And on the topic of whether LLMs can feel, even if you were to grant that LLMs think, may be self-aware, etc., a thoughtful person would have to grapple with the fact that the emotional reactions of humans (and other animals) are deeply tied in to feedback systems of the body, like the endocrine system, and so it's not even entirely clear that you can talk about a disembodied entity feeling (or that the set of feelings they would experience would be at all the same).

Kudos to the AP here. One of the reasons I can't take much of the AI discourse (especially LLM boosters) seriously is because of how much of it is couched in anthropomorphic terms. For any questions related to AI and philosophy of mind, this is essentially begging the question and, given that humans have well-known biases in this regard, actively encouraging irrational thought.

https://futurism.com/artificial-intelligence/ai-bros-meltdown-feelings

Question for #Signal users: I just transferred my Signal to a new phone including all my chat history (and multimedia). My laptop is also a linked device, and once I transferred to the new phone it became unlinked. I'm trying to figure out how to re-link it while retaining all the history (including media) that is currently on my laptop.

The dialogs for re-linking the laptop client (and web pages like https://support.signal.org/hc/en-us/articles/360007320551-Linked-Devices) seem to suggest that I will lose all the multimedia older than 45 days unless I have paid Signal Secure Backups. Is that really true, even if the data is all already on the linked device?

I use local device backups (which I then sync to my NAS) and not the cloud backups through Signal. It seems insane that I would lose a bunch of history on my linked device just because I transferred which device was the primary, especially when I additionally have backup files available containing all the history.

I just installed #GrapheneOS on a new (to me) phone, and man was that shockingly easy. I remembered it being easier than I was expecting the first time I did it (years ago), but either they made it even better or I forgot exactly how quick and simple it was.
#android #privacy

I guess now it's a bit late to warn people not to update the firmware on their #framework laptops, but here is what somebody did to fix them, when the company didn't help. at all.

quantum5.ca/2026/08/16/fixing-…

spoiler: it ends well, but not thanks to framework

also, the ending of the article is *epic*.

I was watching this video about passkeys recently and when they got to the part about a site being able to specify what sort of device was being used for the authentication it struck me again how passkeys will always be a technology that works for the oligopolies who created it, not the user. I hate using passwords for authentication, but for all their faults they are a thing users understand and are ultimately in control of, and sadly, for all their benefits, passkeys will never be.

https://www.youtube.com/watch?v=lypcC79k-gg

I find myself cursing #passkeys yet again. I decided to try to store some passkeys on my #Yubikey for some of my more sensitive accounts that don't allow FIDO for #2FA, but now the passkeys are mysteriously failing to register (using Firefox on Linux).

One of the challenges of FIDO is that there seem to be many ways for it to not work (issues with the site, browser, OS, and whatever you're using for credentials), and the error messages are either non-existent or totally opaque to anyone who doesn't work on this stuff. By contrast, TOTP and passwords are relatively straightforward, mostly just work, and are easy to reason about.

I'd really like to see the end of passwords for authenticating to online services, but it seems like if someone like me, who has been using key-based authentication and encryption his entire adult life, still struggles to make it work then it's not a viable alternative.

RE: https://mastodon.social/@fastmail/116519986875552738

Just reflecting on how nice it is to use @fastmail, where I just pay them for a good product/service. There are no ads, no B.S. AI features shoved down my throat, and it's not a startup likely to flip its business model any minute. I pay them money, and they give me something I want in return. Simple.

Once upon a time that would have sounded like a somewhat low bar, but it's an increasingly rare thing today.

Counterpoint: Human ingenuity is sometimes pretty awesome.

https://youtu.be/YdOXS_9_P4U

Current mood:

I use S/MIME email for work using a certificate provided by my organization. If I wanted to use S/MIME with a personal email account, where would I get a certificate (that recipients would be likely to trust)? Would I need to do it in association with a domain I own? I believe that Let's Encrypt doesn't handle certificates for S/MIME (since that seemed like the first obvious thing to look into). I get the impression that this is a thing few people do.

ETA: I don't care about whether the cert is free; I'm quite willing to pay as long as the fee is not exorbitant.

#encryption #email

I have been using an old-school podcatcher app (i.e. one that subscribes to #RSS feeds and downloads episodes to the device periodically, rather than streaming them in realtime) on my phone to listen to podcasts for a long time. Unfortunately, it seems like the one I've been using is quite abandoned, so I will need to find a replacement.

I'm looking for something I can use on #Android that's not locked in to a specific streaming service and, hopefully, something that looks unlikely to get immediately enshittified (so probably either FOSS or something that's been around for a good while and appears to be in a steady state). I don't mind paying for an app (in fact, if it's not FOSS then I probably *prefer* paying for it). It would also be useful if it supports Android Auto.
#podcast #podcasts

Based on discussion I've seen online, I gather there's another scare about the security of #Signal Sometimes these seem to just be rumors based on nothing. Sometimes they're actually based on something, but often it's being misunderstood or distorted. Does anyone know what this current round is based on?
#security #privacy

Looks like there will be strong #aurora tonight. As of this morning they're predicted to peak between 1 am and 4 am EST on Jan 20 with a Kp index of 8. That's strong enough to be visible in most of the northern half of the US (if the skies are clear and there isn't too much light pollution) and possibly more if they exceed predictions a bit (which certainly can happen). #northernLights #spaceWeather

https://www.swpc.noaa.gov/communities/aurora-dashboard-experimental

I finally had my first online account force me to use a #passkey. I've been afraid of this for a while, half because I prefer to use a hardware-bound FIDO2 key for certain things (but the specific implementation choices of passkeys mean only a limited number of hardware-bound keys can be used on one device) and half because the implementation of passkeys seems to be so buggy on so many sites.

It's frustrating, because getting rid of passwords in favor of something like passkeys sounds great, if only the implementation were a bit better (and not so closely aligned with major tech company wall gardens).

#security

My perception is that mobile OSs have come a long way in terms of implementing sandboxing and limiting permissions for applications, and some desktop OSs, like MasOS, have followed suit to some extent, but mainstream Linux distros don't seem to have (as far as I can tell). Are there Linux distros out there that do implement these sort of protections (and are still usable without becoming a full-time tinkering effort)?

I've read that both Snap and Flatpak incorporate some amount of sandboxing, but I don't know of any distros where that is the main approach to package management, although maybe recent Ubuntu with snaps (I haven't used it recently)? I know that there are things like Qubes OS, but that seems to be using virtualization, which seems like a actually going a level beyond what I'm talking about, and I gather that there are large compromises in terms of usability and performance.

#linux #security #infosec

A friend found a tab open in Safari on their iPhone that they had not (knowingly) opened and whose history indicated it had visited a succession of locations, including multiple web searches and several different websites (some related to the web searches, some not). What are the likely ways that this could have occurred? Could this indicate a larger problem?

My best guess is that a site they opened had ads from some ad network with malvertizing on it that opened the extra tab and caused it to visit the various locations, but I'm really not sure.

#infosec #security #malware #iOS #iPhone

Of course a better solution would be to have some of those dreaded algorithms, to allow more sophisticated prioritization of posts to view than a simple chronological feed. Just something simple like the "quiet posters" feed on Bluesky would be quite useful. Unfortunately, there is a level of phobia around this topic on Mastodon that I think precludes reasonable discussion, so I don't expect to see progress in that direction.

One thing I sort of wish I had on Mastodon is a delayed boost option. If someone has posted something that I'd like to boost, but their account has more followers and the following may have substantial overlap with mine, I figure boosting it shortly after it's posted probably does very little. But boosting it many hours later, when users from different parts of the Earth are likely to be online, seems like it should be useful. Right now I bookmark the posts to boost them later, but it requires that I remember to do it.

Reviews for many types of appliances have become less and less useful for me, because increasingly they focus on how good the appliance's app is and whether it integrates with the Apple, Google, or Amazon ecosystems. But to me these are generally anti-features that mean that the appliance will surveil me, be a security vulnerability on my home network, and potentially lose many or all of its features when some back-end server in a data center somewhere is eventually shut down or access suddenly becomes contingent on a new subscription service at an unreasonable price..

I really wish more reviewers recognized that having an app is a mixed blessing at best. It can be really useful in the short to medium term (if you're willing to give up some privacy), but it's generally bad news in the long term. In some ways the worst option is having an app backed by a service run by the manufacturer that doesn't (yet) have a subscription requirement, because that's an unsustainable model unless the product has manufactured obsolescence.