Thinking about starting a new project that I might be able to convert into a startup later, but also feeling pretty deflated about startups these days. Without millions in VC funding it’s really hard to get anywhere or compete with anyone.
The tool I am planning to build will solve some of the problems with Helm and kubectl, as well as CI/CD problems with k8s deployments. It’s been a pet peeve of mine for a while. Helm leaves much to be desired.
Yep but then you need to have a lot of money saved up, or some other source of income. I am poor (though I have a bunch of illiquid and practically worthless startup stock).
The way the current machine learning revolution was explained to me was simply that “We stopped trying to manually extract features, and instead made the models work on raw data. It turns out with these statistical methods we can find a lot of useful features on dimensions we would never even think of encoding into our manual feature extraction.”
This article is basically a conclusion of that description. Very interesting! It really highlights how arbitrary human classification processes can be – further emphasized by the amount of re-labeling that happens constantly in biology.
Yep! I think that’s a really concise and useful summary. A downside is you lose access to explicit/interpretable feature representations, but you gain potentially much better performance while removing lots of manual labor.
Not strictly necessary. The title “built using HTML and CSS 3D transformations” followed by that pile of Javascript might lead some to call bullshit on the title.
The js is needed to move the camera with the keyboard and mouse, since css can’t react to user input like that. Looking over the rest of the code, it seems like it caches the textures with a canvas an applies them to elements as dara-uri (probably to avoid posible multiple parallel requests on load), but the rendering is done with inline styles.
A mix of HTML, CSS, and Javascript with more of the first two than usual. On DynamicDrive, there’s lots of menus and stuff that are pure HTML and CSS. I just like to differentiate.
Not reading it closely, but it seems like a good bit of it is defining the styles and loading elements onto the page. I think the idea is that this made it possible to write modular functions for creating different objects, rather than copying CSS and HTML all over the place. There’s some more info here.
I haven’t gotten to spend as much time playing with zig as I want to yet. But I just wanted to say that it’s really inspirational to see a project like this moving forward.
Very easy to read, but is this really PLT? Feels more like PLI (Programming Language Implementation), a useful study, but it doesn’t quite prepare you to discuss the fringes.
An excellent PLT introduction is Iverson’s Notation as a Tool of Thought which can really get you thinking about the use and value of good notation (syntax).
Another piece I like recommending is Moseley and Marks Out of the Tar Pit which correctly identifies the most valuable area of research in programming language theory is addressing complexity.
If these feel too advanced, try printing them out and reading them away from your computer.
There are a lot of good resources once you are strong with computer science fundamentals, and I appreciate the want to skip ahead and jump right into the design and implementation of the software you’re dreaming of, but you’ll get much more out of it when you’re a better programmer.
Finally: The decision to bring a new programming language into the world is not one to be taken lightly. We have quite a lot of them already, and I’m of the opinion that study of other - stranger languages is a better use of the young journeyman’s time.
I agree about the naming, but at least in my (albeit limited) experience I think implementation can actually be the most gentle entry point to theory. I like learning about PLT because it’s fun, and I’m not sure it would have seemed that way if I started off with the theory itself. Since then, theory has become more digestible. But the rush of having made my own language that did something cool and nontrivial was a big part of me making it there.
That is a really good point. Building something cool is motivational, so maybe add some next steps to dig into the real theory stuff? Help build why they just did things the way they just did it (and not some other way) might help get them thinking the right way.
By this point, you might be thinking that the obvious solution is to just disable push notifications entirely.
After a few days, I realized that my brain had become so accustomed to my phone telling me what I needed to attend to, that I felt lost when I didn’t have those cues.
How long did you try this for? I’ve completely disabled push notifications on my computer for a few months, and I think over time I’ve ended up compulsively opening email/slack/etc. less. What was important was not just removing the notifications, but other visual cues (for example, having my dock always visible on macOS, which would display little red icons). On the other hand, I think I would have little success if I tried to do this on my phone, so I don’t. I view my phone as less of a working tool anyways, so I’m okay with this. Most importantly, when I’m trying to focus I can reasonably restrict my access to my phone, but less so with my computer.
I’ve personally struggled more with the idea that Something Very Important could have happened since I last checked in. The set of things that trigger this are much smaller, but they exist enough to be a problem at least every few days. For example, when I’m anticipating a results of an exam that I don’t think went particularly well, I tend to open email and check every time my mind drifts. I don’t think consuming content purely in digest form would fix this. Perhaps aggressive filtering would work, but I haven’t tried this and am generally suspicious of “productivity/mindfulness solutions” that require lots of active effort.
I only tried it for about five days. Perhaps a longer experiment is in order. I’ve also done a lot of visual decluttering on my computer, e.g. removing my bar, notifications, etc.
I can attest that both a minimal computing environment and disabling notifications have helped me transition out of a time when I was feeling similar to how you’ve described.
Author here. Although the raw material I used is <10$, as a rule for all DIY projects you should always consider the time and equipment needed. I’ve seen some motorized shades for ~200$, which for a finished/professional product is quite ok.
Thinking about starting a new project that I might be able to convert into a startup later, but also feeling pretty deflated about startups these days. Without millions in VC funding it’s really hard to get anywhere or compete with anyone.
The tool I am planning to build will solve some of the problems with Helm and kubectl, as well as CI/CD problems with k8s deployments. It’s been a pet peeve of mine for a while. Helm leaves much to be desired.
On the other hand, without VC funding you don’t have VC pressure and (possibly) VC-misguidance :)
Yep but then you need to have a lot of money saved up, or some other source of income. I am poor (though I have a bunch of illiquid and practically worthless startup stock).
The way the current machine learning revolution was explained to me was simply that “We stopped trying to manually extract features, and instead made the models work on raw data. It turns out with these statistical methods we can find a lot of useful features on dimensions we would never even think of encoding into our manual feature extraction.”
This article is basically a conclusion of that description. Very interesting! It really highlights how arbitrary human classification processes can be – further emphasized by the amount of re-labeling that happens constantly in biology.
Yep! I think that’s a really concise and useful summary. A downside is you lose access to explicit/interpretable feature representations, but you gain potentially much better performance while removing lots of manual labor.
Tangentially related is this idea.
Javascript also plays a big role there.
Is the giant paste of code really necessary?
Not strictly necessary. The title “built using HTML and CSS 3D transformations” followed by that pile of Javascript might lead some to call bullshit on the title.
If you open the non-interactive demo (https://keithclark.co.uk/labs/css-fps/nojs/) it is only html and css.
The js is needed to move the camera with the keyboard and mouse, since css can’t react to user input like that. Looking over the rest of the code, it seems like it caches the textures with a canvas an applies them to elements as dara-uri (probably to avoid posible multiple parallel requests on load), but the rendering is done with inline styles.
A mix of HTML, CSS, and Javascript with more of the first two than usual. On DynamicDrive, there’s lots of menus and stuff that are pure HTML and CSS. I just like to differentiate.
Not reading it closely, but it seems like a good bit of it is defining the styles and loading elements onto the page. I think the idea is that this made it possible to write modular functions for creating different objects, rather than copying CSS and HTML all over the place. There’s some more info here.
I haven’t gotten to spend as much time playing with zig as I want to yet. But I just wanted to say that it’s really inspirational to see a project like this moving forward.
Very easy to read, but is this really PLT? Feels more like PLI (Programming Language Implementation), a useful study, but it doesn’t quite prepare you to discuss the fringes.
An excellent PLT introduction is Iverson’s Notation as a Tool of Thought which can really get you thinking about the use and value of good notation (syntax).
Another piece I like recommending is Moseley and Marks Out of the Tar Pit which correctly identifies the most valuable area of research in programming language theory is addressing complexity.
If these feel too advanced, try printing them out and reading them away from your computer.
There are a lot of good resources once you are strong with computer science fundamentals, and I appreciate the want to skip ahead and jump right into the design and implementation of the software you’re dreaming of, but you’ll get much more out of it when you’re a better programmer.
Finally: The decision to bring a new programming language into the world is not one to be taken lightly. We have quite a lot of them already, and I’m of the opinion that study of other - stranger languages is a better use of the young journeyman’s time.
I agree about the naming, but at least in my (albeit limited) experience I think implementation can actually be the most gentle entry point to theory. I like learning about PLT because it’s fun, and I’m not sure it would have seemed that way if I started off with the theory itself. Since then, theory has become more digestible. But the rush of having made my own language that did something cool and nontrivial was a big part of me making it there.
That is a really good point. Building something cool is motivational, so maybe add some next steps to dig into the real theory stuff? Help build why they just did things the way they just did it (and not some other way) might help get them thinking the right way.
For anyone interested, Jonathan Blow also streams some of the development and posts recording to youtube here!
How long did you try this for? I’ve completely disabled push notifications on my computer for a few months, and I think over time I’ve ended up compulsively opening email/slack/etc. less. What was important was not just removing the notifications, but other visual cues (for example, having my dock always visible on macOS, which would display little red icons). On the other hand, I think I would have little success if I tried to do this on my phone, so I don’t. I view my phone as less of a working tool anyways, so I’m okay with this. Most importantly, when I’m trying to focus I can reasonably restrict my access to my phone, but less so with my computer.
I’ve personally struggled more with the idea that Something Very Important could have happened since I last checked in. The set of things that trigger this are much smaller, but they exist enough to be a problem at least every few days. For example, when I’m anticipating a results of an exam that I don’t think went particularly well, I tend to open email and check every time my mind drifts. I don’t think consuming content purely in digest form would fix this. Perhaps aggressive filtering would work, but I haven’t tried this and am generally suspicious of “productivity/mindfulness solutions” that require lots of active effort.
I only tried it for about five days. Perhaps a longer experiment is in order. I’ve also done a lot of visual decluttering on my computer, e.g. removing my bar, notifications, etc.
I can attest that both a minimal computing environment and disabling notifications have helped me transition out of a time when I was feeling similar to how you’ve described.
This is really reassuring to hear, thank you :)
Cool article, but not what I expected from the phrase window manager!
In hindsight, I guess the tag should have given it away. But I enjoyed the surprise as well :)
Missed the hardware tag too. Been looking at electric blinds recently and the off the shelf ones seem prohibitively expensive.
TechCrunch Disrupt 2019: “Electric Blinds Meets Machine Learning Monitored and Controlled via a $5 Droplet!”
Author here. Although the raw material I used is <10$, as a rule for all DIY projects you should always consider the time and equipment needed. I’ve seen some motorized shades for ~200$, which for a finished/professional product is quite ok.
It doesn’t even support ICCCM!