Subscribe
Share
Share
Embed
Screaming in the Cloud with Corey Quinn features conversations with domain experts in the world of Cloud Computing. Topics discussed include AWS, GCP, Azure, Oracle Cloud, and the "why" behind how businesses are coming to think about the Cloud.
Xe Iaso: I am actually a huge fan of bad ideas, and bad ideas is a gateway to good ideas, because A lot of the times, bad ideas do actually come from a place of like, genuine care, intuition, and thought. It's just that, for one reason or another, they're completely inviolable.
Corey: Welcome to Screaming in the Cloud, I'm Corey Quinn. One of the fun parts about being me, uh, not that there are many, but there are occasional breakthroughs, is that I get to encounter fascinating people doing really neat things throughout the industry, and then I get to sit in the curb and clap as they go by.
But I also get to drag them here to have a talk with me when basically you ask them in a position when they cannot possibly refuse. Xe Iaso is the Senior Technophilosopher at Fly. io. Xe , how are you? It's been several days since we last spent time together at a conference.
Xe Iaso: I'm pretty good. I had some coffee in me, so hopefully I'm ready.
Corey: Yeah, for those who aren't watching this on the YouTubes, I can't help but notice you're wearing a reInvent hoodie, which is awesome. Everyone loves to give out the shirts and the jackets and the sweatshirts and the rest, Socks as well. I want someone to start giving out like sweatpants or something just because then I can finally go and drape myself head to toe in vendor gear.
The only company I've seen that does this in their online store so far is Cisco. But I don't know that I really want to buy a Cisco underpants just yet.
Xe Iaso: Um, you may want to look at Leidas Tech Tips. Uh, they're a YouTube channel but they have a merch store and they have very surreal things like, uh, Underwear and sweatpants.
Corey: I did buy the Linus Tech Tips screwdriver, which bugs me because I feel like a fanboy. I don't even watch the channel, but it is the best screwdriver out there. I read like, I watched three hours of screwdriver reviews for God's sake. Yeah, it's a great, it's a great tool, I just, the endorsement almost cheapens it in some ways.
I digress. Uh, you are relatively recent to the DevRel role slash TechnoPhilosopher role over at Fly. io. What does Fly do exactly?
Xe Iaso: Your app, in a Docker image, 35 data centers, treat it like it's one. Basically, push button, receive server. I did
Corey: this with my former LastTweetInAWS. com Twitter threader before a douchebag bought and broke that entire site.
But LastTootInAWS. com does the same thing over on Mastodon. Unfortunately, threads aren't really as much of a thing. thing over there. But yeah, I started building that for my own uses and people complained it was slow. It's like, why is that? And they said, I'm in Germany. And I said, well, okay. So I deployed it to every region simultaneously with the CDK.
I had to build a weird CICD approach to make all of this work. And I had to commit some small war crimes to do it. And it, it works. But the fact that I had to step through all of those different, uh, different processes. Was more than a little bit annoying, so this is no small thing you're talking about.
Xe Iaso: Oh yeah, I have worked at a bunch of companies like Heroku and other things, and this is the only place that I have ever seen outside of big companies, like presumably Google has some thing to do it easily with Org. But this is the only place that I've ever seen where it's just push button, receive server.
One command lets me deploy to 35 datacenters for my CDN thing, and I don't even have to think about it. It's glorious.
Corey: That's been the dream for a long time and technically I suppose you could come up with any one liner and if with enough pearls slapped together The problem is is that it's it's something that's that you actually put some production polish on that There's a sense of I go ahead and I push the button and it works and not only that it's discoverable It's maintainable someone else who isn't me could come in and not immediately start raving over what has happened
Xe Iaso: The best part is that it's a one command thing and you don't have to write the Perl script yourself because it's actually written in Go.
Corey: Oh, I was going to say, it's, Perl is, I'm not one for language bigotry. I can write terrible nonsense that doesn't run in almost any language. But there was a time when Perl was something I was getting into a fair bit. As the more I spent time with folks, the more Go took over and of course Python has sort of become a lingua franca.
And now we see TypeScript everywhere, but Perl has really fallen by the wayside. You've done a lot of work over the course of your career with Go, to my understanding. What is your technical history with languages?
Xe Iaso: I did Python 2 for a bit, and then the Python 3 changeover happened. And someone on IRC suggested I pick up Go, and I didn't know that they were joking at the time.
But I took the IRC bot that I was working on. Well, it's less of a bot, and it more that it, like, spoke the server to server linking protocol and did some bot things, I guess. No, no, that qualifies.
Corey: I spent way too long helping run Freenode to discount that. Oh, it's only backend. It doesn't count.
Xe Iaso: I mean, I did have to reverse engineer some C code in order to understand how the linking protocol works since the documentation was nascent.
Corey: Was this IRCD 7?
Xe Iaso: Uh, no, it was Okay, so IRCD 7 is a fork of Charybdis, which is a fork of IRCD RATBOX, which is, you know, comes from IRCU and then European IRCD. The one that I used was based on a different fork of Ratbox called Shadow IRCD, which we then forked to Elemental IRCD, because we needed to add a channel owner mode. And doing that gave us the amazing opportunity to let you mode plus YOLO a channel.
Corey: Wow. That is, it's amazing how much, how much history there is behind IRC for what is, I'm going to get some flack for this, I suspect, but it's largely become a dead Protocol. Like, I spend time idling in a bunch of the places I used to idle, uh, now that it became Libera. And the only time it really comes to life is when social media breaks.
Xe Iaso: Yeah. I'm in a couple private channels because I'm part of the IRC Illuminati. And those channels are fairly active. It's mostly people that don't want to deal with Telegram, Discord, or whatever that Gen Z uses these days.
Corey: I'd miss the, frankly, one of the things that I always found amazing was Slack. It was building an IRC or Jabber equivalent for a company that I could just Send an invite to someone in accounting or marketing or someone, not that I'm disparaging those roles, but because they don't spend their days configuring things on a command line and suddenly they would just show up and it would have persistence across multiple devices and and it was amazing and then they apparently ran completely out of ideas and started redoing the interface at willy nilly and pouring all of this nonsense into AI and You're you're just a chat protocol. Don't don't overthink it.
Xe Iaso: Yeah, I sometimes miss the brutalist simplicity of IRC. I still use it a fair bit. Like, I'm active on IRC daily. I haven't, I also haven't touched my IRC client config in like, 6 years? But, that one configuration folder has survived being moved across 12 machines. And it probably has some of the oldest configuration and log files I own.
Corey: Yeah, I have a few things like that lurking around there. Technically, it's Genesis dates back to so long ago, it's old enough to drink legally. You are someone that I've always gravitated towards in one particular way. Namely, you seem to take it As a personal challenge, much as I do, when there seems to be an implicit statement of, well, this is the thing we've built, but surely you wouldn't be able to shitpost about it.
At which point you are, you are there with a level of technical skill I frankly envy. One of my personal favorites was, I believe you used S3 as a message queue to get around cross AZ data transfer. And I would be lying if I hadn't suggested such things to a number of clients. To use as a stick with which to beat AWS.
Xe Iaso: If I recall, that one happened when we were both drunk at a, at I think it was a bar.
Corey: I stop myself after two in public these days, otherwise I tend to break out in handcuffs. Doesn't go very well.
Xe Iaso: Yeah, handcuffs don't really do well, but on the way back from that thing, I ended up Flying home with Air Canada on the day that their IT service decided that it was time to shit the bed, and flights across the entire world got delayed by 12 hours or cancelled, and I was in an airport, in an airport purgatory, in a very low oxygen environment, giggling like a maniac on my laptop, making this horrific crime against humanity work.
And the most terrifying part was when I got a Prometheus response over HTTP, and it took seven seconds.
Corey: That's kind of amazing.
Xe Iaso: I was amazed that it worked because I was doing really naive code, but I saw the objects build up. I saw, you know, the TCP SYN, the TCP ACK, the beginning of, the beginning of transmission, and because I set this stuff really big, I actually saved the packets as a bunch of bytes, and I still have that packet to this day, and it has an HTTP response. Um, I can dig it up for you.
Corey: Yeah, I've considered doing something similar with EFS. I don't recall if NFS lets you have a socket living on it. I don't know that it does, but
Xe Iaso: I don't know about NFS letting you do socket devices. As far as I know, socket devices are pretty deep into the kernel in the VFS layer.
That being said, I am more than welcome to be completely f ing wrong there, but if, if that worked, that would be absolutely ridiculous.
Corey: Advantage yourself of the free cross AZ data transfer that AWS helps themselves to and doesn't pass on to anyone else. So yeah, if you use one of their managed services, you get free replication traffic. I mean, technically, I could do the thing I always told people not to do, and set up replication in RDS instances and then use MySQL as a queue.
Xe Iaso: Oh, that's horrible. I think that's just about as horrible as the time that I heard that someone replaced Kafka with email.
Corey: On some level, email's the original base level API for everything.
Corey: Yeah, the
Xe Iaso: thing is, you can actually represent an email as a durable message queue like Kafka. Because the main thing that Kafka gives is the ability to scroll up and see events before what's happening now. And with the way most email servers work with the IMAP protocol, you effectively have a pointer that you use to iterate through the entire list of possible emails.
And then, bam, you know, you've basically got Kafka at home, but your sysadmin hates you.
Corey: Yeah, I love that. Like, these ideas lead to weird places, because no, someone should not absolutely run this in production, let's be clear on that. But it does get you thinking about problems in a different way. You've always had a flair for making salient points shitposts.
Xe Iaso: Oh yes, I am actually a huge fan of Bad ideas and bad ideas as a gateway to good ideas because a lot of the times bad ideas do actually come from a place of like genuine care, intuition, and thought. It's just that, for one reason or another, they're completely inviolable, like using email as a message queue.
That's horrible, like, making your program speak SMTP is, well, we're running out of goats, and if we keep sacrificing them, we're going to have a goat extinction on our hands, and how is Chrome going to teleport goats if there's no more goats to teleport?
Corey: Rory sutherland is the vice chairman of Ogilvy, the ad agency, and it feels like we've discovered a corollary to what he often says, which is that the opposite of a good idea is often another good idea.
It feels like, yeah, but I started with a bad idea. You can go for worse ideas and then somehow through some shitpost alchemy, it turns into something that is absolutely worth
pursuing. Yes, that's basically the stuff that I love to do and the stuff that is annoying because your human numbers cannot measure the impact of a shitpost.
Well, not with that attitude. Okay, you can measure the impact of a shitpost. It's just very invasive and personal and not stuff I like to do. I
remember back when I used to have a real job and my boss wandered past and like you'd hide up whatever thing you weren't supposed to be doing at work these days.
Now, not that I have a boss, but it's so what did I spend my day doing? Shitposting on the internet and that is in many ways directly beneficial to the company if you do it right. The trick of course is doing it such a way that doesn't sound forced. Oh, I've seen a lot of Unfortunate talks in the DevRel space that tended to take this particular pattern.
Where it's someone rolls up and, Hi, I came halfway around the world for this conference. I'm sponsored by and then they, not sponsored by, but I'm employed by whatever the company does. They talk for 10 seconds at what it does and, and now my conference talks is how to pick the best standing desk for your home office.
And that's what I'll talk about for 45 minutes. And, It just, it strikes a dissonant chord there. Similarly, if I were to make every conference talk I give about AWS bills, that would not get me invited back to a whole lot of conferences.
Xe Iaso: Yeah, it's the, that balance of like, repping the brand and giving people signal is so hard.
Um, for the record, what I mean by signal, is there's this theory of communication called signal noise analysis, where the signal is, you know, the juicy technical stuff that people want to hear at a talk, but the noise is brought to you by our sponsor, Squadcast. We're Uber, but for recording podcasts, or something, right?
And people that you work with, you know, bless their hearts. They mean well, but they're gonna want to tend you to like rep the bee or have stuff be a product demo. And you have to find that right balance of latent product demos and actual engaging stuff. One of the talks I gave recently was at an Olama meetup.
I knew that it was going to be a product demo. I warned the organizer that it was going to be a product demo. And at the end of it, I actually made a joke of the fact that it was a product demo by acknowledging it and saying, If I missed your question and you want to ask me, email productdemo at domain name.
Corey: I love that.
Xe Iaso: At some level, being self aware about it helps. But at other levels, you know, it's a balance. And finding that balance is always hard, always something you have to find out the hard way. And, uh, hopefully you can stay employed trying to figure it out. We can, let's hope.
Corey: Now, the, the problem I keep smacking into with so many of these things is that it's, when you get too close to a particular product, you start to see that it's the solution to everything.
And sometimes it's because the product is legitimately good. I don't think a lot of people work at places they think the product is garbage at. But. Sometimes it looks like every tool becomes a hammer and every problem looks like hours of fun, beating it to death. And that, I feel like people lose sight of it where they're, when their compensation is tied so heavily to that product doing well.
I've always taken the approach of if I'm going to go and shoot my mouth off in public about things that are cloud adjacent, that'll build an audience and people will stick around. And if I periodically, organically mention that I fix AWS bills for a living, When people suddenly have that expensive problem hit, they'll remember who I am and where to find me.
And it seemed like a very tenuous connection, but seven years in, we don't do mass market marketing approaches. We have no billboards in San Francisco. I'm not very good at kicking people's door in and say, you want to buy some cloud optimization? Doesn't work very well. But being notable, being noisy, and being highly targeted, Works.
Targeted a problem, not specific people, to be clear, because I don't know if someone is having a problem with their AWS bill. They don't give signs of that in public once they're at a serious
company.
Xe Iaso: Yeah, and a lot of the times, many people don't even know what the AWS bill is or, like, what the factors are, because it's Well, because it's so hard to purchase things at companies, they've made a habit of being able to use AWS as the payment mechanism for that.
And that's like an actual huge part of the product that is essential and load bearing at many institutions at this point, which is just hilarious to me.
Corey: Changing gears slightly, uh, we were at Scale, Southern California area Linux Expo last week. And mid sentence, as I tend to tangent like this myself too, which I felt great, you basically just whipped out your camera and took a picture of me on the fly.
Uh, since then, I have purchased the rights for you from that, and that's become my new profile picture everywhere, because it's very on brand, it's very much something that is in line with how I present, and it was an absolute candid shot, middle of a sentence, and it works. I don't love aspects of the photo and the camera, I mean, there's something wrong with your lens, it gives it a really nice bokeh effect in the background, it also makes me look old, and I don't like that.
It also makes me look unshaven and bleary eyed and honestly like I've had just too much dealing with people and man, what is in that camera?
Xe Iaso: Uh, this is a vintage lens from the Soviet Union called the Helios 44. It is probably the most common camera lens that you can find on the market because they produced like three million of those and the bokeh effect is because they tried to copy a Carl Zeiss design out of East Germany.
And they f ed it up in the most glorious way, to the point that cinematographers will actually, like, rehouse these lenses just for that, that, like, super swirly bokeh effect. You can see it in The Batman, you can see it in Dune. And actually, I don't think the showrunners or whatever have announced this, but I'm pretty sure I saw it in use on the Halo TV series during some of the dream sequences where they want stuff to be like, the background to be kind of dreamy and out of focus.
Corey: I wound up removing the background from most versions I use and just slapping it. I'm slowly iterating through all the different sites that have a profile picture on it. Just because it, my, it more or less looks the same pose as my usual face, which is my resting open mouth dumb face. But the, the problem I run into at some point is, okay, that picture is almost 10 years old.
That's like more or less being on a dating site. But here's a great picture of what I look like as a teenager. Like that, that's a little misleading at some point.
Xe Iaso: Photography is surprisingly difficult, and my goal is to never make photography my profession. Which, given that I'm doing DevRel, is kinda weird because I also do video, and I use the same camera for video.
But, I'm also grabbing some of my better photos over the years, and I'm donating a couple to smaller Linux distributions to use as wallpapers. I got a gorgeous one of the Space Needle that's going to be the default in Bazight Linux, which is like SteamOS, but built on top of Red Hat.
Corey: Nice. One last topic I want to get into.
We talked about IRC earlier, and I don't know if you recall, but there were some, uh, operating system support channels, some of which I may have participated in at one point, ahem, ahem, where someone would show up asking for help. Like, I don't understand how to do this thing. And the response was basically, well, that's probably because you're an idiot.
And it was about that level of kindness and empathy and support to the point where things were so confusing in many cases that the promising projects died on the vine just because driving away anyone who might be interested in using the thing didn't work. Wasn't a great approach. Uh, now, Nix is an area in which you've spent a fair bit of time, and I want to be clear.
They are, they have not been toxic in any way. They have been warm and welcoming, and in my experience, almost completely incapable of articulating why you would want to do a thing, and a step by step guide to getting from wherever you happen to be, to achieving that thing. I've tried three times now to get into Nix, and every time, it's been a The documentation has been clearly written for someone who already knows how it works.
And that's awesome and surprisingly unhelpful for me. So my question for you, quite simply, is that given that you are so gifted at explaining complex concepts to simple people, by which I mean of course myself, how is it you're involved with this and it hasn't gotten fixed yet?
Xe Iaso: Because I haven't been paid to.
Corey: Excellent answer. Oh, I like that.
Xe Iaso: At one point, I was going to work at, uh, I think it was Determinate Systems to basically be DevRel for Nix and fix it. But that fell through because I had another job at the time and wasn't open to new opportunities. The other reason why it's failed is because, like, as I'm sure you're aware, I have, I do a lot.
I write a lot. I also do photography and video. And I'm trying to, like, keep my weight down so I have to allocate time to exercise, and I just don't have energy at the end of the day to do that stuff as much. Oh, and, I have been getting the beginnings of RSI from writing too much with bad posture. Don't worry, I'm fine.
This thing right here, this 400 microphone, er, 500 microphone, replaces a keyboard for a lot of, for a lot of tasks.
Corey: That's amazing. I'm pretty sure that like, like effectively using a microphone and dictation to wind up speaking the entirety of a way through a Kubernetes command line, uh, discussion at some point, like you'll be institutionalized halfway through because, Oh my God, they started speaking in tongues. And that's sort of the end of it.
Xe Iaso: Yeah, my husband has, when I do voice coding, sometimes I'll forget to close my office door, and my husband has described it as demonic chanting, which I like that, I think is one of my favorite things that he's ever used to describe it. He's actually tried it for himself because he does simulation racing in VR, and it turns out that the best input method for a computer when you're in VR, and you know, your hands are coded in these controller things, is voice.
And occasionally in VRChat, we'll like, hang out with some other people. I'll forget to mute myself in VRChat and then just rattle off a series of commands or like, Discord messages to people. The demonic chanting thing really does come out because, especially with letters. Typing letters, you type letters with words.
So, like, Air Pit Pit Look Each for Apple. A P P L E, Air Pit Pit Look Each. End. It's like the natophonetic alphabet, except everything's one syllable and as phonetically distinct as possible.
Corey: You also to teach yourself a side language in order to get some of these things working correctly at
it would seem.
Xe Iaso: Oh yeah. And I also write a fair bit of stuff, uh, for configuring it. At one point I figured out that gi and GI are similar enough in my IDI elect, like as in gi as in, uh, get the, uh, blockchain that we use for source code, for source control and get the core English verb that has like 50,000 meanings. They sound similar enough that voice control gets confused between them.
So I had to rename Git to Great Calzone and that solves all the
problems.
Corey: How do you get from Git to Great Calzone? I have to ask.
Xe Iaso: I think I, I mentioned it and I mentioned it in some place and some drunk person was like, some drunk person talked about having a really great calzone at the time. It's phonetically distinct.
It's completely out there and I cannot, for the life of me, imagine something involving calzones to ever exist in my space again. So, Git is great calzone and I have some macros for things like pushing, pulling, adding files, checking out a branch, just the basic actions that you use 90 percent of the time.
When you get to the really fun stuff, that's when you have to spell out like gusset, trap, space, word, rebase, or whatever.
Corey: Do you find that You get better results in teaching people things by taking a deliberately, I don't want to say antagonistic, but, uh, but a, I guess a shit posting direction with them.
Because in my case, it's always been something very intentional, but it also is something, let's face it, that there was an outgrowth of my own personality attributes, uh, call them benefits, call them defects, whatever you'd like. Did, have you, was it an intentional choice for you and you just found it worked?
How did, how did you get here?
Xe Iaso: So, f ing around and finding things out is one thing, but the second you write it down, it's science. And the way that I got here was effectively A B testing my way through the entire thing. Back when Twitter was owned by a guy named Jack, I had these long, giant threads of me going through and talking about various technology things.
I think the record is like a Foon style 300 tweet monologue. About trying, I think it was trying to install FreeBSD on a Raspberry Pi or something mundane like that. And people found the snarky sense of humor helped, and I ended up teaching people, and they ended up understanding. Given that it worked, if it works, do more of it.
So, I tried doing more of it in different places, and the reason why I'm here now is because I've apparently shitposted my way to success. I think
Corey: it's refreshing just because there's so much out there in the context of difficult to wrap your head around technologies explained by corporate interests who are, who speak in the particular same bland corporate tone where, at least for me, I, I zone out almost immediately.
You have to bring it to life and make it interesting. Just for my own sake, if not for anyone else's.
Xe Iaso: Oh my gosh. I, I used to work at Salesforce. And whenever you hear someone say they used to work at Salesforce, the correct thing to say in response is, I'm sorry. Or if you're from my side of the Snow Peso border, it's a, I'm sorry, eh?
Or whatever. And I think the record that I saw at Salesforce for lack of meaning in a single, I don't know the right word to use here, so I'm going to use the word utterance, was about five paragraphs to express a single sentence of meaning. And it's, it's, At some level, like, American Corporate English is amazing in its ability to not, to take this tool designed for communication.
and make it non communicative. To have a whole bunch of complicated sounding things that live, that mutually cancel each other
out.
Corey: Chachipity can go both ways in the sense of encapsulation and decapsulation of Chachipity, put this in business email. Tell this person to do their job. And great, it puts in this flowery five paragraph circle the point to death nonsense.
And someone on the other end takes it and it's like, cool, what does this person want? It's like, do your job. Oh,
okay.
Xe Iaso: Yeah, somebody, I think somebody at some point joked about having a, uh, a pipeline where someone says a simple message in one end. And because we're expected to say complicated, you know, flowery corporate things, they have a large langolmangol translate that into, you know, the flowery corporate stuff as like some kind of f ed up encoding.
And then on the other end, it's decoded back into the plain normal English. Okay, I've had a thought about this for you for a while, but I don't know, I haven't found a good place to actually do it. I would love to see what would happen if you tried to do that over and over in a loop to see what the loss of information would be.
Corey: Interesting, like almost like a game of telephone, but you can have arbitrary number of cycle
times. I did try doing that with Stable Diffusion a while ago and it devolves to porn instantly. Not at all surprised. It is absolutely incredible. There's on some level
it's a, it's an aspect and a reflection of humanity and what, and what things are trained upon.
I think that the idea of, oh, whatever, they'd be extremely brand safe. Okay, I get that. But understand the way that humans interact with each other is not purely commercial beings of pure light and money trying to transact with
each other.
Xe Iaso: What's that old saying? Uh, the internet was, is for porn.
Corey: Avenue Q, did his whole musical about it.
Xe Iaso: Back before El goog came in and defined what the internet was, the companies that literally defined how to do video distribution or large scale image distribution or whatever. were things like Pornhub. Like, at one of my jobs, we used this thing called MogileFS, which, you know it's going to be good, because its name is letters mixed around for OMG files, and that was basically the same type of static asset serving platform that Pornhub used, and I only found out, like,
Corey: People don't like talking about adult entertainment, uh, but there's so many technology lessons coming out of that because the scale and a usage pattern that is unlike most other things.
I'm a big fan of the lessons learned from that, but for some reason, companies really don't like using those as references. I mean, I've, I've helped a number of adult entertainment company, uh, website builders and creators and whatnot deal with weird AWS bills because they're freaking out and they're like, Well, look, some of this might be a little risque.
It's like, I promise you, it is nowhere near as offensive AWS bill. Manage NAT gateways, charge what again yeah, awful.
Xe Iaso: I think about part of the reason why people are so loathe to do that is because it risks pissing off the payment processor, and if you piss off the payment processor, your money generator is dead.
And for some reason, payment processors are super puritan and stable diffusion. Being a publicly available thing and getting to the point where it's good enough has really affected the livelihoods of a lot of my artist friends.
Corey: Oh, lord knows, I used to pay for a blank, a flat write static image licensing service that I would use to build slides, and now I'm doing a lot of that with Gen AI.
Just from the perspective of, it's, for easy example, at one point I was talking about something with cloud contracts and whatnot, and I wanted a line of Italian mobsters outside of a restaurant waiting patiently in line to extract their pound of flesh from the proprietor, and as they go through in turn, and I could not find anything in a static site that evoked that, but, you know, a couple of flowery phrases later and suddenly I have several options from which to choose.
It's amazing for the quick and dirty things. Like, I've also For me, it's transformative in that I used to look for rapid response graphic artists where, middle of a keynote, I want to say, Great, here's the picture they just put on stage, like, slap devil horns on it, go. Now you have things that you can come up with on the fly in less than 30 seconds.
There's magic to that, but it's also trained on people's intellectual property, the work that they've done. How are they compensated for that, like, well, they shouldn't have put it on the internet, suckers, is not a valid answer
here.
Xe Iaso: I have found out that my blog is in the training set for ChatGPT, and as a result, because I've written like, oh what is it, it's getting close to 3 3D printed save icons worth of text at this point, I'm not going to be able to go to college like I've wanted to, to end up getting a degree in either philosophy or linguistics, because Even though everybody involved knows that the science for trying to detect if a given string of text is generated by an AI model is total bogus, absolutely doesn't work, but colleges use it.
And I would have to have a conversation with teachers and deans constantly of, I wrote this myself. I took a screen shot, a screencast of me recording it so you can, writing it so that you can see I wrote it myself. My blog is in the training set for ChatGPT and I have written so much over the years that everything I write is just going to be marked with it.
Corey: This was created by Gen AI, yeah, so was the U. S. Constitution, what's your point? It's depressing. All of those detection things have apparently been proven in controlled studies to be snake oil. All this does is basically accuses, it basically gives a good bluffing card to a professor or teacher to yell at a kid to get them to confess, but if not, they've got
nothing.
Xe Iaso: Yes. I'm going to be speaking at a college class next week about AI ethics. That's going to be real fun. Hopefully I won't be, I won't leave some of the poor kids traumatized. Especially after talking about the Silicon Valley model of consent and why that exists why that exists from a UX standpoint. Those yes or remind me in two weeks. That's not just because they don't have a good understanding of consent. That's from a UX design perspective.
Corey: Your tweet was amazing about the Silicon Valley of consent. Yes, or ask me later.
Xe Iaso: The reason why they do the ask me later is because sometimes people might accidentally click on the ask me later and not get exposed to something that they might genuinely be interested in.
But then again, if you want people to be interested in why are you blocking my f ing desktop? Go away, you techno cretin.
Corey: I have said not right now, at least three dozen times now. Uh, to the, do I want to use the Dropbox for Mac beta? No, I don't. You control the, like the corporate files I need to do my job.
I don't play slap and tickle with that stuff. I don't want to use the new fancy stuff, make it the usual thing that gets pushed out or
shut up about it.
Xe Iaso: Like there's, there's real good intent behind that. It's just comes across as invasive.
Corey: It's all about incentives on some level. People want to see people use the new things. The problem too, is that. Companies don't seem to understand that rapid changes to things or significant changes to things are going to always be met with pushback once you have a certain size of a customer base. You change the interface in Microsoft Excel, accountants will march on Redmond before lunch.
Xe Iaso: Yeah,
I encountered this recently with Slack, and for those of you who are in the industry, haven't been in the industry long enough to experience this, I use, like, the accessibility tooling I use integrates with Slack poorly, because nobody fucking allows plugins to be written, or extensions to be written, without breaking some kind of EULA, and if you get prot breaking the EULA, congratulations, you no longer get to do your job. So the way that my accessibility software interacts with Slack is by sending out keyboard shortcuts. The Slack UI change, where they're gonna, like, integrate Gen AI into everything, is Change the, the keyboard shortcuts inconsistently between slack and between the slack sessions I was logged into. So I would be on the go slack and it would work the old way, but then I tabbed over to the work slack and it worked completely different.
And it was a total mess, and I had to give up and just use my hand, uh, used my hand with the mouse. It was just an absolute train fire, not, not a train wreck, not a trash fire, but the train was on fire going down the tracks towards the station. And it was not a good time. Uh, I complained to some friends of mine who work at Slack about it and they're like, Oh, we did not consider the disability access standpoint.
Corey: No kidding.
Xe Iaso: Well, yeah, because Making things accessible doesn't show up on a graph. The only reason why so many iOS apps are accessible is because if you want to do unit testing, you have to have the accessibility framework enabled. Because unit testing goes through the accessibility framework.
Corey: Someone in the app legal system once said if someone says they are an apple developer Uh, you know, used to be the home button, now it's the power button. And that doesn't fire off an accessibility shortcut. Their app is likely trash. Because you make it do all kinds of other things. I, on mine, it turns my screen into grayscale.
Xe Iaso: Oh, I actually, I use the triple click for guided access. Uh, the, for grayscale, I, uh, tap the back of my phone twice.
I find that whenever I open something like whatever we call Twitter now, I, unless I'm, for a while I instantly, I turned on Grayscale before I did it to make it less enticing, enticing? And then I have taken a multi month process of training Twitter's algorithm to only show me cute Xenoblade fanart. And I've gotten to the point where I just open there and there's like maybe some tech posts and mostly cute Xenoblade fanart.
And I am all for it. Like, let me look right now. There's Xenoblade fanart. There's something from Amos. There's something about AI. There's more Xenoblade Hopefully that's not the Xenoblade fan art made with AI. You know, I've trained it. Because I engage with the Xenoblade fan art, I go out of my way to. So if you follow me, you'll probably just randomly get waifus on your feed.
Corey: I curate mine, and I have used a browser extension that forces me, that forces it to always look at the chronological feed, because I don't really care what the algorithm has to say. If there's something important I need to know about, the people I follow will flag it.
Xe Iaso: I only use Twitter out of obligation at this choice, at this point.
Corey: It's where the audience is, I have to be everywhere. I really want to thank you for taking the time to speak with me today. If people want to learn more, where's the best place for them to find you?
Xe Iaso: Probably on my blog, which is apparently a radical statement in this day and age. It's Xe Iaso. net, I'm pretty sure it'll be down in the description.
Corey: It will indeed be in the show notes.
Xe Iaso: Ah, show notes, right, this is podcast, not YouTube. Heck. Yeah, either on my blog, I have a link, links to a bunch of the social media feeds you can find me on. I have been attempting to syndicate things to BlueSky, but Mastodon is where I will, like, I will just put things, and that's where all of my automation is set up to.
I used to have, like, a fairly intricate automation pipeline, so that whenever I posted something, my blog would poke another service that is like my blog, except it has state. And see if there's new articles, and if so, spread them out to the feeds.
Corey: That got harder and harder to do with API restrictions.
Xe Iaso: Oh my gosh, f ing Patreon.
Oh my f ing God. They have decided that they do not want to improve their API, and they never implemented the API call to post things. So I was working on some post metadata, so that I could include things like the description for people on my Patron. And I was never able to because they just didn't implement the one damn call.
And I don't feel like reverse engineering it with the browser inspector.
Corey: Well, definitely put links to that into the show notes. Thank you so much for taking the time to speak with me today. As always, I do appreciate your time.
Xe Iaso: Yeah. Thank you.
Corey: Xe Iaso, the Senior Techno Philosopher at Fly. io. I'm Cloud Economist Corey Quinn, and this is Screaming in the Cloud.
If you've enjoyed this podcast, please leave a five star review on your podcast platform of choice. And if you hated this podcast, please also leave a five star review on your podcast platform of choice, along with an angry, insulting comment, uh, including whatever your complaint is in shitpost demo
form.