It’s so incredible how deep the rabbit-hole goes when you try to solve a real world problem and you have the capacity to understand and pull information from a vast number of seemingly unrelated works.

I really believe that this is the best time to be a polymath, or at least have a broad spectrum of knowledge and references to look into and pull information from; and that being a true generalist that can dive as deep as needed enables you to build great stuff. But maybe that’s just my experience.

I second that, I feel like right now, with the rise of ML tools in audio production, demucs, audio to midi, voice clones etc, the rise of image generation and text. Coupled with some coding skills and interests in many different fields I could not get bored in a million years because there can be so much to jump into and learn/create/explore

> It’s so incredible how deep the rabbit-hole goes when you try to solve a real world problem and you have the capacity to understand and pull information from a vast number of seemingly unrelated works.

It's funny, I considered quoting this other part as well:

> The big prize is to discover a new fractal bud. You notice a crack in the surface of knowledge, pry it open, and there's a whole world inside.

The nature of fractals is that everything is a new fractal bud. There's really endless complexity everywhere. So I don't think that alone is the "big prize". There's some other dimension, like utility or interest. Because I get that discovering a new subfield of topology is different from discovering the new sounds you can make banging on your stove. But it's not just that one has more to it than the other.

Real world problems with disparate fields involved are a rich of source of "medium sized" fractal buds by this unnamed measure. No one is dedicating their life to your application of measure theory to data dashboards, but it's meatier than searching in the absurd and easier to find than breaking ground in pure theory.

> The nature of fractals is that everything is a new fractal bud. There's really endless complexity everywhere. So I don't think that alone is the "big prize". There's some other dimension, like utility or interest.

Here's another dimension: try convincing others of this while they're discussing a specific object level problem and see how that goes.

I don't really follow. You've quoted a few statements and I also don't know what you mean by a specific object level problem.

Is it that it would be hard to convince people there's endless complexity in this domain while they're deciding what to get for lunch? Yeah, probably. They're too hungry.

* * *

oh gosh I understand now and totally agree. I want to wire up an LLM to read a doc and just point out all of these sorts of flaws.

If you haven't read this (not LLM related), I highly recommend it: https://statweb.stanford.edu/~cgates/PERSI/papers/thinking.p...

Re. "this is the best time to be a polymath". I was noticing something like that. For a while it became impossible to know or do "science" ("philosophy", was it?) as a whole. Too broad, too deep. That was not the end of the polymath but it was the end of truly broad expertise in one individual. Then the net in general made so much info available painlessly. (Much faster to dig deep on a narrow issue and switch issues - than say, even with a large academic library.) So that now, it's still not possible to master the forefront of tech or science on a very broad front, but it is possible to dig deep as needed to address this or that problem in the pursuit of what is now just about always a multidisciplinary project.

With the very present danger that many feel that a couple youtube videos is as deep as they ever need to go.

Being able to gauge how deep and broad you have to go for each difficulty you encounter has become an important skill. But polymath seems very possible.

Yes! I've found this approach very useful with my own projects. I might not be out here inventing C++ or Linux, but it's actually not too hard to find projects where you can apply well known computing techniques or technologies to a new domain to do something truly new.

I think this is really motivational because doing something new and showing the world is really fun!

> I might not be out here inventing C++

thank you!

World already has one C++ too many . Thanks op

I don’t know man, farming in most of the world is a totally risky activity.

You might say the climate crisis has made it interesting.

What climate crisis? Row crop yeilds have been on trendline for decades (it's a positive trendline). Sure, you can attribute some of that to superior genetics, new farming techniques, technologies, etc. but any farmer will tell you mother nature trumps all and for yields to consistently reach higher and higher year after year with few exceptions I wonder what climate crisis you're speaking in regards to because I'm not seeing it

Just a self-delusion due to lack of research. It's all due to technology advances, all of it. Against increasing drought and heat.

Sure corn grows in more places now. Because they warmed up - a concern of its own. They'll not stop warming up. Witness https://globalnews.ca/news/9761043/dry-spring-southern-alber...

Some categories are explored or unexplored because either they didn't have a compelling, defining entrant like smart watches or they're Chindōgu like flying cars and VR.

Build something that's currently painful you know there's a definite need for and people would gladly pay money for. Solving a burning pain is far more compelling than incrementally better with the gotcha of introducing the risk of change.

The biggest mistake people make is not letting things soak in a lean, passive income marketable way. They'll build something, shake the trees for customers for a little while, and then turn it off 3 months later when they're not instant internet billionaires. Not working would be 15 years later <$10k/year net profit. Let it simmer with as little engineering investment as possible. Never waste time on churn for churn's sake, or effort that doesn't add end-user UX value.