Aegis works perfectly. My university used Duo and then Okta, and both worked. I was unable to get Authy to work, though.

She independently discovered it haha. I sit her down on the couch next to me while I’m working, and she constantly flops over. It drives me nuts! I would let her just do her thing, but she’s got pretty bad reflux and even just a minute or two of tummy time within like 2 hours of having a bottle will make her ralph.

The other day, I had to get up for a minute or two, and I left her on the couch. My better judgment kicked in and I pushed an ottoman next to the couch to block the space she usually tries to launch into. This is what I came back to see:

That’s the main reason I jumped on them. I’ve seen an increasing trend of people selling old computers with the RAM and HDDs taken out

My wife taught my son English uppercase letters, so I’ve been teaching him Greek lowercases. I can’t help but feel proud every time he says “pi”

In the past, I’ve used Adguard Home, and I liked it. When I tried to my Adguard server as the DNS for my router, though, my WFH corporate VPN wouldn’t connect, so there’s that. Granted, I was using it to remove ads, but people seem to like it for parental controls too.

I saw a deal for a pair of mini PCs with decent specs on eBay right after I got my annual bonus, so I jumped on it because I want to do the same thing for my kids. May be jumping the gun a bit…my son just learned the alphabet (uppercase only) and my daughter just learned how to flop off the couch head-first.

I self-host Gitea! Not your server, not your data.

Yeah, all of your use-cases are what I see as positive use cases for LLMs. I’ve got an Ollama instance hooked up to Home Assistant, but it does not work very well haha. Haven’t had the time to troubleshoot it.

Can the rubber ducky use case really be considered plagiarism? I think it’s unequivocal that the models were trained on copyrighted data in a way that, if not illegal, is at the very least unethical. Letting AI write stuff for you seems a lot more problematic than using it to bounce ideas off of or talk things through.

What about a self-hosted instance?

Thanks for the read, I learned something today! I worry, though, that even if someone could devise a ZPK for age verification, can end-users actually trust that platforms are using it? Say for example Meta provides a biometric-based ZPK for age. Can we trust that they’re not harvesting our biometric data? In the podcast’s examples, it’s easy for Peggy and Victor to understand that they are using a ZPK system. However, the age verification problem most often arises in arrangements where the prover is using a client app into whose inner workings they have no insight (either because it’s closed source, they’re not technologically literate enough, or who has the time to scrutinize the source code for every program they use) and which is most likely developed by the verifier. So the problem kind of moves upstream: how can you trust that ZPK is actually being used?

Before Windows 11, I told people to switch to Linux because open source software is better for the soul. Now, I tell people because the user experience is just better. I used XP/Vista/7 throughout my childhood, and modern Linux desktop environments really do feel closer to that experience than Windows 11. I use Win11 for work, and I can confidently say that it has the worst settings menu I’ve ever used.

If you know the basics of using a desktop computer, most things won’t feel that weird or foreign to you. The hardest part will probably be learning Linux-compatible alternatives for apps that only work on Windows. What kind of programs do you typically use on your Windows system?

Instead of burning the ISO to the flash drive, I recommend burning Ventoy to your flashdrive. Then you can drag and drop ISOs for every distro you want to try without having to burn them every time.

I mentioned it in my original comment! I thoroughly enjoyed it. As an older member of Gen Z, a lot of what’s written there jives with my lived experience and the intuitions I’ve developed around social media. And as a relatively young father, I’m also invested in figuring out how to give my kids the healthiest relationship with the online world possible.

I’m also a strong proponent of digital freedom and privacy. A lot of the age verification technology that’s being rolled is tied to companies like Palantir or organizations like DHS, which seem to have a rather unambiguous interest in neither the freedom nor the privacy (nor really the general wellbeing) of the populace.

I’m of the opinion that any system that could enable or facilitate mass surveillance is not an acceptable solution to the problem of protecting kids online.

The idea I laid out in my original comment was inspired by the idea Jonathan Haidt presents in Chapter 10 (What Governments and Tech Companies Can Do Now), Section 3 (Facilitate Age Verification), 6th paragraph:

There is not, at present, any perfect method of implementing a universal age check. There is no method that could be applied to everyone who comes to a site in a way that is perfectly reliable and raises no privacy or civil liberties objections.[26] But if we drop the need for a universal solution and restrict our focus to helping parents who want the internet to have age gates that apply to their children, then a third approach becomes possible: Parents should have a way of marking their child’s phones, tablets, and laptops as devices belonging to a minor. That mark, which could be written either into the hardware or the software, would act like a sign that tells companies with age restrictions, “This person is underage; do not admit without parental consent.”

Yeah, there’s no doubt in my mind that this tide of “think of the kids” is just a fascist dogwhistle (and one with a double-entendre at that).

The book The Anxious Generation by Jonathan Haidt had a really clever idea. Create a regulation for operating systems that requires that their parental controls include an option that labels a device as belonging to a kid. When that option is toggled, requests will include some sort of header that labels the request as originating from a kid. Then, place onus (probably through some sort of legislation) on web platforms to restrict what content is shown to kids.

I know AI is an emotionally charged topic, but I think your frustration is misdirected. I find that the best way to learn tech stuff is with hand-on experience, and to that end, it works pretty well to try something, ask why it didn’t work like I expected it to, and get instantaneous feedback. Or to start with a working example and pick it apart so I can learn the syntax. I’m not saying it’s a replacement for reading official documentation or figuring things out for yourself, but it makes it a lot easier to get started.

Fundamentally, I’m a humanist. I believe that we should use technology in a way that augments our brain instead of circumventing it. I don’t let AI write code for me, but I don’t really see the harm in having it present information in a digestible format.

I’ve always been bored by lectures and tutorials because they’re not good at meeting me at my level of experience. I don’t think anyone would argue that having a tutor/mentor who gives you individual attention and meets you where you at will help you climb the learning curve way faster. And when you’re in a situation where you don’t have a human mentor, AI can be pretty useful.

I worked at an organization where there were no senior software people and my supervisor told me you “hey, you created this dashboard – now deploy it”. My only relevant experience was having hosted a Minecraft server on Windows 10. After a few months of iterating with ChatGPT, I knew the basics of how to use containerization and deploy an app on a RHEL server. 3 years later, I’m doing it at a tech consulting firm, and I’m the guy everyone goes to for help writing containerfiles and compose files. They promoted me from data scientist (I have an MS in data science) to solutions architect, all because I used AI to learn the basics of Linux devops, and then got a shit ton of practice by self-hosting.

Even for things we’re excited about, truncating the y-axis is never okay 😕

My best friend used to go by @comet on Discord. He’s the one who made my friend group’s server, and he’s essentially the only reason I ever got it in the first place (now, he’s one of the first to join my Matrix instance). The fact that this project is named Commet is NOT good for my “main character syndrome”!

I initially subscribed to ChatGPT because I got a job as the only devops guy at an organization, when I had very limited devops experience, and ChatGPT essentially served as my mentor. I justified keeping it for a long time because it helped my productivity; bugs that I had no idea where to start with could be worked through given a few hours (or days) of back-and-forth.

As I climbed the learning curve, ChatGPT became proportionally less helpful, but I kept it because it’s kind of useful for rubber ducky debugging. I did find Copilot to be pretty handy for writing docstrings (especially for keeping consistent formatting conventions), but the actual code completions were more annoying than anything.

When all was said and done, I cancelled my ChatGPT and Copilot subscriptions because I’m taking on a mortgage tomorrow and I literally just can’t afford them. I have Ollama running on my homelab server, but I only have enough vRAM for a 7B-param model, and it kind of sucks ass, but whatever. At the end of the day, I like using my brain.

UPDATE (because I just thought of it after posting): I do think that “AI-as-a-mentor” is a good use-case of AI. It really helped me cut my teeth on the basics of Linux. I often find that it’s easier to learn when you have a working example of code or config that you can dissect than to bash your head against the wall just trying to figure out how to get something to run at all in the first place.

For my birthday challenge this year, I’m learning how to read and write Devanagari as a surprise to my Indian grandparents. I asked my local qwen model to generate some worksheets for me to practice with, and it totally flopped. It gave away all the answers. I do think ChatGPT would have done better, but maybe I could have gotten sufficient results with a better GPU.