Read Andy’s Evergreen Garden on the topic of knowledge nets and spaced repetition. By that I mean you should at most skim it of course.
As a consequence of the above point, you can always read it faster. Most stuff is fluff, so you can get away with skimming most of the time.
Don’t try to come up with new ideas; just try your utmost to hold on to what you know (preferably via spaced repetition), and new ideas automatically appear.
The Forgetting Curve decays exponentially (there’s a Lindy effect wrt memory), which means reviewing a near-forgotten idea will have exponentially greater return (in terms of retention / opportunities to find a use-case) than learning something new. Resist the recency effect.
Aim your ideas at your own head.
If the ideas you came up with for your own purposes happen to potentially be usefwl for someone, you may teach as a side-project, but making it your main quest is worse for both purposes.
Don’t compromise your purpose.
If you’re sampling for ideas (or anything) that conjunctively score high on N independent criteria, the probability decreases exponentially with N. Assume each variable has the same probability p of scoring above the threshold, and the probability that each of them do is pN. Unless the utility of the idea increases multiplicatively with each IID criterion it satisfies, the conjunctive search is unlikely to be worth it, and you should consider horizontally segmenting your search.
Optimise the generator.
Ahmdal’s law: the efficiency gained by optimizing a single part of a system is limited by the fraction of time that the improved part is actually used.
A marginal improvement in your sleep-quality, or the way you process ideas, may have an immense impact on you as a goodness-generating system.
Because of point 5, one of the most important variables you can optimise within yourself, is how aligned your mind is with the idea of helping others. It is extremely typical [citation needed] for altruists to have confounders in their own motivations & thinking-habits without being aware of it, especially as relates to social belonging, prestige, and not being laughed at. But these compromising confounders have a heavy cost wrt idea-sampling.
Learn the generalized game.
You need fewer parameters to specify a generalized game (eg NxN-sudoku) compared to specific variants, but if you solve the former, that often mostly solves the variants by extrapolation. This is one aspect of the inventor’s paradox, and it implies that it’s sometimes easier to be more ambitious.
Roll the biggest dice
Rolling 1d60 vs 10d6 is the same max and ~same avg, but the former has a 160 chance of max, whereas the latter has only 6−10.
Projects whose outcome depends on fewer variables tend to have longer tails.
初心
Consider all projects first from a perspective where you assume nobody has ever tried it before.
For all questions you encounter, act as if it’s your first time, act as if nobody has ever asked it before.
How does an artist learn to draw? As soon as they stop relying on their teacher’s eyes as proxies for their own—as soon as they learn to see. Unless you do both, you have no internal feedback-loop by which you can independently grow.
知行合一
Knowledge that doesn’t change you is no knowledge at all. You gain wisdom in proportion to how much you change. The one thing you should monitor at all times is your rate of change wrt anything you think you’re learning from.
If you read a whole textbook with no perceptible general changes in your behaviour (knock-on effects due to credentials don’t count), that was worth less to you than a random tweet you somehow found the motivation to process truly and deeply.
Make up a dumb plan, and pursue it at full speed until you learn why it’s dumb.
Then make up another dumb plan, and repeat until you find yourself at the completion of a plan you couldn’t figure out the dumbness of. Maybe it was worth doing after all.
More than anything, dumb plans and naive models provide you with sensitivity to evidence, and lets you learn much faster. Δ.
Read less. Think more.
Your brain is much better than you think, once you learn to unconditionally support it as it clumsily emerges from the confines you’ve kept it in.
Those near you may not like it; but those far from you may benefit.
12 research aphorisms nobody asked for and yet
You can always make it shorter.[1]
Lists are good for that.
Knowledge nets like RemNote and Obsidian even better.
Read Andy’s Evergreen Garden on the topic of knowledge nets and spaced repetition. By that I mean you should at most skim it of course.
As a consequence of the above point, you can always read it faster. Most stuff is fluff, so you can get away with skimming most of the time.
Don’t try to come up with new ideas; just try your utmost to hold on to what you know (preferably via spaced repetition), and new ideas automatically appear.
The Forgetting Curve decays exponentially (there’s a Lindy effect wrt memory), which means reviewing a near-forgotten idea will have exponentially greater return (in terms of retention / opportunities to find a use-case) than learning something new. Resist the recency effect.
Aim your ideas at your own head.
If the ideas you came up with for your own purposes happen to potentially be usefwl for someone, you may teach as a side-project, but making it your main quest is worse for both purposes.
Don’t compromise your purpose.
If you’re sampling for ideas (or anything) that conjunctively score high on N independent criteria, the probability decreases exponentially with N. Assume each variable has the same probability p of scoring above the threshold, and the probability that each of them do is pN. Unless the utility of the idea increases multiplicatively with each IID criterion it satisfies, the conjunctive search is unlikely to be worth it, and you should consider horizontally segmenting your search.
Optimise the generator.
Ahmdal’s law: the efficiency gained by optimizing a single part of a system is limited by the fraction of time that the improved part is actually used.
A marginal improvement in your sleep-quality, or the way you process ideas, may have an immense impact on you as a goodness-generating system.
Because of point 5, one of the most important variables you can optimise within yourself, is how aligned your mind is with the idea of helping others. It is extremely typical [citation needed] for altruists to have confounders in their own motivations & thinking-habits without being aware of it, especially as relates to social belonging, prestige, and not being laughed at. But these compromising confounders have a heavy cost wrt idea-sampling.
Learn the generalized game.
You need fewer parameters to specify a generalized game (eg NxN-sudoku) compared to specific variants, but if you solve the former, that often mostly solves the variants by extrapolation. This is one aspect of the inventor’s paradox, and it implies that it’s sometimes easier to be more ambitious.
Roll the biggest dice
Rolling 1d60 vs 10d6 is the same max and ~same avg, but the former has a 160 chance of max, whereas the latter has only 6−10.
Projects whose outcome depends on fewer variables tend to have longer tails.
初心
Consider all projects first from a perspective where you assume nobody has ever tried it before.
For all questions you encounter, act as if it’s your first time, act as if nobody has ever asked it before.
How does an artist learn to draw? As soon as they stop relying on their teacher’s eyes as proxies for their own—as soon as they learn to see. Unless you do both, you have no internal feedback-loop by which you can independently grow.
知行合一
Knowledge that doesn’t change you is no knowledge at all. You gain wisdom in proportion to how much you change. The one thing you should monitor at all times is your rate of change wrt anything you think you’re learning from.
If you read a whole textbook with no perceptible general changes in your behaviour (knock-on effects due to credentials don’t count), that was worth less to you than a random tweet you somehow found the motivation to process truly and deeply.
Make up a dumb plan, and pursue it at full speed until you learn why it’s dumb.
Then make up another dumb plan, and repeat until you find yourself at the completion of a plan you couldn’t figure out the dumbness of. Maybe it was worth doing after all.
More than anything, dumb plans and naive models provide you with sensitivity to evidence, and lets you learn much faster. Δ.
Read less. Think more.
Your brain is much better than you think, once you learn to unconditionally support it as it clumsily emerges from the confines you’ve kept it in.
Those near you may not like it; but those far from you may benefit.
Bonus aphorism: we teach what we most wish to learn…