Where do I start with _____?

I don't have any advice for you, but I try to pay attention to efforts to take the Facebook out of Oculus. Please let me know if you have any success!

This is something that is ultimately pretty personal and depends on what kind of knowledge you want to store. Personally I've been using Obsidian and enjoying it a lot.

The main thing I've learned about knowledge management (and habits in general) is that it should be as easy as possible. If you're struggling to get in the habit of taking notes, any system that requires complex categorization, or adds barriers to taking a note is detrimental.

While making a fancy categorization/directory structure feels nice, it can be painful in the long run, as it makes it more difficult to do the one thing you want to do: make notes. If I need more categorization, I just create a new note linking to other notes.

To that end, I've settled on a flat directory of markdown files. Each note describes a concept or idea. If I mention a concept I've made a note of before, I just create a link using the [[square bracket]] link notation. There's a nice plugin for Obsidian that makes it easy to automatically commit changes to a git repo, so that's how I sync/backup my notes.

I pretty much make a note of anything I want to remember: grammar/vocab for language learning, math, how to use software/libraries/programming languages, ideas for creative endeavors, recipes, TODO lists, etc.

As another Obsidian user, I'll echo what @tesseractcat said.

Don't concern yourself with how to use zettelkasten properly¹. I did that mistake myself at the beginning because zettelkasten has some pretty loud advocates.

Pick a software that suits your needs the best and then just start taking notes. The more notes you'll take, the more aware you'll be of your current setup's shortfalls and from there you can just adjust according to your needs. Picking a system of any kind without assessing your needs first usually ends up hurting productivity because over time it becomes work to maintain that system and work around its needs whereas it should be working and helping you with your needs.

Since I've been using Obsidian for over a year now, I can wholeheartedly recommend it, but if it doesn't work for you for whatever reason, the only advice I can give you is make sure whatever software you end up using has great portability options. (The mere support for exporting your files is not enough.) Good luck!

^{1: This is the best website that describes what zettelkasten is. As I learned more about it, I come to realize that it's a neat idea that works best in a physical medium more than a digital one. I think people romanticize the system quite a bit for various reasons. More power to them, of course, but I personally don't think zettelkasten is a great method for digital note taking.}

The best advice that I can give you is to pick a piece of software - Obsidian is good because of data portability, Dendron is newer but looks decent - and just start taking notes. You can worry about connecting them later!

For example, I recently started a new job. I started with numerous onboarding meetings and tons of one-on-one chats. Rather than think about what my desired structure was, I just made one note per meeting, wrote down everything, and then (between a few meetings) made an index note which connected every meeting note with a checkbox next to it. Over the next week or so, I spent a bunch of time between meetings (and during boring ones) processing those stream-of-consciousness notes into a more atomic form - one note per idea. Only then did I take the time to connect those notes and discover the emergent structure among them.

I recently found out about Dendron from the presentation they did for the vscode YouTube. It exists right inside vscode which is my usual development environment so that was a big selling point for me compared to Obsidian. It's also open source.

The first question you should ask yourself is whether you even want a Zettelkasten. I tried to properly set up a Zettelkasten multiple times, but ultimately I just used it for referencing things (recipes, bash commands, etc). Eventually I gave up on the Zettelkasten method proper and decided to keep notes in a personal wiki (wiki.js; I've detailed my reasons for using this particular software before).

How do I start zettelcasting effectively?

I have tried a few times, always on Emacs Org Mode + Org Roam Mode. I usually become so invested in the setup that I forget why I wanted it in the first place. The other problem is that Emacs is the perfect platform on the computer. There are no good solutions for bringing Org to mobile, especially with a non-trivial setup like this one (what am talking about? Emacs is never trivial!). And I don't wanna pay a subscription, anything priced in dollar is not cheap for me.

It's easy; when you start off, all of your paintings will be gauche.

(Sorry, couldn't resist the joke. It's actually spelled "gouache" by the way).

Do you have much experience with any other form of painting? Most skills you learn transfer, especially if you're already familliar with watercolors. The biggest difference is that there's more pigmentation with gouache so it will create a matte effect. That means that you can effectively create an opaque layer on top of other colors.

For the record I also have basically zero art skills and have thrown away everything I have ever made.

I’ll say a thing that I hope is helpful, but I’m apologizing in advance in case it’s not:

The only, and I mean only, way to make good art is to make a LOT of bad art. The first hundred pieces of art made by literally every artist, especially the very best artists, were terrible. If you’re lucky, you get all or most of the truly bad stuff out of the way when you’re a young child and haven’t developed an adult’s miserable sense of perfectionism; if you’re not lucky, you just have to plow through, knowing full well that you’re making crap.

That said, for 2D visual art, here’s what I’d suggest:

• Do really small, simple stuff to start. Not because it’s easier, but because it makes it easier to see how you’re improving! Things like still life portraits of a bowl of fruit, or even a single fruit
• Do the same painting more than once, but change something about it when you do. Sit in a different place relative to your subject, use different colors (they don’t even have to be the right colors!), move stuff around, turn off some lights, whatever strikes your fancy
• Speaking of which, use a real life subject. Again, still life portraits are used so heavily for teaching art, despite how boring they are, because they are easy to set up and will stay in one place for days unchanged. It’s really hard to overstate how valuable it is to paint something that it actually physically in front of you
• If you’re really interested in improving, you can even buy a little sketchbook. The skills you practice with graphite (outlining, composition, proportions, shading) will all transfer to painting, and vice versa. Try to sketch something every day; NOTHING will improve your skills faster. Seriously, sketch something every day for six months and make sure you date your sketches; you’ll be amazed when you look back and see how much you’ve improved.

Don’t compare yourself to experts and professionals. Compare yourself to last-week-you. And even then, only every once in a while. Even if it doesn’t come out how you imagined, it’s still cool as hell that you made a painting!! And if you ever want someone to take a look and tell you how awesome you’re doing, don’t hesitate to hit me up!

This is exactly what I wanted to ask. In fact I asked this on tildes before but didn’t get a response.

Does anyone here know sign language? Not necessarily ASL, just any.

If you are ok with non-English lyrics, Mid-air Thief is pretty great IMO.

Some others:

• Parcels is an up-and-coming funk/pop band that's worked with Daft Punk. They have 2 very solid and mature albums out atm
• MGMT has made some very interesting indie and psychedelic rock that diverges from some of their earlier electro-pop hits
• Mac Demarco has dominated indie music lately.
• BADBADNOTGOOD is a nice jazz group. Mostly instrumental with occasional guest singer's

Here you go :

https://rateyourmusic.com/charts/top/album/2000-2029/

You've got to sift through the list to accommodate for your tastes, but if you over a couple of pages you're pretty much guaranteed to find something you'll love.

Do you have Spotify? Sourcing links will be easier if so.

There's actually a bunch of small scale 3D printers on the market, so you've got a lot of things to choose from.

It sounds like you want to print something that requires a good amount of mechanical strength, so you're probably going to want an FDM printer, and with that in mind the one I recommend to all newbies is the Prusa Mini. It might take a while to get one right now because of the holiday rush. There are others I can recommend if you can't wait that long, but it would be worth it if you can; Prusa seems to offer the greatest amount of support you can get from a 3D printer manufacturer outside of extremely expensive commercial environments.

I've printed out a huge number of different things, from decorative objects to functional one-offs, stands, and storage organizers, and whatever else I could think of. The last thing I've been working on was a miniature board game concept. I'm not great at designing 3D models but I get some practice in from time to time. One of my all-time favorite printable objects is this collapsible dice tower.

I helped a friend in college make a board game and had fun with it! We were pretty amateurish, but I had some takeaways:

• Instead of trying to balance everything you can add catch-up mechanics or way for players that are "behind" to sabotage players that are winning. That may end up being a frustrating mechanic for some people if it's excessive.
• Alternate or even cooperative win-conditions are great. It can give players that got unlucky an out or give different personalities some sense of agency.
• It was a great opportunity to get other people involved. My partner helped with the art and it gave me an excuse to ask people for their experiences for the content/theme.

Theme-wise I think you do almost anything. You could make a game about your neighborhood through a lens of magical realism, a hobby, or something outlandish... the success of it would be more in the execution and your feelings about the project more than having a good theme.

The game we made was based on our school. The distance traveled around the board corresponded to completion of the curriculum, and some RPG elements were inspired by the different majors.

Some questions to start with might be:

• Who do you see yourself playing the game with? What sort of experience would you want to have?

• What's something you know a lot about, or something you'd enjoy researching?

• What games have you enjoyed playing? What mechanics made for a fun game loop?

• What's made you give up on games you've tried? Long turns, too simple/complicated?

• What's easy to do with your kit that are hard elsewhere? How could you approximate a mechanic that isn't easy to do with your kit through something more simple?

Hope you have fun with it!

Unfortunately there's not really any way around math, that's the best and only way we have to model these kinds of things if you want to really understand what's going on. Some pointers, though. First, just forget anything analogies revolving around the human brain - they get parroted by reports on deep neural networks because it sounds cool, but neural networks resemble the brain about as much as the tree data structure resembles an actual tree. Relying on the brain analogy on any more than a surface level will just cause problems.

In terms of math, you only need a little multivariable calculus - once you get the hang of gradients and the derivative/integration rules for matrices you're probably fine. Then it's mostly a big exercise in linear algebra and statistics. Linear algebra is particularly key - neural networks on a base level is smashing matrices together, after all. You'll need to know stuff like eigenpairs, eigen decomposition, SVD, lagrange multipliers in order to do the linear optimization techniques used in neural networks.

To be honest, there's surprisingly (and somewhat unfortunately) not that much more math than that - a lot of neural network research is, uh, surprisingly "empirical", so to say, for the field of computer science.

Here is a useful document from my undergrad ML class that is intended to be a broad overview of math fundamentals that are necessary as a prerequisite.

Work towards understanding this paper. It is the first style transfer paper that was published, and likely the easiest algorithm. You will need to understand how classifiers like vgg16 are architected and trained.

Edit: Please consider this edit a low effort joke about how your second to last sentence describes the field of ML.

Edit 2: There’s actually a decent PyTorch tutorial using this algorithm, I highly recommend you use that tutorial to guide your understanding of the paper.

There are definitely going to be people here who are more up-to-date than I am. This is a mangled mess of how I first learned ML and how I am getting back into it.

NOTE: This isn't necessarily the best or easiest path. I am also mixing in my desire to learn functional programming into this

As others have said, if you want a deeper understanding of how ML works there isn't really a way around the math. I personally would suggest All The Mathematics You Missed But Need For Grad School. It isn't always perfect, but I found it's sections on Linear Algebra and Multivariate Calc to be pretty good, and it has a lot of exercises. If nothing else, its a great resource to find your weak points and help tune google searches for the areas you have problems with.

From there, you'll want to start learning about ML itself. I think just about everyone I know started off with this course by Andrew Ng. Its pretty solid, and I think most importantly is that it runs you through coding various machine learning algorithms. You don't use pre-built libraries to do all the ML for you. You will have to show that you understand step-by-step what each algorithm does and how. Really very useful.

After that, the world kind of opens up for you. I don't really know anything about the areas you listed as interests so I really wont be much help there. My next step personally is going to be learning Scala. I've been really liking functional programming as of late and so I'm going to be trying to get my hands dirty with it. Kind of a two-birds-one-stone situation, mixed with a moderate case of I-hate-python-itis.

Here is a youtube video also discussing best resources for self-taught learning from someone I generally respect: https://www.youtube.com/watch?v=-2AcCEZaCZM

I find this creator to be pretty good at explaining stuff:

Neural Networks Demystified

Learning to See

I think starting with the linear regression is always the way to go :

1. Generate some data (with noise)
2. Write down your model
3. Compute its gradient (by hand)
4. Fit the model via gradient descent with a mean square loss function
5. Plot the results

More explorations :

1. Split in testing and training set and explore overfitting (try higher order polynomials)
2. Explore how the final parameter values depend on the initial conditions
3. Explore how the parameter values depend on the realization of the noise
4. Implement another optimizers (add momentum, AdaGrad, ...)
5. Use a library to compute the gradient
6. Add asymmetrical noise and use a likelihood function instead of MSE
7. Add regularization
8. Compute the posterior on the parameters using Bayesian inference

For synths specifically, there's a great youtube channel that covers design rather than just assembly, by a guy called Moritz Klein.

In terms of digital electronics, Ben Eater's videos and associated kits are unparalleled.

I’ve done a few hobbyist electrical projects in my day, but by no means am an electrician.

That said, my advice to you is find something electronic and relatively simple, that you’ve considered buying. Dig around google for “(thing) diy” and related keywords. If a fair amount of results pop up, they’ve probably laid out each detail between all of them. Find one that seems to make the most sense to you, and get to work.

For your use case, I’d suggest an effects pedal or a preamp. These things are fairly simple and have plenty of diy guides available. Just one example, but there are plenty out there.

Best of luck on your ventures!