xnzakg

I guess they're the long term study subjects themselves at this point. honestly, maybe it's a positive thing? lots of testing across a variety of subjects, without all the ethics and paperwork slowing it down.

GLM-5.3-flash is open-weight though. you can run it on your own hardware for a few k$. yes, it will probably be slower, but you're no longer reliant on a subscription, a working network connection, or the servers not being overloaded.

I've had a very similar issue happening randomly. same pattern of the model outputting gibberish, seemingly trying to end its response and starting to panic when it couldn't, repeating a bunch of times.

turned out to be a bug in a patch, causing kv cache indices to be stored in too narrow variable types, making them wrap around.

fascinating, and slightly horrifying, that LLMs are able to notice that their output isn't what they meant to output. reminds me of the mirror test.

not a bad analogy imo. power tools vs hand tools. faster, but if you're not careful you can cause more damage before you notice. like a table saw, where you can cut too far or at an angle and the result will look bad. or an electric screwdriver, where you don't feel the torque as much, so it's easier to cross-thread.

and skills from building with hand tools help you judge the results, but not necessarily with actually using the tools.

Huh, almost as if the governments get something out of allowing the companies to abuse us...

Don't worry, the arbitrary software is probably running under Intel's ME or AMD's PSP.

If the kids are clever enough to have a second browser the parents don't know about, I'm pretty sure they'll find a way to bypass this. For example, who's stopping an older sibling giving remote desktop access to his browser?

It's the same as age limits on energy drinks, alcohol, tobacco etc. More of an inconvenience than a restriction if someone is determined enough.

We should focus on having parents teach their kids what they should and shouldn't be doing, instead of "blocking" them and pretending that solves the issue.

But in many cases, you could split it up into smaller, `inline` functions.

On the other hand the standalone models would all need "basic" knowledge, so I'm assuming this would be much less disk space-efficient.

How would this help against someone submitting an actual, non-compromised version bump, then adding malware once it's accepted?

Question is if this would be thorough enough for this attack? A package with a slightly more involved build process, maybe some patches because it was made to build on a different distro. Maybe you've already installed (and thoroughly inspected) it before, so you're only updating to a newer version, so you're not as thorough with your review. Or an xz-style backdoor.

> ultrasonic chirp authentication

Source? Can't find any good source on this.

I suppose using an LLM-powered autocomplete might be something for you then?

Then you can be writing the code, with the LLM doing the "boring" parts, in chunks small enough you can review them on the fly.

Local/open LLMs are a thing though. You can build a server for hosting decent sized (100-200B) models at home for a few k$. They may not be Opus-level, but hopefully we can get something matching current SOTA, but that we can run locally, before the megacorps get too greedy.

Alternatively you could find some other people to share the HW cost and run some larger models (like Kimi-K2.5 at 1.1T params).

The 2.13 level ("hex racer") is kind of pain. Apparently I'm not fast enough at dividing/multiplying by 16... when I get something like "convert 0xB3 to decimal"

You may wanna see if openrgb isn't able to configure the RGB. Could even do some fun stuff like changing the color once done with a training run or something

Sad to see the quality of the content from Lumafield slowly going down. Feels like the content got less technical after they moved it from "scan of the month" to their blog, and now it feels like the descriptions are just slop, not matching the scans themselves.

Examples:

- Each window is sealed under a thin polymer layer, balancing optical clarity with biocompatibility while preserving the ring’s continuous, scratch-resistant exterior. <- they're on the interior of the ring, and I'm assuming if they're polymer, they're not very scratch resistant.

- This flexible board architecture allows the Oura Ring to maintain its circular form as well as distribute heat [...] <- what heat? And how?

- The charging coil runs along the ring’s outer circumference [...] <- scan shows a small coil on the inside, not running along the circumference

- [...] we can easily visualize the deployment channel and insertion mechanism that guide this filament to its precise depth and angle. <- I can't see the insertion mechanism.

- spiral geometry <- the Bluetooth antenna isn't spiral. Nor does it communicate "through the user's skin"?

- miniature microphones (visible as small cylindrical cavities) <- they're rectangular.

The problem with the "your life objectively sucks" option is when you end up too depressed to actually bother doing anything and just give up. That's another case where drugs can help.

I don't fully agree with the "if something isn’t worth consuming at 1x, it’s not worth consuming at all" part. Sometimes you find a great video or podcast, but the person speaking simply speaks slowly. Or you've just watched/listened to someone who speaks quickly (like Louis Rossman for example), and gotten used to the speed.

Other times you just want to skim through the content, for example if you're already familiar with the topic, although you could argue that it's not really worth spending time on skimming something you're already familiar with.

But I definitely agree with the "quality filter" part. There's so much content out there that just doesn't have much substance to it.

> they are all reactive such that they must be given a task before they will do anything.

Isn't that just because that's what they're being trained on though?

Wonder what you would get if the training data, instead of being task based, would consist of "wanting" to do something "on someone's own initiative".

Of course then one could argue it's just following a task of "doing things on its own initiative"...

Neither this nor the blog post explains what this feature is actually supposed to do?

Also interested in this, but haven't tried it myself. Apparently NSF (Nano Silver Fluoride) is more commonly in non-western countries.

They have a video with some more info here: https://pt.fourthievesvinegar.org/w/9aa66b49-2ec5-497f-9f49-...

And apparently the use of NSF does have a bunch of research papers written about it: https://www.researchgate.net/profile/Amol-Patil-43/publicati...

This one? https://youtu.be/_goIYP3FfO8

...or you can just give someone a "smart" device that requires then to install an app with lots of unnecessary permissions to use it.

> Tastes like bullshit to me.

Does it really? In my opinion, if it stops working and it's under warranty, why not send it out for repair? They did no changes to the actual device, and apparently it was working fine for a few days without network connection, so if it suddenly stops working and it's under warranty that's the manufacturer's/store's problem, not theirs. Trying to fix it/reverse engineer it takes time, and I can see someone with these kinds of skills wanting to spend it on something else than trying to figure out how the manufacturer bricked their vacuum.

In addition, _someone_ is paying for the repairs under warranty, so if enough people were to do it, hopefully it would disincentivize completely blocking devices just because they can't reach a server.

You can always put some extra protection on the external interfaces. Won't make it impossible to fry if you really do something stupid but would reduce the risk significantly.

Hasn't stopped a lot of companies from selling out and then being ran into the ground by their new owners.

Seedboxes often have TBs of storage for relatively cheap.

Roon seems great but the pricing is really steep in my opinion... Costs practically as much as a streaming service, but you still need to get your own music.

At least they have a lifetime purchase option, though it costs $830!

Technically that's four cores, two of which that can be active at once