Top
Best
New

Posted by maksimur 9/10/2025

We can’t circumvent the work needed to train our minds(zettelkasten.de)
385 points | 178 commentspage 3
qwertytyyuu 9/10/2025|
You definitely do not need to remember everything, it’s not worth the effort to try, famously in programming even the best look up things they have looked up before.

Memory is helpful but brains aren’t hard drives, they aren’t designed to store information perfectly.

wolttam 9/10/2025||
I remember things just fine, just not at the sufficient detail to remember all aspects at the drop of a pin. What I hold on to are the core concepts that allow me to hit the ground running when I have to interact with the subject-matter again.
NitpickLawyer 9/10/2025|
> What I hold on to are the core concepts that allow me to hit the ground running when I have to interact with the subject-matter again.

Exactly. And those also come with doing the thing, or watching the thing being done, or reading about the thing, or thinking about the thing. You often don't have to actively try to kastle / cram / grok / etc the information to remember it. Just being exposed to it will make you remember some of it. Knowing where to get more accurate info is often a greater skill / benefit than knowing the details yourself. Especially in fields where knowing all the details yourself is almost impossible.

Weak article.

birdman3131 9/10/2025||
AI is not a replacement for a greybeard.

That said AI, Search and the like can be quite useful and helpful.

wduquette 9/10/2025|
I remember the dotCom bubble. After the bubble burst, people got on with putting storefronts and other kinds of business on-line in a more sober fashion.

I predict the same thing will happen with the current AI tools: the bubble will burst, a bunch of folks will lose their shirts, and the world in general will come to a more realistic and sober understanding of what they are good for. We will figure out how to provide the useful parts without massive data centers and it will become natural. (I remember when things a graphics card can do trivially required a supercomputer with supporting staff.)

Animats 9/10/2025||
"To find what you need online, you require a solid general education and, above all, prior knowledge in the area related to your search." And get off my lawn. That's more a criticism of search engines than AI, anyway.

Insisting that people know exercise physiology to work out is a bit much. That's what trainers and coaches are for. Now drop and give me twenty.

The real problem with LLMs remains that they can't really do the job of thinking yet, because they're wrong too often. They can both hallucinate and get lost. What the AI situation will be in five years, we don't know.

tim333 9/11/2025||
Looking at the example he gives of an exercise plan, I'm not sure the advice is very good. He says don't do the simplistic plan the LLM gave, and be aware of technical stuff like "discussions of Olaf Alexander Bu on the difference between capacity and power numbers".

My experience as a bit of geek is it's easy to get caught up in theory and do stuff that doesn't work in practice. Better than either the LLM or theorising is an experienced coach who can look at what you are doing and say don't to that - it's all wrong - do this etc.

constantcrying 9/10/2025||
But in how many cases is the "why" more important than the "how"?

People can drive a car, without understanding any of the mechanical and electrical systems of a car. In fact understanding these does essentially nothing for what most people use a car for.

>If you can’t produce a comprehensive answer with confidence and on the whim the second you read the question, you don’t have the sufficient background knowledge.

And then what? Why would this matter at all? You can successfully use a workout schedule without understanding what its strengths and weaknesses are and how they align with your goals.

shadowgovt 9/10/2025||
Possibly worth considering the source: the author coaches on a mechanism for collecting and collating information. They have a vested interest in the notion you have to do the work (instead of letting the machine do it for you); indeed, they hope you do the work using this method they coach on...

(That having been said: I've used zettelkasten myself a bit and I'd say it's worth a try. Probably not for everyone but the underlying idea of "building out an artifact to supplement your memory and understanding of what you've seen" is an intriguing approach).

palata 9/11/2025||
I once saw a comment that said "AIs are great for things that are hard to find, but easy to verify". Which I agree with.

In order to verify, we need knowledge.

If the AI is the last stop ("I asked my LLM and it said that"), it's the wrong way to do it. If the AI is part of the process ("The LLM mentioned this, which made sense, and I confirmed it by [...]"), then it may (or may not) be useful.

In any case, "knowing it" is always more efficient than "having to search for it". So the more you know, the better.

procaryote 9/10/2025||
> The reduced engagement with the material reduces the emotional weight of the whole line of action. You mind is an engine that is fuelled by emotion. Without any emotion, you don’t think. Rather, you try to imitate thinking efficiently.

This doesn't sound true and they don't seem to offer any support for the claim.

There's a whole host of emotion-driven cognitive biases, where an effective counter is to reduce the emotional weight of the whole line of action.

Of course, to their credit, it's only by remembering those biases that I could see their error

OmarShehata 9/10/2025||
The first thing that happened in your mind when you read that sentence is (1) a bad feeling. That then triggered (2) a rational, conscious thought that interpreted that bad feeling: "this feels bad because it's not true, here are the reasons why it is not true.

There is ALWAYS an "emotional/intuitive" response that precedes the rational, conscious thought. There's a ton of research on this (see system 1 vs system 2 thinking etc).

There is no way to stop the emotional "thought" from happening before the "rational thought". What you can do is build a loop that self reflects to understand why that emotion was triggered (sometimes, instead of "this feels bad because it's wrong", it's "this feels bad because it points to an inconvenient truth" or "I am hungry and everything I am reading feels bad")

procaryote 9/10/2025||
That's very hard to know without being in an FMRI machine while reading, which I wasn't, sadly.

Just functionally, it seems reasonable that something happened before that bad feeling to trigger it, e.g. "trying to fit this with already known things, and finding it doesn't".

OmarShehata 9/11/2025||
this isn't a "it depends on what computation happens in this specific case" question, this is a "how does human cognition work".

Every website you visit has the payload delivered over the network before any JS is parsed. It has to, there's no other way. Same with intuition followed by rational thought

procaryote 9/11/2025||
I'm not sure human cognition is that clear cut or linear.

I guess if you define all rapid cognition as Intuition and all slower conscious cognition as Rational Thought, you're right by definition, but while that might provide a happy feeling, it might not give you any useful insights upon deeper inspection.

For instance, it doesn't support or disprove the "less emotion is bad for thought" idea

jbreckmckye 9/10/2025||
Isn't your argument a support of his claim?

If emotions did not weigh on recall, surely there would be no "emotion-driven cognitive biases"

procaryote 9/10/2025||
If the claim was "reducing the emotional weight of the action makes your thinking worse" – No.
darepublic 9/10/2025|
At my first software dev internship my manager asked me to code in languages I was not trained in. I told him I would need some time to study up on these. He scoffed and said just look up what you need on Google. Initially I resisted, I felt like it was too shallow. That it was akin to copying answers I didn't really comprehend. However it didn't take long to pick up the habit. Learning is like going to the gym now, it's self enforced discipline
bluGill 9/10/2025|
That is because learning a computer language is not hard. Learning to program in the first place is hard, but once you know that the language itself is not hard to learn. However it is easy to verify someone knows the syntax of some specific language and hard to check if they actually know how to program.
More comments...