Zelizz

In high school, I once volunteered at a local library for a few hours, sorting returned book alphabetically. At the start, I tried just grabbing the alphabetically "first" book repeatedly, moving it to the front. This got old fast, so after a little while, I started looking at just the first letters, and grabbing all the books I could fit in my hand with that letter. When I had them all "roughly" sorted, I did the same with the second letter, etc.

A few years later I learned about sorting algorithms. It's interesting to me that my natural book-sorting intuition was O(n^2) selection sort, but then it wasn't a very big leap for me to tweak it a bit and discover a form of O(n)(ish) radix sort.

In school you don't really spend a few hours manually executing sorting algorithms to build an intuition, so the algorithms can feel a bit like artifacts handed down from the gods. How could someone have come up with these? But then, if that's the specific problem you're trying to solve, developing a fairly efficient and sophisticated (and sometimes opaque to future students) algorithm can feel completely natural.

I think there's free alcohol on international flights, but you pay for it in domestic flights.

It's she, not he: https://twitter.com/b0rk

It's she, not he.

I work on Windows (speaking informally here, these are my own opinions and not those of my employer, etc.)

We care a lot about GDPR and privacy, and there's been a big top-down push to reduce the amount of telemetry collected, including giving good justifications for anything we want to keep collecting. We have to make sure all the data is kept in GDPR-compliant systems that can handle the deletion and security policies, and we regularly take training to tell us what is and isn't compliant use of customer data. My team uses it mostly for seeing the success rate of our feature across different hardware, firmware and OS versions, which helps us identify issues and poke certain partners to improve ecosystem reliability. I can't speak for the Edge/Bing/Whatever teams, but I'd hope they take it as seriously as us.

I definitely wish users had more control over data collection without resorting to group policy, and I wish users had more visibility and control over more of the OS in general, but I feel we've learned the wrong lessons from Apple so everything must be unconfigurable and only give vague feedback like, "Something is happening..." (but I digress). In terms of caring about privacy I'd still give Apple the crown, but I don't think it'd be unfair to give Microsoft second place.

And Love2D, which is free and open-source, also using Lua.

I'm constantly turning my Netflix profile language to Korean so that I can get Korean subtitles for my wife. It's an annoying workaround, but it often works.

I've found it affects how sharp lines look for me on Quest. I like to find a vertical line somewhere and dial in until it looks decently sharp.

That's unreasonably cool. Do you think it consumes more resources than if they just gave you more than 64K to work with, or do you think you were pushed to optimize further because of the limitations?

People usually use it because piped input is a known interface, you don't have to go digging through a man page or help page to figure out what flag is needed to accept input from a file. "Data flows from left-to-right" is also very intuitive, and you don't need to know the specifics of how your shell interprets a command line to figure out where to put the input redirection in the case of a more complicated shell command.

Aren't you effectively asking the compiler to solve the halting problem?

I think the best you could do would be heuristics - having inferred or user-supplied bounds on the complexity of functions, having rough ideas on how disk or network latency will affect the performance of functions, and bubbling that information up the call tree. It wouldn't be perfect, but it could be useful.

The article does mention that it's more durable than vantablack, so perhaps more suitable for a space-based project (if it turns out a really black material is useful for a solar panel of sorts).

The biggest problem, in my opinion, is that there appear to be two settings, one for trackpad, and one for a mouse scrollwheel. You can find them in different locations. They change the same value, though. You can't have natural trackpad scrolling and "pull the wheel towards you to scroll down" scrolling at the same time without a 3rd-party application.

The Medium article says that was intentional:

> That gives the elevation spread of about 1.15% of the mean radius. I’ll use a bit higher value of 2%, so the terrain features are more prominent.

You'd need to emulate/simulate/shim all the graphics calls and state changes, but that shouldn't have any bearing on the actual code architecture. In fact, given that Dolphin uses a JIT, you could argue that this already happens to some degree when you're playing Gamecube games, having the source just allows ahead-of-time compilation.

Is it really more power-efficient? CPUs are often a smaller process node than integrated GPUs and it's much easier to share common work between pixels with CPU rendering, reducing redundant operations.

You might enjoy this video on climate change: https://www.youtube.com/watch?v=RLqXkYrdmjY

What you describe is a natural consequence of adding the side-to-side velocity to the forward-and-back velocity. It is simple to implement, fast to run, and would be completely correct with something like a spaceship, where velocity components would naturally sum up like that.

  velocity.x = (RIGHTKEY-LEFTKEY)*speed;
  velocity.y = (BACKKEY-FORWARDKEY)*speed;

The constraint it doesn't capture is that humans usually run about the same speed in any direction. To capture that constraint, you either normalize the xz components and multiply by speed or do some extra logic when perpendicular directions are input. Either way, it's slower, more complicated, and harder to get right. If it doesn't break the game, why bother?

What you are describing is a coilgun, a railgun uses a slightly different principle.

I think they were suggesting that you could save fuel for once you're out of atmosphere, so that you're not fighting air resistance when you start a burn to get into orbit.

0. Pay more.

They also considered only Lua 5.1, when Lua is on 5.3 already (and soon to be 5.4).

Is this base salary or total compensation?

I think it’s not unusual for any program to be 30% faster than the program it replaces. So far I haven’t seen any specific examples of the C++ or Rust compiler providing a speedup by leveraging additional information input by the programmer. The main speedup seems to be the ability to easily implement more complicated algorithms, and the ability to speedily replace slow one-off implementations with relatively high-quality standard library ones.

How often are you having to hand over your phone to customs? I've literally never been asked to do so.

There's no reason for them to even tell us that they're using Tor, except to raise the specter of some retaliatory bogeyman.

Someone should port this to C and add it to the next version of Lua, given that Lua's main datatype is pretty much a hashtable with some extra magic.

When doing an interview at Google in C, I asked if I could assume I had a hashtable implementation with so-and-so interface, and the interviewer said no ¯\_(ツ)_/¯

Right, I know that you’d want to use radar for longer distances, but if the towed decoy is meant to appear as a juicy target, to take a hit for the fighter, couldn’t you just follow it until you’re in close range, and then switch to a better discriminating sensor (visible light camera in my example) to hit the jet instead?

How well would a towed decoy be able to fool a missile equipped with a camera and a neural net trained on pictures of fighter jets? Are the distances and speeds too extreme for a visible light camera? Is there some technology to "blind" a visible light camera?