your reading as direct practice and your highlights as the raw material for retrieval and retention. During the metalearning phase, highlight recurring recommendations so your research becomes a searchable plan. When learning from video courses, pull a written summary, take only what you need for your current task, and get back to practicing instead of passively watching. Then test yourself on your saved highlights from memory, and revisit them on a widening schedule so the learning sticks long after the project ends.
Conclusion
Ultralearning is easy to misread as a book about superhuman feats, but its real value is the quieter claim underneath them: that teaching yourself a hard skill is a strategy you can run on purpose, not a gift you either have or don't. Map the skill before you start. Practice the actual thing early and badly. Find your weakest link and drill it. Test yourself instead of re-reading, get specific feedback fast, and space your review so the skill lasts.
For a reader, the method folds neatly into a habit you may already have. Your research becomes a plan when you highlight what matters. Your highlights become a quiz when you test yourself from memory. Your saved passages become a retention loop when you resurface them over time. None of that requires a year off work, just a willingness to do the slightly uncomfortable thing, which is the whole secret of the book.
Pick one small skill this week. Spend an hour researching how to learn it with Glasp as your notebook, then go do the real thing badly tomorrow. In a month you won't have an MIT degree. You'll have a skill you built yourself, and a method you can run again on the next one. Then read Young's book for the full picture, intensity, caveats, and all.