snailmailman

From what I understand the Bluetooth is required to ensure proximity. It isn’t actually used to transmit anything of vital importance. The QR is doing most of that. It’s just there to prevent scanning the QR through a video call and logging in from elsewhere.

Which I’m sure is great in theory. But IMO just adds even more complexity to a system that already has several moving parts and is more fragile than it should be.

This is good in theory but in practice doesn't always work. It isn't just a QR code like I would like it to be, its a QR code and you need bluetooth. Maybe there is a hypothetical world where bluetooth drivers actually work on windows machines and can connect to a mobile device seamlessly but that is not my experience. Across multiple windows machines i often have a problem where windows just decides the machine doesn't have bluetooth, and will refuse to display the QR code without any explanation. It doesn't tell you that bluetooth is the problem it just says "insert a hardware token to authenticate".

I would love for this feature to actually work but every time I've needed it to it hasn't. Literally this week I only had a passkey on my phone, but at the time I was in Linux with Firefox, and afaict the qr code workflow basically requires either chrome or windows 10.

An anonymous answer to a question is more trustworthy to me. They have no reason to lie. They are usually answering out of kindness. At least they used to be. Now it’s often actually bots advertising a product or pushing something, pretending to be a helpful user with an anecdote and a good experience using a niche product.

AI shouldn’t have any reason to lie. But its lies aren’t intentional. It’s just actually making things up and “hallucinating” when it pretends that an option or setting exists, or confidently claims something entirely untrue, and makes up a source to go with it. For something Google is willing to shove into the top of every search result it’s crazy the percentage of time the answer is blatantly incorrect.

Apartment advertisements are already doing this. With the caveat that the sense of scale (and literally everything about the apartment space tbh) can be entirely tampered with.

Oh yeah, this tiny apartment can definitely fit all this furniture. It’s not all ai generated at 2/3rds the size of any real furniture.

Weird that it’s inconsistent

Works fine for me in desktop Firefox. But on mobile iOS the “whole video within a jpeg” is 3 frames, all of which are nearly entirely solid color brown->orange->red with a vague cat silhouette. The color changes each frame, so you can tell it’s “working” but it’s certainly not what I’d call a video.

I was surprised when I pulled it up on my desktop and it did actually play like a video. Wonder if it’s causing some weird iOS image decoding glitch/edge case.

I’ve always been under the impression that Signal is for secure, private chats and group chats amongst friends and small groups. While telegram is often used more like discord or irc, with “secure” and “private” group chats that are extremely large and semi-public. “Invite only” but invites are handed out easily. Those chats are pretty obviously not as secure, as basically anybody can join them and decrypt the chat. On the surface its somewhat more secure than discord, where discord will be scanning all chat content.

I asked “Alexa+” (the new AI/LLM Alexa), what time one of the upcoming World Cup games is. It was off by two hours due to time zones. This was a big enough mistake that I caught it, and followed up with “I’m pretty sure that’s wrong” and then it tried to correct itself and gave me a new time that was off by hour in the opposite direction.

I guess my time zone wasn’t in the context? So it just hallucinates the wrong coastal time zone twice when I’m not in either one. But who knows where exactly it messed up because it could have just picked a random hour and a would have had the same outcome.

I’ve only used Alexa a half dozen times since the release of Alexa+ but it has been confidently incorrect about 100% of the queries.

I think it’s been functionally login-only for a while, unfortunately.

Any time I try to use the site while logged out I immediately hit all sorts of rate limits and spam prevention measures. Issues and pull requests stop loading almost immediately. Searches are extremely restricted, like i think literally one search is all you get sometimes unless you login.

It’s hard to tell how much of this is intentional and how much is GitHub infra not being stable. My daily renovate bot that looks for updates for specific packages times out and encounters errors almost every time it runs and it’s barely doing anything. And it has an api key. I know I’m not hitting limits, some queries just don’t work >50% of the time on some repos.

Sometimes I see there are 200 issues, I click the issues tab, and there are zero. Is GitHub down? Am I rate limited? Do I just need to login? Or I know an issue/PR exists, and it just doesn’t appear when I search for the title or keywords verbatim. I just manually bookmark or subscribe to any important issue or PR now because the UI can’t be trusted anymore.

Its a fun challenge.

I used https://squoosh.app to make a pretty good one. Mostly just a resize and then OxiPNG for compression. Managed a 124x62 black/white image. OP has a resolution of 195x53, so I had very similar, but slightly worse i think? mostly a different aspect ratio + map projection i think.

playing with Squoosh.app is very fun, and you can very easily see how the jump from 500b to 1-2kb turns a map from "awful" to "very good" with the right settings.

I think the comment can be left on any video on the channel?

Assuming this is what you are referring to, it was resolved within a few days. The incident being resolved just didn't make headlines. https://sourceforge.net/p/veracrypt/discussion/general/threa...

With fast boot, it is a partial-hibernate (as you said, it is basically logging out then hibernating). But also, you are shutting the device down. You can hibernate a laptop and take the battery out. The device is off. Thats the difference between sleep and hibernate. Sleep still uses power.

Hibernate is a true power off, with instructions for the PC to load ram from disk on next boot. IME hibernate often works more reliably, because sleep states can be buggy, and hibernate just powers off the device. Hibernate is slower though, and can have its own issues (often related to disk encryption)

Sleep being buggy is a mystery to me, but has happened on nearly every laptop i've owned. I think too many things can wake slept device or prevent sleep. Leading to "I closed my laptop and put it in my bag. Then it woke up and got hot and span the fans until it died". But ive also had many instances where some apps are more buggy after waking from sleep (ive had issues like video playback failing after wake, presumably related to hardware video decode), or certain peripherals don't handle it well, and windows reshuffles monitors or changes a microphone since the "default" was briefly gone, or a mouse/keyboard doesnt wake the device from sleep when they should, etc. Phones and tablets seem to handle low power states really well, but ive literally never had a good experience with it on a laptop or desktop.

Yeah, i just simply don't trust the sleep functionality to not be buggy. I walk away with processes running a decent amount of the time and its unclear to me what things can prevent sleep from triggering and what cant.

If i do need to have something persist overnight I use hibernate. And I haven't encountered issues with that. But ive had enough issues with sleep states on my laptops that i just dont bother on my desktop. The monitor goes to sleep if i walk away but the PC doesn't sleep.

I'm confident this will mostly fix my problem but its such a minor thing its not worth disassembling my PC for. But I will likely swap it out next time I disassemble it for any other reason.

It seems to me like this problem can be entirely solved in software if the OS more frequently resynced the clock. I checked just now and it last did a sync 2 days ago? I dont think my PC bios even saves sub-second accuracy, and IMO it should be resyncing after every boot.

The CMOS battery on my ~10yr old linux laptop is also dead, and its a complete non-issue there. The bios complains occasionally and shows me the time is wrong and then i boot in and the OS fixes the clock immediately.

I shut down my desktop every night. No reason to keep it on. I can save my work, all my tabs persist, it’s fine.

My main complaint is that my PC seems to have crazy clock drift? And windows just doesn’t seem to care? It doesn’t actually cause problems but I will check and see that it’s off by >5 seconds and that just shouldn’t ever happen. My phone is always accurate to <0.05s and same in Linux. Idk why windows isn’t.

I definitely had some cheap PC speakers in the mid 2000s that worked as a shitty microphone if you plugged the cable into the microphone port instead of the speaker port. You basically had to yell into them, and the sound quality was atrocious, but it did ‘work’.

A surprising amount of circuits work like this. Generators and motors are basically the same as well, and spinning a fan by hand will induce current into the wire. I assume most circuits are designed to properly handle that happening, but I still worry when blowing the dust out of my PC and will manually prevent the fans from spinning when I blow compressed air at them.

I run an instance of smokeping locally for this purpose. It pings a variety of DNS servers (including my ISPs DNS) and several of the top websites. I periodically update my local DNS server’s upstream accordingly.

All the big DNS servers are in the 5-6ms range for me, but that hasn’t always been the case. My ISPs DNS is about the same but with crazy variance and spikes of up to 50ms, even though they should be able to be the fastest.

This is the real strength of ActivityPub compared to ATProto. I follow a lot of accounts on mastodon, and only a handful of them are on mastodon.social. Globally though looks like 1/4 of the accounts are there.

If the instance goes down, the network stays up and continues to work, minus only the accounts on mastodon.social. This is not the case on bluesky afaik. They got DDosed a few months ago and the whole network was down because of it. https://news.ycombinator.com/item?id=47802330

In ATProto there aren't 'instances', and its technically 'decentralized' but theres really only bluesky.

They have an “all in one” container that supposedly works out of the box.

I didn’t want to give it access to the docker socket, with the ability to spawn its own containers. So instead I just use the nextcloud container directly. (With several other containers, like DB, reverse proxy, collabora, etc) It’s a mess to configure, hence their recommendation to use their “all in one” setup. All sorts of weird defaults with documentation that says “this is the default but you should absolutely change it to do X instead so that it performs better”. Things like setting up a service to generate thumbnails, setting up redis, etc.

Once configured though, it mostly just works. You can’t let it auto update between major versions, but you probably should be doing that anyway. There are usually breaking changes and you have to manually run a command or two between major updates. That doesn’t happen too frequently though.

I can’t speak to the quality of the all-in-one setup. It’s likely easier than what I did - but also what’s the point of putting it in docker and also giving it control of docker? Seems to defeat the point of containerization.

I use nextcloud all the time, my private instance works great and does everything I need it to. But I keep it behind a VPN. It’s got a lot of parts, and thus a lot of surface area. It may be secure but I just assume it isn’t. I rely on the VPN to be the security boundary.

How is the second LLM not also vulnerable from prompt injection? In order to supervise the first, it must receive data (presumably output from the first LLM?). All generated output after the user input is in the context should be considered possibly compromised/prompt injected. Having a second LLM just adds more obfuscation, but prompt injection could be chained.

I’ve seen it directly contradict the citation so many times that i disregard the text and just click the citation or scroll past every single time. Just today i caught it making up the date for an event, and the citation had accurate information when clicked through.

It’s super easy to catch on dates and numbers, but it gets other details wrong all the time too. But so many people won’t be double checking the results.

I thought most modern Bluetooth devices essentially randomize the Bluetooth MAC address periodically, specifically to prevent this sort of tracking? And random MAC addresses too on WiFi.

I had a problem recently where I ran a script with the wrong set of permissions, and accidentally screwed up the ownership of a random mix of files spread across my entire drive. This broke several pieces of software and made the system unusable.

I had enough information to reconstruct what files exactly got screwed up, and while I didn’t have a backup, I had a similar enough system I could pull “known good” file permissions from. I knew a simple script could find the problematic files and fix all of them.

I tried getting an AI to solve this. And it repeatedly gave me scripts that ignored all the details and intricacies of my issue and were functionally just "chown -R user:user /". (A command that will functionally nuke a drive, breaking ownership on every file)

The ai-provided scripts were reasonably complex and did a pretty decent job of obfuscating the disastrous outcomes the scripts would have inflicted on my drive.

After reading the man pages myself I wrote a simple enough script by hand and fixed the issue myself. AI wasted more time than it saved.

On my rokus, I am able to use my phone as a remote via the roku app. This includes typing on mobile via my phone's keyboard. Makes logging into things much easier.

While Texas is quite red. Renewables are surprisingly popular. Why should a farmer in the middle of nowhere have to rely on Texas’ power grid, when they can install a few solar panels and a battery. Especially when storms can take out power lines, or take out the entire grid.

I’m near a big city in Texas, and before any big storms here, generators frequently sell out at stores. Power outages are basically expected during any storms. Lots of people buying into solar (or backup generators/batteries) just for independence from the power grid. Especially after the huge winter storm a few years ago left people without power for days in the cold.

While fossil fuels are huge in Texas, solar and wind are too. Especially out in west Texas where there’s a lot of wide open space, wind turbines are surprisingly frequent. Texas produces the most wind power out of any state. And solar works just about anywhere in Texas. Lots of sun in the summers.

I eventually managed to get far enough into the article that I thought I saw the main stat - the stat that 26% of the citations were hallucinated. Then the scroll threw me back to the top again and I gave up entirely on reading from my phone.

Coming back later on desktop, I see that the percentage keeps climbing the further you manage to make it down the page. The real stat is 60% of the citations were hallucinated.

On mobile, It’s hijacking my scroll in such a way that I literally cannot move further down the page. And “reader mode” is only showing me the first paragraph or so.

I’ll have to try again later on desktop. The content looks interesting but it’s literally impossible to read. I cannot get past the section that introduces Ernst and Young.

LLMs are so frequently inaccurate its crazy to think of it fully replacing search.

I've been trying to use LLMs for things and it makes mistakes all the time. Just this week i had multiple instances of various LLMs basically saying "just run the software with --flag-that-fixes-your-problem" or "edit the config and add solve-your-issue=true" hallucinating non-existant options. Even if i manually link the relevant documentation pages it will still just make basic mistakes. and if im having to read the documentation myself anyway to fix the AI's mistakes, why is the AI even in the loop.

its infecting search too, because blogspam/slop articles are managing to make their way into search results by just making up untrue information, claiming software can do things it cant, or has options that don't exist.