Yup. As a kid I could "entertain" (distract is the better word) myself by "drawing shapes" with the cursor, highlighting random things, switching between random cells in Excel, or just like... browsing through the system without any plan or reason. Procrastination is hell of a drug.

I'm so lucky I didn't have this in the classroom.

To be fair, I did entertain myself by drawing comics on my notebook or playing with my pencil and rubber as if they were toy cars.

I drew a lot of doodles and did things like that as well, but I think that they're less visually stimulating and simply "slower" so there's still some brain capacity left for learning.

Congratulations, you were exercising your literacy and art skills.

Wait, what, Zed is Tauri? How? One of their main things was that they implemented the UI layer completely from scratch using their own GPU-accelerated rendering engine. It's got none of that browser-type stuff.

To extend on that a little bit: they use data centers located in EU, but owned by US cloud providers. They can still pull the plug ofc, so it's only a small difference, but still

Sort of, they have no "hands", LLMs can only respond that they want to execute a tool/command. So they do that a lot to: read files, search for things, compile projects, run tests, run other arbitrary commands, fetch stuff from the internet etc.

Obviously the LLM inference is super heavy, but the actual work / task at hand is being executed on the device.

Models are not local most of the time, no, but all commands execute on "the mac mini" so I wouldn't exactly call it a prop. LLMs accept and respond just with text what stuff to execute. They have no h̶a̶n̶d̶s̶ claws.

But that could just as easily run on an EC2 instance, or in Azure cloud? The only magic sauce is they've set up an environment where the AI can run tools? There's no actual privacy or security on offer.

Yeah, pretty much. A "mac mini" is just easier to set up for the average hype-driven AI "entrepreneur" bro than anything on the cloud. It's mostly a meme though.

This happens to non-native English speakers a lot (like me). My style of writing is heavily influenced by everything I read. And since I also do research using LLMs, I'll probably sound more and more as an AI as well, just by reading its responses constantly.

I just don't know what's supposed to be natural writing anymore. It's not in the books, disappears from the internet, what's left? Some old blogs for now maybe.

The wave of LLM-style writing taking over the internet is definitely a bit scary. Feels like a similar problem to GenAI code/style eventually dominating the data that LLMs are trained on.

But luckily there's a large body of well written books/blogs/talks/speeches out there. Also anecdotally, I feel like a lot of the "bad writing" I see online these days is usually in the tech sphere.

Books definitely have natural writing, read more fiction! I recommend Children of Time by Adrian Tchaikovsky

I'm a firm believer that almost nothing except public services needs that kind of uptime... We've introduced ridiculous amounts of complexity to our infra to achieve this and we've contributed to the increasing costs of both services and development itself (the barrier of entry for current juniors is insane compared to what I've had to deal with in my early 20s).

What do you mean by public services?

All kinds of companies lose millions of dollars of revenue per day if not hour if their sites are not stable.... apple, amazon, google, Shopify, uber, etc etc.

Those companies have decided the extra complexity is worth the reliability.

Even if you're operating a tech company that doesn't need to have that kind of uptime, your developers probably need those services to be productive, and you don't want them just sitting there either.

By public services I mean only important things like healthcare, law enforcement, fire department. Definitely not stores and food delivery. You can wait an hour or even a couple of hours for that.

> Those companies have decided the extra complexity is worth the reliability.

Companies always want more money and yes it makes sense economically. I'm not disagreeing with that. I'm just saying that nobody needs this. I grew up in a world where this wasn't a thing and no, life wasn't worse at all.

Eh, if I'm paying someone to host my git webui, and they are as shitty about it as github has been recently, I'd rather pay someone else to host it or go back to hosting it myself. It is not absolutely required, but it's a differentiating feature I'm happy to pay for

Most likely a Windows or MacOS user, where docker runs in a linux VM. Optimized as much as possible and lightweight, but still a VM.

No, on Windows it is very quick too. On WSL2 compiling Rust programs are almost as fast as Linux on bare metal. However the files need to live inside the Linux filesystem. Sharing with Windows drives actually compiles slower than native Windows.

You can use dev drives instead, I guess.

If you are building natively, yes. However the original comment is about Dev Containers which runs under WSL2.

If you open a native Windows folder in VSCode and activate the Dev Container, it will use the special drvfs mounts that communicate via Plan9 to host Windows OS to access native Windows files from the Docker distro. Since it is a network layer accross two kernels, it is slow as hell.

That is the beauty of Plan 9's design. :)

I haven't tried, but the idea was to map a such a drive into WSL, but I am not sure if it is possible and indeed how much it would help in the end.

You can mount raw VHDX files even raw normal bare metal drives in WSL2, that's true.

However, Linux needs to understand the FS inside.

Windows is a bit "yes but" kind of situation.

First of all it supports containers natively, Windows own ones, and Linux on WSL.

Secondly, because Microsoft did not want to invent their own thing, the OS APIs are exposed the same way as Docker daemon would expect them.

Finally, with the goal to improving Kubernetes support and the ongoing changes for container runtimes in the industry, nowadays it exposes several touch points.

https://learn.microsoft.com/en-us/virtualization/windowscont...

So if you're directly embedding the thing. This is a somewhat rare use case, should not be banned almost anywhere...

The thing is, if you want AI output to be heavily directed, which is probably the case here, I can imagine that thousands of random takes had to be made to make the damn thing follow the director's imagination. If you don't care too much about the output you can make these very quickly, yeah.