I think you’re spot on with LLMs being mostly trained on these kinds of tasks. Can’t say I’m an expert in how to build a training set, but I imagine it’s quite easy to do with these kinds of problems because it’s easy to classify a solution as correct or incorrect. This is in contrast to larger problems which are less guided by algorithmic efficiency and more by sound design/architecture.

Still, I think it’s quite impressive. You don’t have to go very far back in time to have top of the line LLMs unable to solve these kinds of problems.

Also there is no big consequence if they don’t and it’s probably possible to bruteforce (which is how many programming tasks have been solved).

Usually with AoC part 1 is brute-forceable, but part 2 is not. Very often part 1 is to find the 100th number, and part 2 is to find the 1 000 000 000 000th number or something. Last year, out of curiosity, I had a brute-force solution for one problem that successfully completed on ~90% of the input. Solution was multi-threaded and running on a 16 core CPU for about 20 days before I gave up. But the LLMs this year (not sure if this was a problem last year) are in the top list of fastest users to solve the problems.

I hardly see it changed to be honest. I work in the field too and I can imagine LLMs being good at producing decent boilerplate straight out of documentation, but nothing more complex than that.

I think one of the top lists on advent of code this year is a cheater that fully automated the solutions using LLMs. Not sure which LLM though, I use LLMs quite a bit and ChatGPT 4o frequently tells me nonsense like “perhaps subtracting by zero is affecting your results” (issues I thought were already gone in GPT 4, but I guess not, Sonnet 3.5 does a bit better in this regard).

oh, I thought he didn’t care?

Oh, I thought only cats did that!

Didn’t the refactored netscape eventually evolve into Firefox though? Not disputing the poster child status or the fact that it’s a terrible business decision, but the project did not really go stale I think?

What’s IVF in this context?

Never knew it, very neat!

How do I use this feature? I’m a Firefox user since quantum and had no idea this was a thing.

Just write some content without no soul and not a shred of humanity present. I.e. use the platform as intended.

This is so cool, first MQTT-based sensor I’ve set up. Already had a broker set up with HA, but how can HA automatically discover which topic to listen to, know the vendor name and how to interpret all the data?

Interesting, so I guess those API-calls are just fetching the cached calendar on my HA Yellow. Wonder why it’s so slow, but I guess there’s not much to do about that then. :(

Not exactly. My main use-case here is for my girlfriend and me to see each both of our calendars in one place, and HA had support for it and is a web portal we both have access to. To do automations on them is secondary.

Currently, whenever I look at the calendar control panel it will load for a bit while pulling all the calendars, and sometimes timeout and not show anything. I believe this to be because it’s pulling from Fastmail / iCloud everytime and might be rate limited or just have a poor connection, this wouldn’t be an issue if the calendars were stored on the instance itself because then it would only miss the latest entries.

The idea that maybe I can self-host an app that does it is that if HA can’t do the caching, then maybe this self-hosted app can and it wouldn’t matter that HA fetches it remotely each time since the remote is on the same local network. Having them as separate calendars is still desirable since that gives some additional information.

They already have something kinda like this. All public wifis require that you sign in with phone number and SMS-verification. It might not be as air-tight as whatever the article is about (like a chad I only read headlines), but in practice it seems pretty darn tight IMO.

Is immich in a usable state yet? I was looking for a self-hosted image service a while back, but eventually I just went with pigallery2 mostly due to the extremely simple file storage (just point to a folder and you’re good to go), but I do miss being able to manage images/albums from the website and having a more mobile friendly version. I kind of avoided immich due to the repo saying it’s under very active development (#scary).

Back in the day, before streaming was a thing, there were lots of people saying that they’d gladly pay for content if it was served to them in a convenient way. But why would you pay for a worse experience (at that time physical media, often at lower quality, and lower availability) when you can get a better one for free?

Along came streaming. Lo and behold, piracy decreased. Where the fuck do you even go to pirate music anymore? All the big sites have shut down. Video piracy is kinda still going strong, probably mostly due to bullshit concerning exclusives, but it’s way less than it used to be.

Its their platform, they can do whatever they want with it I guess, but this trend is definitely gonna be a big boost to piracy.

Isn’t it a local filesystem though, so I can’t expand the filesystem with other drives on my network?

That’s very helpful because glusterfs and ceph are probably my top two candidates. Will probably try it out.

Could be difficult. Name is a bit stained.

Why not?

the biggest selling point for me is that I’ll have a mounted folder or two, a shell script for creating the container, and then if I want to move the service to a new computer I just move these files/folders and run the script. it’s awesome. the initial setup is also a lot easier because all dependencies and stuff are bundled with the app.

in short, it’s basically the exe-file of the server world

runs everything as root (not many well built images with proper useranagement it seems)

that’s true I guess, but for the most part shit’s stuck inside the container anyway so how much does it really matter?

you cannot really know which stuff is in the images: you must trust who built it

you kinda can, reading a Dockerfile is pretty much like reading a very basic shell script for the most part. regardless, I do trust most creators of images I use. most of the images I have running are either created by the people who made the app, or official docker images. if I trust them enough to run their apps, why wouldn’t I trust their images?

lots of mess in the system (mounts, fake networks, rules…)

that’s sort of the point, isn’t it? stuff is isolated