When I look at this article and ask “What is the strongest counter-example that comes into my head”, and I think something like “Being a mathematician or physicist”. I think I have the g-factor to get a math degree if I really wanted, but it’s not at all clear how I could reach the level of a working mathematician in six months, despite already being a decent programmer. There are a few replies my inner Habryka might say, and I’m curious if any of them seem accurate, and if some are very much not what you intended to say:
This advice isn’t meant for everyone. In the same way that I’m expecting people to be 140+ IQ to use this advice, I also expect them to be at least +2.5 SD of conscientiousness. If you can put in 60+ hours a week of high quality mental work into this task, you actually could achieve this—most students of mathematics do a fraction of this time. You underestimate how much progress can be made with real focus and drive paired with strong conscientiousness and smarts—it’s a multiplicative effect that adds up very quickly. Such a person really could get to that level in six months, but they’re rare enough that standard career paths can’t accommodate them, which is why society needs specialists in the first place.
When you imagine being a working mathematician, you imagine having many years of schooling that teaches you about a very broad variety of maths. This is not what I am imagining—I am imagining having the level of ability to perform a specific type of job a mathematician might do. For example, take this SLT theorist job by Timaeus. https://timaeus.co/blog/updates/2026-04-09-hiring This doesn’t require you to know everything an undergraduate maths degree holds. It requires you to know an important subset and to grasp some universal skills like writing rigorous proofs and doing mathematical research at all. Rolling your own curriculum could get you there in 6 months. Think about your own CS career—does someone need to have learned everything you learned in order to do what you do? Clearly not—if they’d aimed at your specific target from the start, they’d get there in six months.
The kind of person who would be an actual research-grade mathematician really is a level of expertise beyond this post. It’s a level beyond, say, a solid mid-level programmer in their field, and to get to that level it really does take years even if you’re good. It would take months to get to the level of doing an easier class of problems that people would still pay you for, and that’s the level I would expect a Lightcone-level generalist to be able to get to in a different area in the same domain within six months.
My current belief is that people around 2-3 SDs above average can learn everything in an undergraduate degree at a top university in around 6-9 months, if they actually spend all their time doing focused study (I remember there was a guy with the last name “Young” whose full name I don’t quite remember who did this for an MIT undergraduate education and wrote a blog about it).
And then, I do think as things go, being a research-grade mathematician is playing on one of the hardest difficulties available. That said, I remember Aubre De Gray, anti-aging guy not generally widely known for his math research, actually made a pretty serious contribution a few years back, and that felt like a good validation of a generalist world theory: https://www.quantamagazine.org/decades-old-graph-problem-yields-to-amateur-mathematician-20180417/
For the benefit of later readers, this was Scott Young’s MIT challenge, doing a full CS degree in 12 months. I saw this as rather incredible the first time I saw it, but maybe the more incredible thing was having both the circumstances and the agency to attempt it in the first place. https://www.scotthyoung.com/blog/myprojects/mit-challenge-2/
Yep, that’s it! I really love that he did that. I remembered 6-9 months because he did it in just under 12, and also didn’t do it full-time for something like the second half of it. But seems like it took a bit longer.
I did my CS undergrad in about cumulative 3 months (or maybe a bit more if you count a couple physics lectures I took before I switched to CS and optimized studying hard). (At TUM which is the best German university for CS. I didn’t optimize for good grades—final grade was 2.5 - probably slightly below median.) This was over the course of 4 years during which I spent almost all my time on AI safety research.
If the curriculum had only been exams it would’ve been even faster. I spent like an average of 10h per exam learning (where the usual expectation by the credit system is like 180h).
Although I already had a bunch of CS knowledge from autodidactic learning during highschool and I am probably >+3SD, but a lot of it actually came down to optimizing study techniques.
i actually found friends who i could pay to tutor me for quite a lot of lectures. especially one friend was really good at that. but just looking through the slides and asking an AI whenever you have questions works quite decently too i think.
try to understand. see it as structured knowledge. don’t just accept and memorize facts—in CS things are usually at least a somewhat sensible solution to a problem.
recall after each section and each session
find practice exams where solutions exist. after doing an exercise, check immediately the solution. clearly understand mistakes. (if there aren’t sufficiently many public exams then use the homework exercises. there are usually solutions.
overall was probably roughly 50-50 split between learning and understanding the content and doing exercises. you can also skip some stuff that doesn’t seem worth learning.
take sufficient breaks. (i often did 1.5-2h tutor sessions with 1-2 5min breaks.)
also be rested sufficiently in general. non-stimulating rest is better than youtube or so. takes a while to remove such addictions though. you usually don’t want to deplete dopamine low enough that you have a too strong pull to stimulated rest. rather do some walking/hiking or whatever active recovery works for you.
and ofc obvious basics like sleep, reasonably healthy food, sports.
While still significant and hard, my intuition is that a problem ‘of that shape’ requires less front loaded theory. (while still being very smarts and math skill heavy!)
#2 deserves more attention here. In my experience, real expertise has a few key differences from shallow expertise. First, a real expert knows what’s important and what can be ignored. Shallow experts often waste quite a bit of time on things that could, theoretically, matter but in practice don’t.
Also, real experts know the limits of their techniques much better than shallow experts. They’re quicker to say no. 1% of the time this was a mistake and the shallow expert will succeed, but 99% of the time the shallow expert spends years and money pursuing paths that the real expert knew immediately couldn’t work. My first startup was a sensor company, and if we’d known to ask a certain question about market requirements before we started we’d have saved two years and a few million dollars. And the company.
I’ll also add that 6 months sounds like a reasonable time for a smart person to learn coding, but there are a lot of other technical skills in the world and many are much more complicated or inherently much slower. Designing an iPhone is not the sort of thing you can learn in six months, nor is building a fighter jet. Six months on a farm gets you one crop, typically—there’s not enough time to iterate toward expertise.
When I look at this article and ask “What is the strongest counter-example that comes into my head”, and I think something like “Being a mathematician or physicist”. I think I have the g-factor to get a math degree if I really wanted, but it’s not at all clear how I could reach the level of a working mathematician in six months, despite already being a decent programmer. There are a few replies my inner Habryka might say, and I’m curious if any of them seem accurate, and if some are very much not what you intended to say:
This advice isn’t meant for everyone. In the same way that I’m expecting people to be 140+ IQ to use this advice, I also expect them to be at least +2.5 SD of conscientiousness. If you can put in 60+ hours a week of high quality mental work into this task, you actually could achieve this—most students of mathematics do a fraction of this time. You underestimate how much progress can be made with real focus and drive paired with strong conscientiousness and smarts—it’s a multiplicative effect that adds up very quickly. Such a person really could get to that level in six months, but they’re rare enough that standard career paths can’t accommodate them, which is why society needs specialists in the first place.
When you imagine being a working mathematician, you imagine having many years of schooling that teaches you about a very broad variety of maths. This is not what I am imagining—I am imagining having the level of ability to perform a specific type of job a mathematician might do. For example, take this SLT theorist job by Timaeus. https://timaeus.co/blog/updates/2026-04-09-hiring This doesn’t require you to know everything an undergraduate maths degree holds. It requires you to know an important subset and to grasp some universal skills like writing rigorous proofs and doing mathematical research at all. Rolling your own curriculum could get you there in 6 months. Think about your own CS career—does someone need to have learned everything you learned in order to do what you do? Clearly not—if they’d aimed at your specific target from the start, they’d get there in six months.
The kind of person who would be an actual research-grade mathematician really is a level of expertise beyond this post. It’s a level beyond, say, a solid mid-level programmer in their field, and to get to that level it really does take years even if you’re good. It would take months to get to the level of doing an easier class of problems that people would still pay you for, and that’s the level I would expect a Lightcone-level generalist to be able to get to in a different area in the same domain within six months.
Or it could be something else entirely.
It’s IMO a mixture of 1 and 3.
My current belief is that people around 2-3 SDs above average can learn everything in an undergraduate degree at a top university in around 6-9 months, if they actually spend all their time doing focused study (I remember there was a guy with the last name “Young” whose full name I don’t quite remember who did this for an MIT undergraduate education and wrote a blog about it).
And then, I do think as things go, being a research-grade mathematician is playing on one of the hardest difficulties available. That said, I remember Aubre De Gray, anti-aging guy not generally widely known for his math research, actually made a pretty serious contribution a few years back, and that felt like a good validation of a generalist world theory: https://www.quantamagazine.org/decades-old-graph-problem-yields-to-amateur-mathematician-20180417/
For the benefit of later readers, this was Scott Young’s MIT challenge, doing a full CS degree in 12 months. I saw this as rather incredible the first time I saw it, but maybe the more incredible thing was having both the circumstances and the agency to attempt it in the first place. https://www.scotthyoung.com/blog/myprojects/mit-challenge-2/
Yep, that’s it! I really love that he did that. I remembered 6-9 months because he did it in just under 12, and also didn’t do it full-time for something like the second half of it. But seems like it took a bit longer.
I did my CS undergrad in about cumulative 3 months (or maybe a bit more if you count a couple physics lectures I took before I switched to CS and optimized studying hard). (At TUM which is the best German university for CS. I didn’t optimize for good grades—final grade was 2.5 - probably slightly below median.) This was over the course of 4 years during which I spent almost all my time on AI safety research.
If the curriculum had only been exams it would’ve been even faster. I spent like an average of 10h per exam learning (where the usual expectation by the credit system is like 180h).
Although I already had a bunch of CS knowledge from autodidactic learning during highschool and I am probably >+3SD, but a lot of it actually came down to optimizing study techniques.
Have you written about your study technique optimizations anywhere? It would be useful if you could share them.
Not much, but here’s one post: https://www.lesswrong.com/posts/aq84rfx3XRyLd9y2v/optimizing-feedback-to-learn-faster
Some things I did when learning:
i actually found friends who i could pay to tutor me for quite a lot of lectures. especially one friend was really good at that. but just looking through the slides and asking an AI whenever you have questions works quite decently too i think.
try to understand. see it as structured knowledge. don’t just accept and memorize facts—in CS things are usually at least a somewhat sensible solution to a problem.
recall after each section and each session
find practice exams where solutions exist. after doing an exercise, check immediately the solution. clearly understand mistakes. (if there aren’t sufficiently many public exams then use the homework exercises. there are usually solutions.
overall was probably roughly 50-50 split between learning and understanding the content and doing exercises. you can also skip some stuff that doesn’t seem worth learning.
take sufficient breaks. (i often did 1.5-2h tutor sessions with 1-2 5min breaks.)
also be rested sufficiently in general. non-stimulating rest is better than youtube or so. takes a while to remove such addictions though. you usually don’t want to deplete dopamine low enough that you have a too strong pull to stimulated rest. rather do some walking/hiking or whatever active recovery works for you.
and ofc obvious basics like sleep, reasonably healthy food, sports.
While still significant and hard, my intuition is that a problem ‘of that shape’ requires less front loaded theory. (while still being very smarts and math skill heavy!)
#2 deserves more attention here. In my experience, real expertise has a few key differences from shallow expertise. First, a real expert knows what’s important and what can be ignored. Shallow experts often waste quite a bit of time on things that could, theoretically, matter but in practice don’t.
Also, real experts know the limits of their techniques much better than shallow experts. They’re quicker to say no. 1% of the time this was a mistake and the shallow expert will succeed, but 99% of the time the shallow expert spends years and money pursuing paths that the real expert knew immediately couldn’t work. My first startup was a sensor company, and if we’d known to ask a certain question about market requirements before we started we’d have saved two years and a few million dollars. And the company.
I’ll also add that 6 months sounds like a reasonable time for a smart person to learn coding, but there are a lot of other technical skills in the world and many are much more complicated or inherently much slower. Designing an iPhone is not the sort of thing you can learn in six months, nor is building a fighter jet. Six months on a farm gets you one crop, typically—there’s not enough time to iterate toward expertise.