Hacker News Reader

Comment by gpttpg | Hacker News Reader

How can I effectively narrow down the search space? I often observe myself to be going around circles/stuck in rabbit holes. Lack of clarity perhaps.

I was thinking of something along the lines of CEA (I felt that I could relate to this) with better heuristics if it makes sense. Sort of like optimized bruteforce methods.

If it doesn't, then it would be helpful to know why and what can I do to improve on these flaws.

I asked this question partly because I have seen people with really good investigative/debugging skills and it reminded me of Sherlock Holmes. "How do they do it?" This question would often strike my mind.

View on HN

It's about understanding the system holistically, and recognizing patterns. If I understand 'p -> q', and I observe 'q', I can deduce 'p'. Obviously, this is the simplest example, and real systems are more complicated, but it all works off of understanding causality. If you understand projectile physics, and you find a projectile, you can deduce the path taken. (The arrow of time is one of physic's big mysteries though.)

What concrete approaches would you take to internalize this?

I think the major driving factor would be curiosity, which is why I like the Conan Doyle quote - all the information is there, but it must be observed, and often the way to do that is to keep asking "why?" "Q" is true - why? Well, "P -> Q". Then, why is "P"? Toyota uses this strategy, they call it "Five whys."[1] Unfortunately, society is not super fond of this, as it is probing, and rather difficult. Just think of kids asking "Why is the sky blue?" Not a popular question with adults.

A second big point would be to realize that everything is part of a system. Nothing exists in a vacuum. As part of science, we using modelling which assumes vacuums or other inaccuracies to make reasoning possible. Think of the "spherical cow".[2] These simplifications make reasoning easier, but have hard limits that must be recognized. Quantum mechanics are true at a tiny scale, and relativity is true at a macro scale, but we do not have a well tested unifying theory (string theory and such are possible explanations).[3] Physicists are still asking why to understand.

As for growing these skills, that's tricky. Arthur Conon Doyale was a physician[4]; he studied an incredibly wide range of topics to inform himself. Knowledge is required, and unfortunately, the way to knowledge is through experience. The story of the master ship builder who looks for an hour, and taps once to fix a broken ship is true - you've identified it yourself. The process of gaining experience is one that people would love to hack, but unfortunately, in real life, there are no silver bullets. Your best "hack" here is probably meditation, though it sounds strange. The other "hack" might be one of "always saying yes" like in the Jim Carrey movie "Yes Man."[5] You get a lot of experience rather quickly always saying yes, but it seems rather dangerous too. Reading a lot helps: HN is a great source of diverse information which is nice.

One of, if not the biggest, problem with Capitalistic commodification of labor is the reduction of human capability. Briefly, we see this in a few places: one is the devaluation of experience, and the other is the preference for credentials over skills.

Experience allows humans to grow - a junior dev can, in fact, become a senior dev. Most commodities cannot do this: your tier 2 generic branded Orange Juice can never improve to become a tier 1 Tropicana OJ. So, experience isn't valuable to some people who treat humans like commodities.

The credentials come into play because they can be used, theoretically and incorrectly, to identify you as Tier 1 Tropicana or Tier 2 generic brand OJ.

Capitalism baby.

[1] https://en.wikipedia.org/wiki/Five_whys

[2] https://en.wikipedia.org/wiki/Spherical_cow

[3] https://en.wikipedia.org/wiki/Fundamental_interaction

[4] https://en.wikipedia.org/wiki/Arthur_Conan_Doyle

[5] https://www.imdb.com/title/tt1068680/

Interesting. That's why I come to HN. I really appreciate the detailed post.

I have another question.

How would you go about solving unsolved problems? It's quite to difficult to not end up in rabbit holes and lose track of what you are doing.

I have often ended up chasing the wrong thing/going in the wrong direction.

My pleasure!

This one is a little trickier for me, for some reason. Probably the first step is to identify the specific problem you need to solve. We need to find 'Q', or the relation between P and Q. Maybe you need to break the problem up into peices, and solve one at a time. How to build a website? How to build a UI? How to host?

Then identify and capture the facts as you know them. X -> Y, P, Q. This is a research phase where you want to understand all that you can, within reason. You don't need quantum mechanics for your average application design. So you need login for your site? Pictures? Forms? Charts?

Then conceive of a framework for analyzing. On a 2D plane, polar or Cartesian coordinates for instance, as well as picking the origin. Maybe you need logic, or something else. Richard Feyman invented his quantum Feyman diagrams to help him understand quantum mechanics. Calculus, derivatives, and integration were invented for calculating the areas of spaces between complex curves, and two people, Leibnitz and Newton, each came up with their own notation to represent the ideas of adding infinitely tiny slices. Sometimes you might need to create a tool. For webdev I think this is like, single page app or server side generation or whatever.

Once you have data, and a framework, you can analyze for a solution. I also think of something about graph searches might be applicable: breath first is best when the solution is close to the source, and depth first is better for farther answers/more exhaustive searches. Ok, server side generation has these options with these pros and cons, jsx, templating, whatever.

Also, one of the best strategies in programming (and life) is: "small steps, always," from the pragmatic programmer, which I highly recommend. Feymans book are also incredible, he teaches learning as much as physics.

Reply on news.ycombinator.com