624: Do Less Math in Computers
30 Jan 2025DeepSeek’s impact, Apple’s position in AI, updates on scrolling and home LEDs, and an adaptive apology.
Episode Description:
- 🗣️ New ATP Movie Club Special! Star Trek IV: The Voyage Home
- Follow-up:
- Honda Prologue vs. GM equivalents (via Matt Rigby)
- Apple insists it’s still going to
make fetch happenship next-generation CarPlay - Handheld gaming PCs are having a moment
- Ryan London is trolling John
- John has so much RAM, he can’t even remember how much he has
- Hyperspace scrolling updates
- Not all phone jacks are created equal (via John Haydon)
- Casey’s notification/LED project
- Garage door ↔ home assistant was already sorted
- ESP32
- Did you know you can just straight-up cut some LED strips‽ (via Kiel Olesen)
- Historical commission application process
- Switch replacement options:
- HomeSeer HS-WX300 (via Drew Stephens)
- Inovelli Smart Dimmer (via everyone, including Drew Stephens)
- DeepSeek causes a tech stock sell-off
- Siri Is Super Dumb and Getting Dumber
- Ben Thompson’s FAQ
- DeepSeek-V3 paper
- DeepSeek distilled GPT-4o?
- R1 is more open than OpenAI
- Interview with DeepSeek CEO Liang Wenfeng
- R1 censorship
- Post-show: Marco wants to do some local web development
- Members-only ATP Overtime: Sonos Shakeup
Sponsored by:
- Members like you! Become a member for ad-free episodes, member specials, and our early-release, unedited “bootleg” feed!
Chapters
- An adaptive apology
- ATP Movie Club: Star Trek IV
- CarPlay updates
- Hand-held gaming PCs
- Leather case update! …ish
- Follow-up: John’s RAM
- John’s scrolling updates
- Phone jacks: sometimes Ethernet!
- Casey’s LED updates
- DeepSeek
- Ending theme
- Local web dev
An adaptive apology
⏹️ ▶️ Marco So I need to issue an apology.
⏹️ ▶️ Marco I’d like to apologize to Airpods Pro adaptive noise cancellation mode.
⏹️ ▶️ Marco A few months ago, we were talking about the Airpods Pro noise cancelling modes and I
⏹️ ▶️ Marco had said I had tried the adaptive mode for a little while and I didn’t like it
⏹️ ▶️ Marco and I switched back and I disabled it from my like, you know, my cycling when you hold down the stem
⏹️ ▶️ Marco because it was too it blocked out too much street noise. noise when I was walking around Manhattan, and I just,
⏹️ ▶️ Casey and I didn’t like it. I don’t entirely agree with you, but I, I didn’t, I never got to the point that I wanted it removed
⏹️ ▶️ Casey from my, you know, cycle of options, but I find I very rarely choose
⏹️ ▶️ Casey adaptive. Typically I’m either doing transparency or full-on cancellation.
⏹️ ▶️ John And I’m pretty sure the way we talked about it was because I got the AirPods 4, and I was asking you guys about the different
⏹️ ▶️ John modes that they do, because I didn’t know which modes and what they were good for. I think at least at that point, you reiterated your opinion that you
⏹️ ▶️ John didn’t like adaptive.
⏹️ ▶️ Marco Yeah, and we got a couple of notes from listeners basically saying, try it again, it’s really good.
⏹️ ▶️ Marco And I tried it again for like, you know, a couple of days and I hated it and I went back.
⏹️ ▶️ Marco Well, in the meantime, I tried it again,
⏹️ ▶️ Marco like another time, and I just left it on.
⏹️ ▶️ Marco And I realized over the last week or so that I have just left it
⏹️ ▶️ Marco on for like months now. And it’s
⏹️ ▶️ Marco fine. I don’t really think about it anymore. I don’t really change modes
⏹️ ▶️ Marco anymore. Almost ever. And so I actually have come around
⏹️ ▶️ Marco that whatever the current, the new version or the current version of adaptive noise cancellation mode
⏹️ ▶️ Marco is indeed good enough to use as my main mode almost all
⏹️ ▶️ Marco the time. I’ve almost never switched it, like since the most recent try. I almost never switch
⏹️ ▶️ Marco it to any other mode now. It just adjusts, and it’s fine.
⏹️ ▶️ Marco So, yeah, sorry Adaptive Mode, I was wrong about you and you’ve won me over.
ATP Movie Club: Star Trek IV
⏹️ ▶️ Casey We released a new ATP member special. So this is, if you recall,
⏹️ ▶️ Casey is one of the perks of being an ATP member, which you can join by going to ATP.FM.Join.
⏹️ ▶️ Casey Once a month, we try really hard. We’ve been pretty consistent so far. Once a month, we do some sort
⏹️ ▶️ Casey of bonus content that is not really canon, I guess you could say. It’s just usually way
⏹️ ▶️ Casey off in a different world, so to speak, in this case, kind of literally. And sometimes we do what
⏹️ ▶️ Casey we call ATP Movie Club. Club, and this month we returned to ATP Movie Club and did Star
⏹️ ▶️ Casey Trek IV, The Voyage Home. John, can you tell me why and how we landed on this particular topic?
⏹️ ▶️ John Greenewald-Marco made a comment on an earlier episode that anyone listening would have thought
⏹️ ▶️ John was Marco making a reference to a famous scene in a Star Trek movie, but it turns out Marco has never seen any Star
⏹️ ▶️ John Trek movies or any Star Trek anything, and I said, oh, well, we have to do that as a member special. We
⏹️ ▶️ John have to show him the movie that he was unknowingly referencing and that Casey and I were making jokes about. And that’s That’s exactly what we
⏹️ ▶️ John did. There’s no other reason than that. And it was fun. So check it out. Star Trek 4,
⏹️ ▶️ Casey Yeah, it did not go the way I expected. And it is definitely worth checking out. And we will
⏹️ ▶️ Casey remind you, like I said, atp.fm slash join. You can do the Syracuse approach, which
⏹️ ▶️ Casey is wrong, and join for a month and then, you know, cancel your membership, which is very easy
⏹️ ▶️ Casey to do. And you can slurp, you know, slurp up all the member specials while during that month.
⏹️ ▶️ Casey then you can, you know, walk away and listen to them all. Or you can do the right thing, which is
⏹️ ▶️ Casey join and just let it ride. You can join us for a month, a year, however long as you
⏹️ ▶️ Casey as you feel, and just enjoy all of the member specials and the bootleg and the ad free
⏹️ ▶️ Casey shows and so on and so forth. So forget
⏹️ ▶️ Casey And over time. Thank you. I almost forgot. So So, hp.fm.join.
CarPlay updates
⏹️ ▶️ Casey John, we talked about the Honda Prologue and how it had been outselling
⏹️ ▶️ Casey equivalent GM vehicles that are based on the same platform. Why do you think that is?
⏹️ ▶️ John A lot of people wrote in, including Matt Rigby, to tell us that they think it’s because Honda has a CarPlay
⏹️ ▶️ John and the GM cars do not. I’m not sure how much of a factor that is. I still feel like the Honda brand is the
⏹️ ▶️ John bigger factor, but that is a factor. The GM ones don’t have CarPlay because GM
⏹️ ▶️ John wants that subscription revenue and does not want to have CarPlay. So
⏹️ ▶️ John throw that in the mix.
⏹️ ▶️ Casey Yeah, in my head canon, that is 100% the reason. That may not be reality, but in my head canon, that’s
⏹️ ▶️ Casey why. And speaking of CarPlay, Apple is still trying to make fetch happen. They
⏹️ ▶️ Casey insist that next generation CarPlay is still a thing. So reading from MacRumors,
⏹️ ▶️ Casey there was a statement from Apple. We continue to work closely with several automakers, enabling them to
⏹️ ▶️ Casey showcase their unique brand and visual design philosophies in the next generation of CarPlay. Each car brand will
⏹️ ▶️ Casey share more details as they near the announcements of their models that will support the next generation of CarPlay. So MacRumors
⏹️ ▶️ Casey adds, Apple also remains committed to its current CarPlay platform and said it is available in over 98% of new cars
⏹️ ▶️ Casey sold in the US over the past few years. Apple previously said committed car makers included
⏹️ ▶️ Casey Acura, Audi, Ford, Honda, Infinity, Jaguar, Land Rover, Lincoln, Mercedes-Benz, Nissan, Polestar, Porsche, or
⏹️ ▶️ Casey excuse me, Porsche, Renault, and Volvo. In December, 2023, Aston
⏹️ ▶️ Casey Martin and Porsche previewed their next generation CarPlay designs, but have yet to deliver. It is
⏹️ ▶️ Casey unclear which car makers are currently working with Apple. Why isn’t Ferrari on that list? Isn’t Q still
⏹️ ▶️ John Yeah, I don’t know. Is that the list of next generation cars? It’s gotta be the list of next generation CarPlay. I
⏹️ ▶️ John, Marco think they’re
⏹️ ▶️ John committed to it, but like things happen so slowly in the car world. I don’t understand why Apple made all these pronouncements
⏹️ ▶️ John that 2024 was gonna be the year. 2024 was not the year, but Apple just wanted to make a statement,
⏹️ ▶️ John an official statement saying, next generation CarPlay is still a thing and it’s still coming someday.
⏹️ ▶️ John That’s the nature of the name next generation. It’s always off in the distance. It’s the next, it’s not the current generation. It’s
⏹️ ▶️ John the next generation. Someday it will arrive and you’ll be sure to hear about it here when
⏹️ ▶️ John the first car ships with it because it’s definitely going to be weird. But right now Apple says it is still a thing. So just
⏹️ ▶️ John be patient, I guess. It’s the race between next generation CarPlay and a new monitor.
⏹️ ▶️ Casey Now, John, can we get a commitment from you that if enough members join at atp.fm
⏹️ ▶️ Casey slash join that you will purchase the first car with the CarPlay 2
⏹️ ▶️ Casey situation within it?
⏹️ ▶️ John I don’t think we have enough people who listen to the show to make that possible, even if they all converted into members, so.
⏹️ ▶️ John, Casey Well, it’s not
⏹️ ▶️ Casey, John attitude, this is never gonna work.
⏹️ ▶️ John Yeah, it’s expensive, but I’ll definitely read articles about it and watch YouTube videos about it and talk about it on the show.
⏹️ ▶️ Casey We all know that even if somebody handed you a million dollars with which to buy a car,
⏹️ ▶️ Casey you are incapable of buying anything that is not Honda.
⏹️ ▶️ John If the only thing I could do with the million dollars was buy one of these cars, I would do it.
⏹️ ▶️ Casey Yeah, you would wait for the Honda version though. You would not buy anything but a Honda.
⏹️ ▶️ John One of these days. Well, Honda’s got their ASIMO operating system. I don’t know what they’re doing with that. Like, it’s not
⏹️ ▶️ John even clear. Are they using Android Automotive and ASIMO is just something like on top of it or alongside it? Anyway,
⏹️ ▶️ John we’ll find out when the Acura RDX EV comes out with the ASIMO operating system in 2026,
⏹️ ▶️ John if I recall correctly. Yeah,
Hand-held gaming PCs
⏹️ ▶️ Casey Tell me about handheld gaming PCs. Is this something that’s relevant to your world?
⏹️ ▶️ Casey Are you interested in this? Actually, this might even be better as a Marco question. I don’t know.
⏹️ ▶️ John Acura RSX. Sorry for that.
⏹️ ▶️ John talked about this in overtime in a past episode where we talked about the Switch 2. And
⏹️ ▶️ John I mentioned that there has been a lot of increased activity in the realm of handheld gaming
⏹️ ▶️ John PCs because the technology is available for it. You can make low-power, high-performing
⏹️ ▶️ John processors. can play quote unquote PC games pretty well, well enough to be on a little
⏹️ ▶️ John screen that you can hold on your hand. Another reason why more of these are appearing
⏹️ ▶️ John is from an article from the Verge from earlier in the month. Valve will officially let you install
⏹️ ▶️ John SteamOS on other handhelds as soon as this April. So Steam Deck is a Valve’s handheld thing and it runs
⏹️ ▶️ John a variant of Linux that has some libraries to let it run Windows games called SteamOS and Valve has said for a while
⏹️ ▶️ John that they were always supposed to license SteamOS to other manufacturers, now they’re doing that. Lenovo
⏹️ ▶️ John is going to ship the first third-party SteamOS handheld in May, and supposedly it will let people
⏹️ ▶️ John install SteamOS on other handhelds even sooner than that thing ships. So Valve is making a play here to
⏹️ ▶️ John say, we don’t just wanna sell the Steam Deck, we wanna be like the Microsoft of handheld gaming PCs.
⏹️ ▶️ John If any, quote-unquote, PC manufacturer wants to make a handheld gaming platform, and there’s a bunch of them, I think Asus makes
⏹️ ▶️ John one, Lenovo’s gonna make one, there’s a couple of other manufacturers that I can’t remember, Like, and if you don’t want to
⏹️ ▶️ John like run, literally run Windows on it, you can run SteamOS, which is just Linux that runs Windows games with that as
⏹️ ▶️ John a compatibility layer or whatever. We’ll see how this works for them. I know the Steam Deck is very popular, but
⏹️ ▶️ John not as popular as the switch to, I imagine, but yeah, it could be like, you know, so this is something we kind of accept in
⏹️ ▶️ John the realm of games that you play while sitting in a chair that there’s PC gaming and then there’s console
⏹️ ▶️ John gaming. Uh, and there’s sort of a rivalry there, but it’s like, oh, well, console gaming, there’s
⏹️ ▶️ John a handful of consoles, very few, or as time goes on, it seems. But PC is a million PCs.
⏹️ ▶️ John You can build build your own game. You see, you can buy gaming PCs, right? We can get any old PC. And this is trying
⏹️ ▶️ John to make that happen in the world of handheld as well, because historically it’s like, well, handheld, there’s
⏹️ ▶️ John no equivalent handheld. You get it from one of the console makers or that’s it. Like there’s no handheld PC.
⏹️ ▶️ John But now they’re like, oh, well, you can get a switch or maybe that weird PlayStation thing that just remote plays to your PlayStation
⏹️ ▶️ John four or five, or you can get any one of these umpteen handheld gaming PCs.
⏹️ ▶️ John The twist here is that if Valve has its way, they’ll be running Linux instead of Windows, but this will be running quote
⏹️ ▶️ John unquote Windows games. So as usual, everything in the PC world is a little bit of a mess, but also
⏹️ ▶️ Casey kind of exciting. Very good. you
Leather case update! …ish
⏹️ ▶️ Casey Uh, apparently Ryan London is out to troll you. Can you tell me more about this?
⏹️ ▶️ John So excited. Uh, there was, someone had sent me something that they talked to. I think I mentioned on the show, they talked to Ryan London back
⏹️ ▶️ John in December at through customer service and they, and the customer service person said, Oh, I know you want one of
⏹️ ▶️ John our leather cases that has the little Sapphire button for the camera control. Well, we’re totally going to make that should be ready
⏹️ ▶️ John by mid December. That didn’t happen. Right. But here we are in mid January and low and beer at the end of January.
⏹️ ▶️ John And lo and behold, the Ryan London website has a big banner on the top. It’s like, hey, here it is. Leather cases with
⏹️ ▶️ John a Sapphire button for the camera control. New, new, click through, buy one now. And I did exactly that.
⏹️ ▶️ John And I clicked through and bought so quickly that I didn’t realize until I saw the receipt screen that
⏹️ ▶️ John what they’re selling with the Sapphire camera control button is their variant of the leather case that has a
⏹️ ▶️ John metal ring around the cameras. And I do not like that. So I had to cancel that order.
⏹️ ▶️ John I’m like, well, I’ll just cancel that order and I’ll go buy the right one. It’s very easy to get confused on the website.
⏹️ ▶️ John Some of the pictures look very similar, but no, I don’t want them with the metal ring I want the one that just
⏹️ ▶️ John has smooth leather smooth leather lump around the cameras It’s looks the same as the bull strap
⏹️ ▶️ John one and the 17 other manufacturers that sell the same case under different brand names but unfortunately
⏹️ ▶️ John Ryan London is not selling the one with the leather lump and the Sapphire thing the only one that has
⏹️ ▶️ John a sapphire button is the one with the metal ring around right now I’m sad and they are teasing me. I was so excited
⏹️ ▶️ John for a moment and then I wasn’t so I’m well first of all I’m glad somebody is, you know, they’re
⏹️ ▶️ John not just gonna say, well, we’ll wait until the iPhone 17, no sense in changing our cases now. They actually are changing their case for the 16.
⏹️ ▶️ John And you know, I’m gonna have the 16 for another like year and a half, right? So I’m
⏹️ ▶️ John ready to buy a second case for this thing, but today is not that day. Ryan London,
⏹️ ▶️ John you fooled me. I’m so sorry for your loss. I don’t understand the connection. Like
⏹️ ▶️ John why only the one with the metal ring? Is it just didn’t get around to the other one? Do they have too much of the ones without the metal ring in stock
⏹️ ▶️ John and they haven’t sold through it? I don’t know what the deal is, but please, Ryan London, convert all your cases.
⏹️ ▶️ John Don’t, like it’s an option. Like when you go to the one with the metal thing, you can have personalization, yes, no,
⏹️ ▶️ John and camera cutout or camera control, either a cutout or a Sapphire crystal button. Sapphire crystal button should be an option
⏹️ ▶️ John to all your cases, please make that. So I will buy one immediately.
⏹️ ▶️ Casey But this doesn’t matter anyway, because it’s not a naked bottom or whatever you call
⏹️ ▶️ Casey, John it. No, it is,
⏹️ ▶️ John it totally, it is, right?
⏹️ ▶️ Casey Not the one you put in the show notes.
⏹️ ▶️ John Well, maybe that’s another reason I shouldn’t buy this one. I’m
⏹️ ▶️ Casey looking at the black one, in the link you put in the
⏹️ ▶️ John document. Yeah, yeah, the metal one, I think the metal one doesn’t have the, maybe that’s another problem with the metal one that I totally didn’t
⏹️ ▶️ John notice. Yeah, this is no good for you. I know, but I don’t want the metal thing anyway.
⏹️ ▶️ John, Casey Like, I just want
⏹️ ▶️ John the one that doesn’t have, anyway, I’m waiting, I’m still waiting. They have the technology, they have the
⏹️ ▶️ John buttons, they can make them, just not yet.
⏹️ ▶️ Casey Yeah, it’s definitely the entire case industry that’s wrong, not you.
⏹️ ▶️ John No, everyone wants the Sapphire buttons, it’s just that the case manufacturers couldn’t make them, you know, or didn’t know how to make them
⏹️ ▶️ John yet, And so they all did cutouts and now they’re switching. Yeah, they still sell the naked bottom one. It just cut
⏹️ ▶️ John out as the only option on it right
⏹️ ▶️ Casey now. I was mostly referring to the naked bottom part, but that’s okay.
Follow-up: John’s RAM
⏹️ ▶️ Casey Speaking of being wrong about things, hey-oh, tell me, how much RAM do you have in your computer again?
⏹️ ▶️ John More than 192 megabytes. 192 megabytes would have been a lot back in the
⏹️ ▶️ John, Marco classic Mac OS
⏹️ ▶️ John days on my platinum-colored plastic Macs. But alas, on an episode where I
⏹️ ▶️ John was talking about the performance of my powerful computer
⏹️ ▶️ John trying to scroll a list of items, I kept saying that it had either 192 megabytes of RAM, which which is way too little, or 192 gigabytes of RAM,
⏹️ ▶️ John which is way too much. In fact, my Mac has 96 gigabytes of RAM.
⏹️ ▶️ John I very often forget the exact number because as I mentioned when I ordered the computer and later when I talked about it,
⏹️ ▶️ John 96 gigs of RAM at the current point in time is enough for me to essentially forget that I even have RAM.
⏹️ ▶️ John Nothing I do exhausts it or stresses it. I can run the computer for a month at a time.
⏹️ ▶️ John Oh, look at activity monitor, no swap is in use. 96 gigs is adequate for my
⏹️ ▶️ John current needs, but I do not have 192 gigabytes and 192 megabytes would not be enough. Fair enough.
John’s scrolling updates
⏹️ ▶️ Casey All right, tell me about what’s going on with your scrolling adventure.
⏹️ ▶️ John Last time we were talking about AppKit versus SwiftUI and then
⏹️ ▶️ John with a side tangent into WebKit to see how smooth that scrolling is. And there was a lot of feedback about that
⏹️ ▶️ John on Mastodon and through email. A lot of people making demo apps saying, I don’t know what your problem is.
⏹️ ▶️ John AppKit is plenty fast for me. Maybe I didn’t emphasize it enough when we discussed it, but once
⏹️ ▶️ John I converted to AppKit, the performance was good. Like it wasn’t bad anymore. Like it was was fine, right? It’s
⏹️ ▶️ John just that WebKit, I was so impressed at how smooth it was, and I still felt like my AppKit version was not
⏹️ ▶️ John quite as smooth as the WebKit version. And that annoyed me. I felt like it should be much
⏹️ ▶️ John better, noticeably better. Instead, it was like the same or maybe slightly worse, right? But it was fine, it was
⏹️ ▶️ John adequate. I wasn’t worried about the performance anymore. But so many people were making demo apps and asking me about things.
⏹️ ▶️ John One of the things somebody mentioned was, hey, are you using NSL or view-based
⏹️ ▶️ John tables? For a little bit of background, NSTableView is a really old class back from the next days, I
⏹️ ▶️ John think, or if not very close to then. And it was originally designed for much less powerful
⏹️ ▶️ John computers with a special class called NSCell that’s used to
⏹️ ▶️ John populate each cell in the table. And NSCell is like a lightweight thing that it’s like, it’s not
⏹️ ▶️ John a full-blown NSView that could have anything in it. It’s just a very small lightweight thing because we know
⏹️ ▶️ John you’re just gonna be a table cell. You’re not gonna be some arbitrary view, you’re probably just gonna show some text or something, like a
⏹️ ▶️ John number or maybe like an image or something. It’s a very limited thing. And as time went on and
⏹️ ▶️ John what was then known as Cocoa development, people were like, oh, NS table view, it’s so annoying with stupid NS cells.
⏹️ ▶️ John I have all these NS views in my app, but I can’t use them with tables. I just want to put my NS views into the tables, but I have
⏹️ ▶️ John to convert everything so it to wedge it into these NS cells. And it’s very limiting and it’s annoying. And that
⏹️ ▶️ John performance enhancement is no longer useful. I wish they would get rid of NS cells. So many, many
⏹️ ▶️ John years ago, eventually Apple said, okay, now you can make NS table views and you can just stick plain
⏹️ ▶️ John old NSViews inside them. You don’t have to deal with NSCell. You’ve got NSViews in the rest of
⏹️ ▶️ John your app. If you just want to put them in a table, you just stick them into a cell and they’ll show there,
⏹️ ▶️ John done and done. So much so that the cell-based NSTableView has been deprecated
⏹️ ▶️ John since I think Mac OS 10.10. I don’t even remember what version that was. Was that Yosemite? I
⏹️ ▶️ John don’t know. Maybe it was earlier than that, but anyway. It’s been deprecated for a while,
⏹️ ▶️ John but people were saying, asking me, are using view-based or cell-based. And I said, I’m using view-based because,
⏹️ ▶️ John like, what year is it? Like, we should be using view-based, right? Apple promoted it to WWDC. They said it’s time
⏹️ ▶️ John to get a ditch NSCell. I’d heard all these bad things about NSCell. And, you know,
⏹️ ▶️ John it’s deprecated. But I said, you know, fine. Like, I’ll do implementation number six or 5.5.
⏹️ ▶️ John I converted my NSTableView to use NSCell instead of NSView,
⏹️ ▶️ John just to see if it would make any difference. And because I, why not, right? And the thing
⏹️ ▶️ John is, it did make a little bit of a difference. And I had to do the thing where like, I had two copies
⏹️ ▶️ John of the app running side by side to be like, am I imagining things? Scroll, scroll. Because like I said, the
⏹️ ▶️ John NSTableView one with NSViews, it was fine. Like you’d look at it and you would think there’s nothing wrong
⏹️ ▶️ John with it. But I’ve been obsessing over it for so long. Now it’s like, scroll this one. Now scroll that one. Now see, does it feel?
⏹️ ▶️ John Can I move the pointer like off the scroll thumb? Like if I shake it real
⏹️ ▶️ John fast, hmm. Are they the same? I was like, I swear the NSL one is a little bit better.
⏹️ ▶️ John And I didn’t know what to do with that. I’m like, well, I’ve re-implemented an NSL and it’s fine. Maybe I’ll just
⏹️ ▶️ John leave it like that. But guess what? NSL is annoying. Everyone was right. They’re
⏹️ ▶️ John like, it makes the code more complicated and annoying and it’s just like, well, but I already did it but
⏹️ ▶️ John it’s kind of gross. Like, what should I do? Like, it is better. I wish it wasn’t better, but I can tell
⏹️ ▶️ John that it’s better. So instead of what I decided to do was spend an entire day trying to figure out why
⏹️ ▶️ John it better? Well, like,
⏹️ ▶️ John, Marco and I wish I was
⏹️ ▶️ John better at instruments. I’m not good at instruments. I’ve watched all the W2AC sessions. I’m still not good at it.
⏹️ ▶️ John, Marco better at instruments. Someone,
⏹️ ▶️ John yeah, like if you are good at the like performance analysis tools, it makes your life so much easier.
⏹️ ▶️ John I know this from my career with performance analysis tools that I did know how to use,
⏹️ ▶️ John but this is not one of them. So I was just like, I’ll figure, I have two, I have the code
⏹️ ▶️ John for both of them here. Surely I can figure out why are you ever so slightly slower than you
⏹️ ▶️ John over here. And so I just went methodically through it and tried to figure out what is making
⏹️ ▶️ John this slower. And I figured it out. I figured out what it was. It was yet another
⏹️ ▶️ John corner of the language, the Swift language that I’m not familiar enough with, combined with a careless mistake.
⏹️ ▶️ John I have a bunch of subclasses because that’s what you do in AppKit, you subclass things. It is a big cascade of subclasses
⏹️ ▶️ John for populating my NSTableViews. And you guys familiar with the whole like designated
⏹️ ▶️ John Yeah, well, you got to call a designated initializer, and you can have convenience initializers that are what you want to do, but
⏹️ ▶️ John you don’t get to pick the designated initializers if you’re using some class that Apple defines. You
⏹️ ▶️ John got to call their initializers, even if they’re initialized with some crap that you don’t care about. Anyway, I had this big cascade of
⏹️ ▶️ John initializers. And being a dutiful little object-oriented person, I was shoving the common functionality down
⏹️ ▶️ John into the base classes so I don’t have to repeat it in every subclass.
⏹️ ▶️ John And one of the things that I had shoved down was setting a very important attribute,
⏹️ ▶️ John the identifier for the object. And it turns
⏹️ ▶️ John out one of my derived classes was calling through a sequence of inits that
⏹️ ▶️ John never hit the init that set the identifier. It was passed into the constructor and passed down, but it was like,
⏹️ ▶️ John because you have to call the designated initializer sooner than you think, or at least I was calling it sooner than I thought I was,
⏹️ ▶️ John I had skipped over that part. So what ended up happening was
⏹️ ▶️ John of my cells, just a dinky little cell, was not getting its identifier set, which
⏹️ ▶️ John meant that every time I needed one of those, it would make me a new one. And it was the one
⏹️ ▶️ John with the little eyeball, like the preview thing. Like it’s literally the same thing every time. There’s
⏹️ ▶️ John no data, right? It’s hard-coded. So it was really fast, but not as fast as
⏹️ ▶️ John not making it. So that was it.
⏹️ ▶️ John I set the identifier, because I’m already passing the identifier. It was there in the constructor. I set the identifier in like the subclass
⏹️ ▶️ John init, even though it’s like duplication and you know, don’t repeat yourself. Well, anyway, I set the initializer
⏹️ ▶️ John because it wasn’t getting set because I wasn’t going through the right init path elsewhere. And then all of a sudden the
⏹️ ▶️ John view-based one was exactly the same as the cell-based one. And I was very happy.
⏹️ ▶️ John I celebrated and I hope I never have to re-implement that view ever again. I threw away the NS
⏹️ ▶️ John cell-based one, reverted to the view-based one, did the two line fix, And
⏹️ ▶️ John honestly, you can’t really tell the difference.
⏹️ ▶️ John you A, B test it, it looks exactly the same as it was before. But I know it’s ever so slightly better, and
⏹️ ▶️ John that’s what matters. And the final thing on this topic, a bunch of people are asking about
⏹️ ▶️ John web kit performance and scrolling performance. And someone pointed
⏹️ ▶️ John me to a blog post about a website where someone wanted
⏹️ ▶️ John to make a web page where you can scroll through every UUID. That’s bananas,
⏹️ ▶️ John like every UUID. Wait,
⏹️ ▶️ Marco how, like just iterating? Just,
⏹️ ▶️ John well, the idea is a webpage and you scroll it and the top is the first UUID and at the bottom is
⏹️ ▶️ John the last one and in between are all the other possible UUIDs between those values. Now, and this is like
⏹️ ▶️ John an example of like, well, how far can you push the whole like, you know, like we mentioned this before, if you’re recycling
⏹️ ▶️ John the cells, it shouldn’t matter how many things there are in the list, like the performance should be the same with 10,
⏹️ ▶️ John a thousand, a hundred thousand, a million, a billion, the performance should be exactly the same all the time, right?
⏹️ ▶️ John And this is kind of a demonstration of that. Now I feel like they cheated because when you go to the website, everyuuid.com,
⏹️ ▶️ John you will see how they cheated. It’s not really scrolling it, it’s really just showing a fixed
⏹️ ▶️ John list of cells with values that change. So, and it’s a fake scroll bar, so.
⏹️ ▶️ John But anyway, the blog post about how they implemented it is fun because obviously there’s no sort of data
⏹️ ▶️ John Like you can generate all of these and I think they’re essentially like sequential or whatever. But I thought it was interesting,
⏹️ ▶️ John interesting enough to be in the show notes. If you wanna see one of the challenges of trying to have
⏹️ ▶️ John a not infinite, but very, very large scrolling list, everyuuid.com.
Phone jacks: sometimes Ethernet!
⏹️ ▶️ Casey With regard to the Ask ATP, the one and only, I believe, Ask
⏹️ ▶️ Casey ATP topic from last week, we were talking about what should you look for if you’re about
⏹️ ▶️ Casey to buy a house and you’re a nerd like us. And we had talked about, you know, whether or not
⏹️ ▶️ Casey things would be pre-wired, you know, like if there would already be Ethernet in the house. And John Hayden writes, my house
⏹️ ▶️ Casey had phone jacks in every room. However, I looked behind and found that they were actually Cat6, which
⏹️ ▶️ Casey is, you know, Ethernet cabling, but only using one or two twisted pairs, which is to say only a subset of the wires
⏹️ ▶️ Casey within it. They all routed outside the home just like the coax. I tried pulling the cables from the outside wall,
⏹️ ▶️ Casey but they were likely stapled down. So I cut them and rerouted them to what is now my server closet. Life
⏹️ ▶️ Casey lesson, if you have phone jacks, see if you have an Ethernet cable behind it.
⏹️ ▶️ John It’s like finding an extra room in your house that you never discovered before. You didn’t realize your whole house is wired
⏹️ ▶️ John with the internet, you just remove one of the phone switches, and there’s an ethernet cable in here. What a heartwarming
⏹️ ▶️ John story that will never happen to anybody else.
Casey’s LED updates
⏹️ ▶️ Casey All right, and I have some updates with regard to my grand notification project. First of all,
⏹️ ▶️ Casey I got a ton of feedback. This is one of those things where I thought I was very clear about
⏹️ ▶️ Casey this, and either I wasn’t or maybe people just didn’t hear. But one way or another,
⏹️ ▶️ Casey it doesn’t really matter. I already have the connection,
⏹️ ▶️ Casey if you will, between the garage door and home assistant. That’s sorted. That was sorted last week. It’s been sorted
⏹️ ▶️ Casey since I went to home assistant. There is a—I was going to say bespoke, I don’t know if that’s really fair—but
⏹️ ▶️ Casey there is a integration with the particular weirdo flavor of Garage Door that
⏹️ ▶️ Casey I have that I’ve been using since I went to Home Assistant a few months
⏹️ ▶️ Casey ago. And one of the funny things about if you look at an integration in the—or
⏹️ ▶️ Casey on the Home Assistant website, you can see how many people have it installed. And
⏹️ ▶️ Casey the integration I’m talking about is the NiceGeo integration. It was introduced
⏹️ ▶️ Casey in Home Assistant 2024.9, and it is used by gentlemen 36
⏹️ ▶️ Casey active installations, of which I am one of them. So not terribly popular, but it works.
⏹️ ▶️ Casey So I just wanted to make it clear. A lot of people were talking about like a RATGO. I don’t recall what the acronym stands
⏹️ ▶️ Casey for, but I think it’s like RATGD or something like that. Anyways, there are many, many mechanisms
⏹️ ▶️ Casey to get a dumb garage door opener into like HomeKit or Home Assistant,
⏹️ ▶️ Casey what have you, but that’s already been solved for me. Also, a lot of people brought up, and this one I am
⏹️ ▶️ Casey pretty sure I didn’t say anything about, a lot of people brought up the ESP32
⏹️ ▶️ Casey as an alternative to like a Raspberry Pi or Arduino. These are exceedingly cheap, Wi-Fi
⏹️ ▶️ Casey enabled, little programmable, basically circuit boards, and they seem
⏹️ ▶️ Casey to be the popular way to do this sort of LED kind of dance that I’m talking
⏹️ ▶️ Casey about. And in fact, there’s a software project called ESPHome
⏹️ ▶️ Casey that is allegedly really, really good at doing LED related stuff and then
⏹️ ▶️ Casey exposing that in Home Assistant. I don’t think I talked about either of these on the show last week,
⏹️ ▶️ Casey but I very justifiably, and I appreciate it, got a lot of feedback from people saying this is what you need to do.
⏹️ ▶️ Casey I also got a lot of people saying to me, just get an LED strip, man.
⏹️ ▶️ Casey And I was like, okay, I don’t want a hundred LEDs. I want
⏹️ ▶️ Casey three. Why would I get an LED strip of, you know, a meter, two meters, three meters,
⏹️ ▶️ Casey four meters, 10 meters, or what have you. I don’t want that. I want three, not three meters.
⏹️ ▶️ Casey Mind you, I want literally three LEDs. And in a conversation on Slack
⏹️ ▶️ Casey with Kiel Olson, he said to me, well, you can just cut it. Wait, wait, I’m sorry. What?
⏹️ ▶️ Casey Yeah, you can just cut an LED strip. If it’s the right kind of strip, you can just cut it. What?
⏹️ ▶️ Casey So it turns out that there is a style of LED strip and the most popular one
⏹️ ▶️ Casey is the WS2812B. I will put a link to the data
⏹️ ▶️ Casey sheet in the show notes. I couldn’t find a better link for like just the thing. You can do an
⏹️ ▶️ Casey Amazon search or what have you if you want to see it. But suffice to say, these are addressable
⏹️ ▶️ Casey LEDs. So imagine a strip of say a hundred LEDs And you can actually target
⏹️ ▶️ Casey any specific LED. And the way it works, and I’m going to butcher the specifics, but the general gist of the way it works is
⏹️ ▶️ Casey the first LED has a data connection and a power connection to whatever’s powering it. And
⏹️ ▶️ Casey it then daisy chains those connections to all the LEDs behind
⏹️ ▶️ Casey it. So it will look at the data, say, is this about me? Okay, then
⏹️ ▶️ Casey I’m going to turn myself on or off or no, pass it along to the next one. And if you cut
⏹️ ▶️ Casey them at the particular places where they allow you to physically cut them, and they’re usually
⏹️ ▶️ Casey labeled, I guess, you can just cut three LEDs off of a strip of 100. I had no freaking clue this
⏹️ ▶️ Casey was a thing. This is blowing my mind. Now I feel like I wanna just play with LED strips because I can.
⏹️ ▶️ Casey I have no purpose really for them, but I just wanna get some and play with them just because I think that’s incredible.
⏹️ ▶️ John You could combine this with Marco’s holiday lights. You can run those LED strips all over his house. What was
⏹️ ▶️ John, Casey the name of that
⏹️ ▶️ John candy where you constantly eat paper because they stick to it. What was
⏹️ ▶️ John, Marco it? Is it dots? Dot candy? No,
⏹️ ▶️ John dots are the chewy things in the box. What the? Ribbon
⏹️ ▶️ Casey candy? No, no, no, it was something dots. I know what you’re thinking of. It was a row of like four or five.
⏹️ ▶️ John Yeah, it was like a piece of paper and had hard little pieces of sugar stuck to it. And the idea was that you would scrape the
⏹️ ▶️ John little hard pieces of sugar off the paper with your teeth or your fingers, but you inevitably end up eating paper. They were a terrible
⏹️ ▶️ John candy and I’m glad I can’t remember their name.
⏹️ ▶️ Casey Oh, they were the best, but they were also terrible. You’re not wrong about that.
⏹️ ▶️ John If you like eating paper, they’re great.
⏹️ ▶️ Marco It was an activity and a candy in one.
⏹️ ▶️ John Like a fun day. Is the activity picking paper out of your teeth? Is that
⏹️ ▶️ John, Marco the activity? Yeah,
⏹️ ▶️ Casey pretty much. Yeah. Oh, cheesy peasy. In any case, so that’s the thing. I had no idea that was the thing.
⏹️ ▶️ Casey I think everyone who was telling me, oh, just get a 2812, yada, yada, yada. They all just
⏹️ ▶️ Casey assumed that I knew that you could just cut off a slice of LEDs. I had no
⏹️ ▶️ Casey idea until Kiel said something. So now I know.
⏹️ ▶️ Marco Yeah, I didn’t I knew that you could cut led light strips that like, you know, just were regular ones that don’t have like
⏹️ ▶️ Marco individually addressable You know components I figure you know Cuz there’s lots of led light strips that are just like long
⏹️ ▶️ Marco strings of LEDs and they have little Markings on where you can cut them and it’s
⏹️ ▶️ Marco, Casey fine Yeah, I had no idea but
⏹️ ▶️ Marco yeah, but I didn’t realize that that extended to be like addressable cool kind as well
⏹️ ▶️ Casey Yeah, and then eat real-time follow-up John I think
⏹️ ▶️ John, Casey candy buttons that’s
⏹️ ▶️ John or and this is a great one. Pox. Would you like some pox? No, thank you.
⏹️ ▶️ John, Casey I do not like some pox. I will
⏹️ ▶️ John pass on the pox. A pox on both of your teeth with paper that’s going to get you.
⏹️ ▶️ Casey Oh man, that’s good. But I love the, I mean, everything you said is correct. You constantly had
⏹️ ▶️ Casey paper coming with them, but God, I love them so much. They were one of my favorites. All right. Just a little bit more with regard to the notification
⏹️ ▶️ Casey stuff. I ran all this by the historical commission for the second time. I
⏹️ ▶️ Casey had kind of done a casual flyby with the historical commission and I had forgotten that the historical
⏹️ ▶️ Casey commission had strong opinions about this and the historical commission has informed
⏹️ ▶️ Casey me that this is not going to be a project that I will be pursuing any further or at least not
⏹️ ▶️ Casey establishing it in the kitchen. And it’s too bad because as I think I’d said to you last time, like there’s, and
⏹️ ▶️ Casey I don’t want to to share this photo publicly, but in our slack I shared with the boys, there’s
⏹️ ▶️ Casey an empty outlet like almost at eye level, right in the kitchen, right where we pass
⏹️ ▶️ Casey by all the time, and would have been a perfect spot for it. And the historical commission has informed
⏹️ ▶️ Casey me this is not going to be passing. My, not my permission slip,
⏹️ ▶️ Casey my application has been denied. And most of my privileges might have been revoked. That’s to be
⏹️ ▶️ Marco Yeah, I think I would side with the historical commotion on this one. What
⏹️ ▶️ Marco, John actually went there?
⏹️ ▶️ John Because it is literally a blank wall plate. Like, there’s no, it’s like, it’s the same wall plate you would see if there was an outlet or a
⏹️ ▶️ John switch, but there’s nothing. It
⏹️ ▶️ Casey piece of glass. Well, I just installed that blank. What was there until literally a week or two ago, which kind of what started
⏹️ ▶️ Casey this whole process, was that used to be an RJ11 jack, a telephone jack, right
⏹️ ▶️ Casey, John there. Oh, there you go. That makes
⏹️ ▶️ Casey sense. And, you know, the house was built in the late 90s. And so you want the kitchen phone. And that And that was, I mean, you can’t see you
⏹️ ▶️ Casey guys. Well, none of the listeners can see this photo and I apologize, but the boys looking at the photo and behind
⏹️ ▶️ Casey me as I took the pictures, like the kitchen and the kitchen table and whatnot. But it’s the perfect spot for this like
⏹️ ▶️ Casey notification system. Unfortunately, the historical commission has denied my application and my building permit has
⏹️ ▶️ Casey, Marco prejudice. That’s what I was looking for, thank you.
⏹️ ▶️ Casey That being said, right near this outlet this
⏹️ ▶️ Casey former RJ11 outlet is a three-gang set of traditional light switches.
⏹️ ▶️ Casey And one of them is the kitchen, one of them is the kitchen table, and one of them is the hallway.
⏹️ ▶️ Casey And a couple of people wrote in, Drew Stevens in particular wrote in and said, what
⏹️ ▶️ Casey about a home seer, H-O-M-E-S-E-E-R, H-S-W-X-300,
⏹️ ▶️ Casey all these names just roll right off the tongue.
⏹️ ▶️ Marco Is that like a GPU?
⏹️ ▶️ Casey Right, exactly. It might as well be, it’s
⏹️ ▶️ Casey It does, that’s true. So this is a, you know, like paddle or decorous style switch
⏹️ ▶️ Casey that has individually addressable LEDs on the side of it. You can even change the colors of them. This
⏹️ ▶️ Casey looks freaking perfect for what I want. And what I can do
⏹️ ▶️ Casey hypothetically is I can replace one of the three switches in this three gang box
⏹️ ▶️ Casey that actually do things. I can replace one of those switches, specifically the kitchen table switch, which not only
⏹️ ▶️ Casey is in the center of this three gang box, which obviously I can move it, but also is a single switch rather
⏹️ ▶️ Casey than, you know, part of a multi-switch setup, a three-way setup or what have you. Anyways,
⏹️ ▶️ Casey I could replace the kitchen table switch with one of these and it would be perfect. And I
⏹️ ▶️ Casey am ready to buy this thing, money no object, and I go to buy it
⏹️ ▶️ Casey and it’s sold out on Amazon. Leo
⏹️ ▶️ John Dion Aren’t you getting a little ahead of yourself here? Have we not learned nothing about the permitting process?
⏹️ ▶️ John, Casey I know I know but I thought maybe
⏹️ ▶️ Casey I could slide it in on the you know get in and slide
⏹️ ▶️ John I think this is going to be noticeable I don’t think this is gonna slide under the radar cuz like especially in the
⏹️ ▶️ John in the pictures on the website These LEDs are not subtle.
⏹️ ▶️ Casey You’re not wrong But I just hear me out though. So here it is I go to Amazon like this gonna be
⏹️ ▶️ Casey freaking great and I don’t remember what the cost was But it was under $100 which granted for a switch is a lot of well,
⏹️ ▶️ Casey you know You can get a $3 like dumb switch and so $100 is a lot of money for switch but this would be perfect.
⏹️ ▶️ Casey But it’s sold out. So I said, okay, fine, I’ll go to the actual home seer website. I will buy one for
⏹️ ▶️ Casey that. It’s sold out there too. So apparently, I cannot get one of these.
⏹️ ▶️ Casey And I’m really sad about it, even though I think you’re right, John, that ultimately the permitting process would
⏹️ ▶️ Casey deny me. What was it? With prejudice? Thank you.
⏹️ ▶️ Casey, John There it is.
⏹️ ▶️ John You got to go through that process before you buy this.
⏹️ ▶️ John speaking of that, how do you think that terminal with no vowels is going to go over? Has that been discussed?
⏹️ ▶️ John Lightly discussed. Marco did the thing where he got, he bought it first and then slid it in and just waited to see
⏹️ ▶️ John if anyone would notice. And I’m wondering if that strategy might not work in your household.
⏹️ ▶️ Casey I don’t know that it would. Uh, it’s been lightly discussed, but the thought of having a
⏹️ ▶️ Casey automatically updatable calendar was met with enthusiasm because I think I’ve talked about this in the past.
⏹️ ▶️ Casey I print at the beginning of the month, a physical calendar and put it on the kitchen, uh, or excuse me, on the refrigerator,
⏹️ ▶️ Casey just because I like, both of us actually like having a vague notion of what we’re doing and we can
⏹️ ▶️ Casey always write stuff on there if we want. The system of record is 100% our shared Apple calendar,
⏹️ ▶️ Casey but having a glanceable thing in the kitchen is kind of nice. Now, I don’t think we’ll be mounting the terminal on the refrigerator,
⏹️ ▶️ Casey although I guess we technically could since it’s battery powered. I don’t know. We’ll see what happens when we get there.
⏹️ ▶️ Casey But one way or another, this is irrelevant because I can’t put my hands
⏹️ ▶️ Casey on one of these home sear HSWX300s. And a ton of people recommended
⏹️ ▶️ Casey the Inovelli Smart Dimmer. And in fact, people are recommending that in the chat right now. This
⏹️ ▶️ Casey is way more popular in terms of the number of recommendations I got. However,
⏹️ ▶️ Casey it’s not exactly what I want. It’s close, but it’s not exactly there. And it does have a light, like
⏹️ ▶️ Casey an LED light bar on the side of it. It’s a switch with a light bar on the side as opposed to individual LEDs.
⏹️ ▶️ Casey And it wasn’t until, well, Drew Stevens, who was the same one who recommended the Holmes here, recommended this as well.
⏹️ ▶️ Casey And then David Minima pointed out to me, no, no, no. Yes, it’s presented as one LED
⏹️ ▶️ Casey bar, but there’s actually several addressable LEDs in there. But the thing of it
⏹️ ▶️ Casey is, is that they’re all in like one shroud or lens or what have you. So I don’t really love this,
⏹️ ▶️ Casey although a lot of people reached out and said the Inovelli is very, very good. So we’ll see what happens.
⏹️ ▶️ Casey Again, the Historical Commission, I think, has opinions about this. And naturally, I want to appease the Historical Commission.
⏹️ ▶️ Casey I recommend that path, yes. So I think this is all for naught anyway.
⏹️ ▶️ Casey One diffuser, David Shop writes, and I think that’s the correct term for it. Thank you. But as a couple of quick final
⏹️ ▶️ Casey notes, because I can’t leave well enough alone and because my permitting
⏹️ ▶️ Casey has been denied, I thought, well, what’s the next best thing? Well, I can put this junk in my menu bar, because
⏹️ ▶️ Casey why wouldn’t I? And I was looking at Swift Bar, which is my
⏹️ ▶️ Casey thing that will let me put random stuff in my menu bar, which I really, really love. I’ve talked about it many
⏹️ ▶️ Casey times in the past. And one of the advantages of putting all this data on a, you know,
⏹️ ▶️ Casey MQTT setup, a PubSub sort of setup, is that anything can subscribe to these,
⏹️ ▶️ Casey you know, basically, is it bad or is it good messages? And
⏹️ ▶️ Casey the way SwiftBar works though, is that it pings away, like
⏹️ ▶️ Casey every five seconds or ten seconds or five minutes or ten minutes or what have you, it makes another request and gets the latest version in
⏹️ ▶️ Casey the world. And yeah, I could like just do this every second or two, but
⏹️ ▶️ Casey it seems so wasteful when the whole idea of MQTT is you say, I would like to get updated and
⏹️ ▶️ Casey then it sends you updates. Well, come to find out, and I didn’t realize this until earlier today, SwiftBar
⏹️ ▶️ Casey actually has the idea of, shoot, I forgot the name of it and I’m trying to stall for time while I
⏹️ ▶️ Casey look at it, but it has the idea of streamable, there we go, streamable plugins
⏹️ ▶️ Casey where instead of just running a script and then getting a result and walking away, it will actually
⏹️ ▶️ Casey start a new thread and run a script and wait for it to update standard out. And when it updates standard
⏹️ ▶️ Casey out, it’ll update your menu bar. And it took me a little bit of time to figure out how to get this right. And I had to
⏹️ ▶️ Casey actually engage with the author of Swift Bar who went above and
⏹️ ▶️ Casey beyond. This is Alec Maznov, went above and beyond and it seemed to install their
⏹️ ▶️ Casey own MQTT setup just to test my IBS, incredible customer service, especially
⏹️ ▶️ Casey for a free app, but even in general, incredible customer service. And anyways, and between
⏹️ ▶️ Casey Alex and myself, we got it squared away. So now as MQTT messages come in, I will occasionally
⏹️ ▶️ Casey see an envelope appear in my menu bar if I need to check the mail. I will see a little charging
⏹️ ▶️ Casey car in my menu bar if Aaron’s car is charging, and I will always see the state of the garage door because that’s
⏹️ ▶️ Casey what I want to do, either open or closed. And that makes me very happy.
⏹️ ▶️ Casey, John And you can all judge
⏹️ ▶️ Casey finally, finally, we talked about, hey, when I leave the house, I might take all these Caseta switches and
⏹️ ▶️ Casey so on and so forth. And we had a brief conversation between the three of us as to whether or not I could.
⏹️ ▶️ Casey And as per a handful of people, but I think the first one I saw was Mark Bramhill, reached out and said, hey, according
⏹️ ▶️ Casey to their realtor, and obviously the rules may vary where you are, anything you show, and the phrase
⏹️ ▶️ Casey that Mark quoted from his realtor was anything affixed to the wall has to stay.
⏹️ ▶️ Casey So if you show the house with Caseta switches, guess what? going to be the new owners, Caseta Switches.
⏹️ ▶️ Casey But if you rip them out before you start showing the house, fair game, baby. So that makes sense. But I didn’t
⏹️ ▶️ Casey know that and I didn’t have that summarized until Mark had told me about it. So there you
⏹️ ▶️ John go. Didn’t some Germans also say that they take everything with them?
⏹️ ▶️ John, Casey Oh, yes, that’s right.
⏹️ ▶️ John Take their kitchen appliances, their kitchen cabinets, and one of them said their kitchen countertops.
⏹️ ▶️ John What good are the countertops going to do you unless every countertop is the same size and shape? Like, you
⏹️ ▶️ John know, these are our marble countertops. We’re taking them with us. Like they’re not going to fit in your new place at all unless
⏹️ ▶️ John it’s exactly the same as your old. But
⏹️ ▶️ Marco anyway, I mean, you are talking about Germany, like maybe literally all of their counters are exactly the same
⏹️ ▶️ Marco size and shape. Like there’s no other place that I would think that’s a possibility. But in Germany, maybe, I
⏹️ ▶️ John don’t know. Oh yeah. Like you get a, you get a new place and the only thing in the kitchen is just like bare walls, bare floor and
⏹️ ▶️ John like some like electrical wires dangling out somewhere. Maybe like a loose gas line. I don’t know. I don’t know what’s going on over
⏹️ ▶️ John there in Germany, but anyway,
⏹️ ▶️ Casey this is bananas to Americans. And obviously in America, you know, you can write into a contract, you know, I’m taking the
⏹️ ▶️ Casey stove or I’m taking this or I’m taking that. But generally speaking, normally
⏹️ ▶️ Casey kitchen fixtures, well, the fixtures particularly like cabinets and countertops and whatnot, those always stay. I’ve never
⏹️ ▶️ Casey heard of those going. But even like stoves and ovens and in a lot of cases, microwaves,
⏹️ ▶️ Casey if they’re, you know, built-ins, all of those tend to stay.
⏹️ ▶️ John The sellers usually don’t want them. The sellers usually say, please take our crappy old, you know, we’re not taking
⏹️ ▶️ John our fridge with us. We’re not taking your stove. They’re yours if you want them. Like they’re just because they assume they’re they’re going to get new stuff in their
⏹️ ▶️ John new place. And yeah, but anyway, different strokes.
DeepSeek
⏹️ ▶️ Casey All right, so there’s been a big brouhaha over the last several days about DeepSeek.
⏹️ ▶️ Casey And we’re going to read, probably I’ll be reading quite a lot of different things for better and for worse.
⏹️ ▶️ Casey But DeepSeek is a new AI thing
⏹️ ▶️ Casey from this Chinese company that I don’t think any, well, not literally of course, but most Americans hadn’t heard of.
⏹️ ▶️ Casey I certainly hadn’t heard of it. And they released some stuff, and we’ll talk about what here in a second. And
⏹️ ▶️ Casey it done shooketh the American stock markets and a lot of big tech here in America.
⏹️ ▶️ Casey And a lot of big tech took a bath in the stock market over the last week. So reading
⏹️ ▶️ Casey from Ars Technica, on Monday, NVIDIA stock lost 17% amid
⏹️ ▶️ Casey worries over the rise of the Chinese AI company DeepSeek, whose R1 reasoning model
⏹️ ▶️ Casey stunned industry observers last week by challenging American AI supremacy with a low-cost freely
⏹️ ▶️ Casey available AI model, and whose AI assistant app jumped to the top of the iPhone app store’s free apps category
⏹️ ▶️ Casey over the weekend, overtaking chat GPT. The drama started around January 20th
⏹️ ▶️ Casey when the Chinese AI startup DeepSeek announced R1, a new simulated reasoning, or SR, model
⏹️ ▶️ Casey that it claimed could match OpenAI’s O1 reasoning benchmarks.
⏹️ ▶️ Casey There are three elements of the DeepSeek R1 that really shocked experts. First, the Chinese startup appears to have trained the model
⏹️ ▶️ Casey for only about $6 million that’s American, reportedly about 3% of the cost of training
⏹️ ▶️ Casey O1. And as a so-called, quote-unquote, side project, while using
⏹️ ▶️ Casey less powerful NVIDIA H800 AI acceleration chips due to the US export restrictions
⏹️ ▶️ Casey on cutting-edge GPUs. Second, it appeared just four months after OpenAI announced O1 in September
⏹️ ▶️ Casey of 24. And finally, and perhaps most importantly, DeepSeek released the model weights for free with an open
⏹️ ▶️ Casey MIT license, meaning anyone can download it, run it, and fine-tune or modify it.
⏹️ ▶️ John That’s a lot. I mean, I never would have thought Nvidia stock would ever go down again. I just thought it would,
⏹️ ▶️ John they would just go higher forever. And there was nothing about their stock price that was irrational or bubble-like,
⏹️ ▶️ John of videos where they’re still, they’re still the best GPU company out there. But yeah, I think they were a little bit overinflated and, uh,
⏹️ ▶️ John companies taking a hit on this, you know, Nvidia, uh, taking a hit, maybe taking a hit on this a little weird because
⏹️ ▶️ John as this story you just read alludes to, uh, there are or were I think I’m not sure
⏹️ ▶️ John if they’re still in effect, who even knows. But anyway, there were U.S. export restrictions for the most powerful
⏹️ ▶️ John GPUs to China, and the idea being let’s not
⏹️ ▶️ John give our best technology to China because then they’ll develop a I. And I don’t know, take over the world with their AI instead
⏹️ ▶️ John of our AI. Whatever. It’s just like USA, USA. We want to do everything ourselves with our own technology.
⏹️ ▶️ John We don’t want to export it to China. They can buy the crappy old like last year,
⏹️ ▶️ John a year before our previous generation model, but we’re not going to sell them the best stuff. And presumably
⏹️ ▶️ John that will help maintain the U S lead in AI because open AI is an American company and a lot of the
⏹️ ▶️ John other big AI startups are also American companies. Uh, and the Chinese company said, uh,
⏹️ ▶️ John no, we’ll do the same thing you’re doing, but for less money and with crappier hardware. And they
⏹️ ▶️ John did. And it was very upsetting to the stock market because they said, well, I guess,
⏹️ ▶️ John Uh, all those export restrictions did not have the intended effect. And I guess,
⏹️ ▶️ John uh, what open AI is doing is not that unique. We’ve talked about this in many past episodes.
⏹️ ▶️ John Uh, the phrase you will hear all the time, which is annoying is, uh, uh, what kind of moat do
⏹️ ▶️ John the AI companies have? Is there anything about what open AI is doing that makes it
⏹️ ▶️ John special and unique that makes competitors not able to compete? And I think we’ve all said in all past shows, not really,
⏹️ ▶️ John because. Facebook has its open source Lama models. Apple’s got its foundation models. Like
⏹️ ▶️ John the foundation of all of these things is the large language model scientific
⏹️ ▶️ John papers and the study of how to create them. That is all public knowledge. So anybody can make
⏹️ ▶️ John one of these things. And the question was, is there some kind of special sauce that OpenAI
⏹️ ▶️ John has? Okay, well, the technology everybody knows, but we do it in a better way than anyone else.
⏹️ ▶️ John And therefore we have a moat. We’re the best at it. Chat GPT is the best. Everyone’s
⏹️ ▶️ John got a large dynamic models, but we’re just a little bit better than all of them. And that’s why we need $500 billion or whatever
⏹️ ▶️ John to build new data centers to train the next model, blah, blah, blah, blah, blah. And here comes this Chinese company saying, well, we
⏹️ ▶️ John read all the same papers and we have crappier GPUs and we spent less money, but our thing is
⏹️ ▶️ John basically as good as yours open AI. So what do you think of that?
⏹️ ▶️ Marco Not only that, but like, you know, running inference on our thing, which is like, you know, executing the AI models and using
⏹️ ▶️ Marco them for everybody else is way cheaper than you’re thinking.
⏹️ ▶️ John Everything’s cheaper. Everything about it is cheap. It was cheaper to train and cheaper to run, to actually use.
⏹️ ▶️ John And that’s one of the reasons that one of these stock prices that did not take a hit was Apple, because I guess the theory
⏹️ ▶️ John that like, well, if inference becomes cheaper and Apple likes to do lots of on-device AI,
⏹️ ▶️ John that’s good for Apple. Now, it’s not like Apple is using DeepSeek, like in their operating system, but just
⏹️ ▶️ John conceptually, if the cost of inference goes down for equal performance,
⏹️ ▶️ John I guess that benefits Apple because they’re doing a lot of inference on device or whatever, but we’ll see. I think
⏹️ ▶️ John like this, this whole kerfuffle is just kind of, I feel like a correction to some inflated stock prices, but in general,
⏹️ ▶️ John being able to do the thing better and for less money with less power is what we expect
⏹️ ▶️ John with technological progress. What we don’t expect is like every year we will take even more power
⏹️ ▶️ John and you know, like we think things to get better, but keep in mind the deep seek is not like massively
⏹️ ▶️ John better than open AI. roughly about the same with some caveats that we’ll get to in a little bit.
⏹️ ▶️ John But the whole point is, yeah, it’s the same, but cheaper and better and lower power and blah, blah, blah. Right? And I’m like, great,
⏹️ ▶️ John that’s what I expect. I expect like, you know, the MacBook Air that you can get
⏹️ ▶️ John now should be roughly the same performance as like an old MacBook Pro,
⏹️ ▶️ John right? But lower power and better, like I expect that to happen. But I guess people were startled that it happened so quickly, especially
⏹️ ▶️ John since open AI has always just been making noises of like the only way we can surpass a one to make the next generation
⏹️ ▶️ John is for you to give us billions more dollars. Uh, and yeah, I apparently even just to do
⏹️ ▶️ John Oh one caliber stuff, you did not need that much money. You just need to be a little bit more clever. And the fun
⏹️ ▶️ John thing about the cleverness, which we’ll get to in a little bit is kind of like the, the, the
⏹️ ▶️ John saying that like constraints lead to better creative output because this Chinese company had to
⏹️ ▶️ John work with previous generation hardware, they were forced to figure out how to extract
⏹️ ▶️ John the maximum performance from this They had to make compromises. They had to do approximations.
⏹️ ▶️ John They had to come up with new technologies that said, we can’t do it the way OpenAI did it. We don’t have the money. We don’t have the
⏹️ ▶️ John time. We don’t have the technology. We need to find out a way to essentially get the same
⏹️ ▶️ John result, but doing way less work. And apparently, they did.
⏹️ ▶️ Marco Yeah, and I think we’ve been in a pretty long span
⏹️ ▶️ Marco of technology companies and technology stocks and technology stocks and technology
⏹️ ▶️ Marco earnings and profits being pretty mature until the
⏹️ ▶️ Marco big LLM and AI boom of the last couple of years. And I think it was
⏹️ ▶️ Marco easy for us to assume that the technology industry
⏹️ ▶️ Marco is stable now. But in the past, there have been periods where the tech
⏹️ ▶️ Marco industry was not so stable. There were big boom times. There were big technological changes.
⏹️ ▶️ Marco like the birth of the personal computer was a pretty big deal, shook a lot of stuff up. Then
⏹️ ▶️ Marco later on, the internet for home users really shook a lot of stuff up. The consumer
⏹️ ▶️ Marco web came on with e-commerce. That shook a lot of stuff up. And then mobile
⏹️ ▶️ Marco happened, and that shook a lot of stuff up.
⏹️ ▶️ John And in between, though, there was periods of stability. There was a seemingly long period of stability where it was Windows desktop
⏹️ ▶️ John PCs running in Dell CPUs, and then that same stability period, but with the internet, and then mobile came.
⏹️ ▶️ John And so there’s always these kind of periods where you’re like, yes, this is just the way computers are. You buy a personal computer, you put things
⏹️ ▶️ John floppy disks on them and you run them, and that’s all there is. And it seems like that’s going to be it. But then the next
⏹️ ▶️ John inflection point comes and there’s chaos and there’s winners and losers, and then there’s another stable period. And so, yeah, I think we were
⏹️ ▶️ John in a pretty long stable period with we were currently in the PCs exist, mobile exists, the Internet
⏹️ ▶️ John exists. And then we’re in kind of mostly stable period. And then LL Ames came and said, now
⏹️ ▶️ John nobody knows what’s up. Everyone’s scrambling who are going to be the winners, are going to be the losers. And yeah, we’re in the middle of that right now.
⏹️ ▶️ Marco Yeah. And I think, you know, the, the LLM based AI phase, I mean, I guess we’re just
⏹️ ▶️ Marco calling it AI now. So I’m going to stop saying LLMs, even though it’s more correct, but whatever. So the, the
⏹️ ▶️ Marco AI era is upon us now. And all of these things that have been stable
⏹️ ▶️ Marco are suddenly not so stable. It does create bubble dynamics, bubbles do burst. It does create
⏹️ ▶️ Marco a bunch of volatility in, in every market that, that is touched by it, which is in this case, many markets.
⏹️ ▶️ Marco So we have to assume like, you know, I mean, look, geez, like Google now has
⏹️ ▶️ Marco disruption to their core search product for the first time ever. Like in their entire
⏹️ ▶️ Marco existence they have like more disruption and more threat to Google search
⏹️ ▶️ Marco than we’ve ever had before. You know, this is a big deal and
⏹️ ▶️ Marco in fact, I mean, you know, I’ll leave Tim Cook alone for this episode for the most part, but
⏹️ ▶️ Marco you know, I do think we will look back on this time and say Apple was really
⏹️ ▶️ Marco behind on LLMs and you know they they spent their time making a car
⏹️ ▶️ Marco and a vision Pro and while everyone else was doing this and you know they are
⏹️ ▶️ Marco behind here and I really hope they catch up and it’s not a bigger deal but we’ll see how that goes.
⏹️ ▶️ John Well there’s a question of whether them being behind is an advantage or disadvantage though like the reason their stock price is up
⏹️ ▶️ John it’s like this is further evidence that LLM technology that nobody really has a
⏹️ ▶️ John moat that even if you are the best at making these AI things, there’s nothing
⏹️ ▶️ John you’re doing that someone else can’t also do because everything you’re doing is essentially based
⏹️ ▶️ John on technology and techniques that everyone understands. You know what I mean? And the reason people think Apple
⏹️ ▶️ John has a moat is because Apple’s just making like computers that run software. Like there’s no mystery
⏹️ ▶️ John about what they do. It’s computer chips, it’s software, it’s hardware. All right. Apple’s moat is
⏹️ ▶️ John yeah, anyone can do this, but we’re the only ones who know how to do it with
⏹️ ▶️ John taste, with style, with the right feature set, with, you know, like all the Apple sort of more intangible
⏹️ ▶️ John things. There’s nothing technologically speaking, even in the Apple Silicon era, really,
⏹️ ▶️ John that it’s like, well, nobody else could do this except for Apple. They have a secret sauce. Their secret sauce is how they combine
⏹️ ▶️ John the ingredients. So to stretch the cooking analogy here, right? Everyone knows everyone’s got the recipe, right? It’s just
⏹️ ▶️ John a question of, can you put it, can you put it all together, But that’s
⏹️ ▶️ John Apple’s mode. That’s why Apple’s been so successful for so long. It’s not like they have a secret technology
⏹️ ▶️ John that nobody knows about. And it’s the same thing with the AI companies. And OpenAI, I think, kind of felt
⏹️ ▶️ John like it was the Apple of AI. It’s like, well, yeah, everybody knows how to make an LM. And we publish all these papers to say how
⏹️ ▶️ John we’re doing it. And we know how to train them. We do all these different techniques. And we talk about it. But we’re so
⏹️ ▶️ John good at it. We’re the Apple of AI. And we can combine things in a way that no one else can copy. no
⏹️ ▶️ John one else can make a product that is a substitute for ours. And DeepSeek said, well, I think we can
⏹️ ▶️ John do that. And so far, the public is like, maybe it’s just a fad, and people are trying
⏹️ ▶️ John it. And we’ll get to some reasons why you might not want to do that in a little bit. But practically speaking,
⏹️ ▶️ John it’s like, is this a substitute for the other thing? Does it do what the other thing did? And when it comes to personal
⏹️ ▶️ John computers, a lot of people say, well, a Windows PC kind of technically does the same thing as a Mac, but not really. Android
⏹️ ▶️ John phone versus iPhone, there’s still that differentiation. Apple still has a moat and it’s been sustained,
⏹️ ▶️ John a sustainable moat for a long time, basically made of intangibles. And I don’t think
⏹️ ▶️ John OpenAI has that kind of a moat where everyone has the same technology, but they’re just doing it a little
⏹️ ▶️ John bit special. But like down to the, you know, this is the Samsung effect. But if you go to the DeepSeek
⏹️ ▶️ John website, you could be forgiven if you squint your eyes and think, is this the ChatGPT website?
⏹️ ▶️ John It’s a Samsung style ripoff. Like it looks so much like ChatGPT.com, but it’s
⏹️ ▶️ John not, the icons are slightly different. And the Apple is similar, right? So
⏹️ ▶️ John I think in this area where Apple’s kind of like behind, it’s like, look, I feel, I think Apple feels
⏹️ ▶️ John if people are talking about what is the number one app on our store, we’re still winning.
⏹️ ▶️ John Like we don’t need to have the number one app. We just need Chats GPT and DeepSeek and whatever competitor we’ve never heard
⏹️ ▶️ John of to be duking it out on our platform. That shows we’re still in the game here. And
⏹️ ▶️ John we’ll just wait and see which LLM is the most successful. And we’ll partner with them
⏹️ ▶️ John and we’ll leverage their technology and we’ll work on our own. And there’s still out there, which I keep
⏹️ ▶️ John mentioning every time we talk about LLMs and AI, the question of how useful
⏹️ ▶️ John this technology is and in what context we are currently in the phase, the chaos bubble phase,
⏹️ ▶️ John where it’s like, this technology is good for everything and should be used everywhere. That is not going to be true. It’s
⏹️ ▶️ John only going to be useful in some places for some things. But right now everyone is trying everything. And
⏹️ ▶️ John we’ll find out where is it useful and where is it not. With what Apple has done in Apple intelligence,
⏹️ ▶️ John I don’t think they’ve done anything where you look at it and say, I can’t live
⏹️ ▶️ John without this AI feature now. It has made such a big change in my life. I’m glad it’s there. I will be
⏹️ ▶️ John sad if it ever goes away. I don’t, you know, like where is the utility and has Apple harnessed
⏹️ ▶️ John that in its operating system? I don’t think they have yet, right? But they got to keep trying. And same thing with everybody else.
⏹️ ▶️ John Everyone else is trying. So I think Apple’s in a good position right now. And until and unless someone
⏹️ ▶️ John out there sort of tries to usurp Apple’s sort of platform control,
⏹️ ▶️ John I think Apple’s fine content to just keep trying different approaches to mixing AI into its
⏹️ ▶️ John platform and wait to see who’s left standing at the end of all this.
⏹️ ▶️ John Is it gonna be OpenAI, NVIDIA, DeepSeek, Anthropic, some other new company we’ve never heard of? Like
⏹️ ▶️ John who wants to partner with us later? Well, you know, you guys duke it out and we’ll just figure out who has
⏹️ ▶️ John the best, maybe we’ll buy one of the ones that has the best one or whatever, it’s a risky move because I don’t think Google’s thinking
⏹️ ▶️ John that because Google’s thinking these people are a direct threat but Apple’s like, we can wait and see, work on our models
⏹️ ▶️ John and just keep trying to integrate it into our apps and see if anything sticks.
⏹️ ▶️ Marco I don’t know, I think the dynamics are a little worse than that for Apple. I mean, they do have advantages in
⏹️ ▶️ Marco the sense that obviously they have infinite money and so if they do end up needing to buy somebody at the end of the day, they can.
⏹️ ▶️ Marco I don’t think that’s their style in this kind of scale but they could if they really had to. But I think the bigger
⏹️ ▶️ Marco challenge is like, Apple’s whole thing about owning and controlling core technologies for
⏹️ ▶️ Marco their products. There’s obviously a huge role now and in the future
⏹️ ▶️ Marco for LLMs and AI type models being core technologies of their products.
⏹️ ▶️ John Is it a core technology of their product?
⏹️ ▶️ Marco Yes, of course. I mean, look at how many features are going to be based on it. Is
⏹️ ▶️ John it now? You think it’s a core technology of their product now? It’s a core technology of their, a core part of their marketing.
⏹️ ▶️ John I don’t think it’s a core part. You could turn off Apple intelligence on people’s iPhones and see how long it takes them to notice.
⏹️ ▶️ John Like it is not a core technology in the same way as like Apple Silicon or their operating
⏹️ ▶️ John system or their app store. Like, I mean, I’m not saying it’s not going to be, but right now I feel like.
⏹️ ▶️ John Apple intelligence, if you had to say, does this fit the Tim Cook doctrine of we need to own and control, blah, blah, blah. Sure,
⏹️ ▶️ John it does yet. I mean, it sure seems like it’s going to. So it’s a good idea for them to be pursuing that. But right now,
⏹️ ▶️ John I don’t think they’ve proven it.
⏹️ ▶️ Marco I think it’s rapidly becoming an assumed feature on computing
⏹️ ▶️ Marco platforms in various contexts. And it’s only going, I mean, look, LLMs have only really been a thing in the consumer world
⏹️ ▶️ Marco for like two years. They’re still brand new and they’re already, like people are expecting chat
⏹️ ▶️ Marco GPT-like functionality all over the place. Hell, look, it’s Siri. But even just going to things like improving
⏹️ ▶️ Marco the quality of dictation or text-to-speech and speech to text and
⏹️ ▶️ Marco image recognition of things. Is it improving that? Well, it can in other
⏹️ ▶️ Marco, John platforms. It does, I think. Not at Apple’s
⏹️ ▶️ John hands, apparently, but other people you
⏹️ ▶️ John, Marco can use it. Well,
⏹️ ▶️ Marco right. And so the other things that Apple has historically been kind of bad at that
⏹️ ▶️ Marco are kind of big data or big infrastructure problems, things like search indexes.
⏹️ ▶️ Marco Apple is not a great search company in many ways. They kind of have their
⏹️ ▶️ Marco own search stuff going on in a few places. They’re not great at search. They’re not
⏹️ ▶️ Marco great at running consumer-facing web services. That’s just not one of the things
⏹️ ▶️ Marco they are really good at. They’re able to avoid those things being a problem for them in many ways.
⏹️ ▶️ Marco Maybe a bigger example is voice assistants. Apple is
⏹️ ▶️ Marco not good at making voice assistants. But they’ve been able to get by in part
⏹️ ▶️ Marco because of the massive lock-in they have that you can’t make a competing
⏹️ ▶️ Marco voice assistant on iOS. You just can’t. But, and also in part because it turned
⏹️ ▶️ Marco out voice assistants were not that important. They were not that big of a threat. If yours
⏹️ ▶️ Marco sucks as much as Siri has sucked for its entire life, it doesn’t make people not buy your product. Well,
⏹️ ▶️ Marco LLM and LLM-based features are, I think, going to be somewhere on that spectrum.
⏹️ ▶️ Marco We don’t know where yet. It’s possible that it’s gonna be really important that people
⏹️ ▶️ Marco will start assuming these features will be there and will work better than they do on Apple’s platforms.
⏹️ ▶️ Marco And if Apple never takes this more seriously and puts and, you know, develops
⏹️ ▶️ Marco more culture and engineering and infrastructure around this, the way they never
⏹️ ▶️ Marco got into web services and never got into voice assistants very well, if they miss on this,
⏹️ ▶️ Marco it might be more important to their customers. We don’t know. They will still have the lockout
⏹️ ▶️ Marco problem with, you know, locking out any competitors, which in, I think in their case, this will actually
⏹️ ▶️ Marco hurt them a little bit here because that will just make the iPhone
⏹️ ▶️ Marco work worse for iPhone customers in these ways. I don’t know. I
⏹️ ▶️ Marco think this kind of shakeup in technology, we have seen this dramatically
⏹️ ▶️ Marco disrupt really established competitors. We’ve seen like, look, when the iPhone first came
⏹️ ▶️ Marco out, it kind of sucked at a lot of things, but we loved it and we use it anyway. And
⏹️ ▶️ Marco eventually it stopped sucking at those things. we all love them and during that time
⏹️ ▶️ Marco between sucking and not sucking on a lot of things a lot of people talked
⏹️ ▶️ Marco a lot about iPhones on PCs and Windows PCs even
⏹️ ▶️ Marco many people talked about their iPhone using their Windows PC and many other people
⏹️ ▶️ Marco talked about on their Windows PC we’re fine we have 90% of the market what’s the problem here
⏹️ ▶️ Marco and then phones massively disrupted the entire computer industry and Microsoft was screwed
⏹️ ▶️ Marco because they weren’t taking mobile seriously enough. That can happen to Apple. They seem
⏹️ ▶️ Marco impossibly big and established at this moment, but there’s this huge area
⏹️ ▶️ Marco of technology that’s disrupting a lot of things and that has pretty big promise for the future that Apple
⏹️ ▶️ Marco has shown no core competency in and not much competitiveness,
⏹️ ▶️ Marco not really taking it seriously. Clearly, they were caught off guard. Clearly, they started their AI efforts
⏹️ ▶️ Marco way later than everybody else. Clearly, they are way behind and they don’t seem to have that
⏹️ ▶️ Marco kind of talent in the company at anywhere near the levels that their competitors do. So I
⏹️ ▶️ Marco think Apple is extremely vulnerable to disruption from AI and I don’t
⏹️ ▶️ Marco think they’re taking it seriously enough. I don’t think they are prepared. I don’t think they started early enough and
⏹️ ▶️ Marco we’ll see if they can recover but so far with with what we’ve seen so far from them
⏹️ ▶️ Marco I don’t see any reason to be optimistic on this. Maybe we’ll see it in the future. Maybe they’ll pull out of
⏹️ ▶️ Marco this nosedive that they seem to be in with AI and actually, you know, finally
⏹️ ▶️ Marco get their footing and kind of take off. I hope so. I know I’m mixing a lot of metaphors there. I hope they take
⏹️ ▶️ Marco off and I hope they can actually take this way more seriously than they
⏹️ ▶️ Marco appear to be taking it so far because they could be in a great space. Like hardware
⏹️ ▶️ Marco wise, they’re in a great place to run inference on their devices because their devices have all this memory the GPU
⏹️ ▶️ Marco can use. So they’re in a great place there. And of course they have the neural engines and everything like
⏹️ ▶️ Marco they make all their custom silicon like Apple’s in a great place hardware wise for the devices
⏹️ ▶️ Marco to run inference where they seem to be way behind
⏹️ ▶️ Marco is just in the other areas the software the models and that’s that’s what I have concerns about that it seems
⏹️ ▶️ Marco like this huge opportunity for disruption is aimed right
⏹️ ▶️ Marco at them and they don’t seem prepared or necessarily taking it seriously
⏹️ ▶️ Marco you know Apple among among their problems, you know, again, we like Apple a lot, they make a lot of great stuff,
⏹️ ▶️ Marco but they also have debilitating hubris. And I don’t
⏹️ ▶️ Marco know if they know how much they are under threat by this in the future, potentially.
⏹️ ▶️ Marco I think they think they’re in a really good spot. I think they think Apple Intelligence is great because
⏹️ ▶️ Marco why else would they have called it Apple Intelligence and taken the huge risk of putting their brand name on
⏹️ ▶️ Marco it like that? I think they are super confident that they’ll be fine. And I
⏹️ ▶️ Marco don’t know that gonna be fine like in the short term sure in the long term this
⏹️ ▶️ Marco could be like bomber missing mobile we don’t know yet but that’s the potential that is
⏹️ ▶️ Marco that is what this has the potential to do and so I I hope they are taking it that seriously
⏹️ ▶️ Marco we have not yet seen signs that they are
⏹️ ▶️ John I think they’re taking it very seriously like they’re making a big push for it but like the the history you mentioned like well first of all
⏹️ ▶️ John voice assistants that is the obviously the area where the farthest behind they were far behind that before LMS like we all know that,
⏹️ ▶️ John right? There was, they were just not doing well with that. And somehow able to limp along with Siri being terrible
⏹️ ▶️ John for all those years. Um, they also don’t make a search engine. They’ve just been, they essentially leaning on Google and other
⏹️ ▶️ John companies cause that is apparently not core enough part of their operating system with the AI stuff. Apple feels like
⏹️ ▶️ John they need to incorporate it because it is a potential disrupting threat. Absolutely. But potential keyword being potential,
⏹️ ▶️ John but like, you know, you got to cover your bets. So they, however many years ago they decided we’re just going to go on on this Apple intelligence
⏹️ ▶️ John thing. They probably waited too long. They haven’t, it’s been taking them forever to do what they
⏹️ ▶️ John said they were going to do, right? So it’s, it’s going slowly and the things they’re rolling out are not that impressive,
⏹️ ▶️ John but I feel like they are really bought into it. Like every aspect of the company is focused on this. They just
⏹️ ▶️ John haven’t figured out how to do anything that’s particularly compelling with it.
⏹️ ▶️ John And meanwhile, their voice assistant is sitting there still sucking, right? And the competition, the
⏹️ ▶️ John competing voice assistants are getting better and better because LLM is helping them and LLMs are not really helping
⏹️ ▶️ John Apple and its traditional weaknesses. But, you know, they last, they existed a long time with a crappy
⏹️ ▶️ John voice assistant. But they can exist with a crappy one for a while longer, as long
⏹️ ▶️ John as those voice assistants that everyone else is doing don’t get much, much
⏹️ ▶️ John better. And obviously, they’re already better than Siri. But like there’s this breaking point, like there’s, you know, how much
⏹️ ▶️ John better is one cell phone than another? The iPhone was across that breaking point of like, this isn’t just
⏹️ ▶️ John a little bit better than your Nokia candy bar phone. This is a whole different thing. Arguably, we’re kind of
⏹️ ▶️ John already there with voice assistants, but maybe not because the whole LLM’s
⏹️ ▶️ John reliability problem. But anyway, if someone figures this out, yeah, it’s good for Apple to to
⏹️ ▶️ John be in the game. They’re trying to get in the game. I think they are serious about it. I think they’re spending a lot of time and money
⏹️ ▶️ John to the detriment of all the other things they could be doing to try to put Apple intelligence everywhere to try to get
⏹️ ▶️ John better at it. But I’m not I’m not optimistic because I, you know, like
⏹️ ▶️ John I’m pessimistic, not because I think they’re not putting in the effort. I’m pessimistic because it doesn’t seem like a thing, like you said, that they’ve
⏹️ ▶️ John historically been good at. And so no matter how much effort they put in, it’s like, well, you can be really
⏹️ ▶️ John serious about this and put a lot of effort into it, but if you are not able to acquire those skills, will
⏹️ ▶️ John you ever be fruitful? But yeah, we’ll see how this goes.
⏹️ ▶️ John I totally think they’re doing the right thing. It is a potential threat, and you shouldn’t wait for it to be
⏹️ ▶️ John a life-threatening thing before you get serious about it. that you can’t afford to wait, which is why everybody is scrambling to
⏹️ ▶️ John do AI everything, because they’re like, well, it might be huge. Look how much it’s
⏹️ ▶️ John advanced in the past two years. So we better got it on this ASAP. But right now it
⏹️ ▶️ John isn’t, but it might be. So we better do everything that we can. And it’s kind of
⏹️ ▶️ John sad seeing Apple flail with Apple intelligence, because it’s like they’re trying to do stuff, but
⏹️ ▶️ John what they’re producing is like, I don’t know, it’s not compelling. Like they did the partnership with chat GBT to say like,
⏹️ ▶️ John We don’t have that. We’re not going to have that. We’ll partner with them. And somehow they found a way to screw that up, like with Gruber’s
⏹️ ▶️ John story where they’re trying to ask when the Super Bowls are. And when Apple asks ChatGPT, it still
⏹️ ▶️ John gets the wrong answer. But if you ask ChatGPT directly, it gets the right answer. So they’ve somehow partnered with ChatGPT
⏹️ ▶️ John and made it less functional. Is that the Apple touch? I don’t know.
⏹️ ▶️ Marco Well, I mean, and it could end up, look, it could end up being something like Siri, where Apple
⏹️ ▶️ Marco is just limping along in mediocrity forever, buoyed by their own
⏹️ ▶️ Marco lock-in that they have on their platforms. Or…
⏹️ ▶️ Marco have enough other advantages that they can make up for it. You know, they can have something like this chat GPT thing where they’re integrating
⏹️ ▶️ Marco somebody else’s, but I see that going the direction of Maps. The iPhone
⏹️ ▶️ Marco used to have Google Maps built into its Maps app. There was, it wasn’t a separate Google Maps app, it was just
⏹️ ▶️ Marco the Maps app on the phone, and it was Google Maps behind it. And then we had the Apple Maps fiasco
⏹️ ▶️ Marco back when those companies split that relationship up for lots of pretty good reasons.
⏹️ ▶️ Marco And Apple, they needed maps as a core feature of the phone, but
⏹️ ▶️ Marco it took them, what, a decade before their version of maps was actually
⏹️ ▶️ Marco decent? And for most of that time, they suffered because their
⏹️ ▶️ Marco maps kind of sucked. At least you could run Google Maps on the phone, and that helped in many, many ways.
⏹️ ▶️ Marco But if that’s what their LLM efforts end up looking like, where they’re okay
⏹️ ▶️ Marco now kind of backfilling their capabilities with chat GPT, but they’re
⏹️ ▶️ Marco going to have to use their own model. That’s not going to be a long-term solution.
⏹️ ▶️ John What if it’s like search though? What if OpenAI pays them $20 billion a
⏹️ ▶️ John, Marco year to make
⏹️ ▶️ John OpenAI the default voice assistant on Apple? Apple’s forced to open the voice
⏹️ ▶️ John assistant thing to third parties because of the EU. And then OpenAI doesn’t have its own platform. again, if
⏹️ ▶️ John I think Apple just loves the fact that people care what’s number one in the app store still like this, the magic of the
⏹️ ▶️ John app store, like all these all these apps want to be on the iPhone, right? So it might be like maps.
⏹️ ▶️ John And part of the thing that made maps come to a head was that Google demanded access to customer data that Apple
⏹️ ▶️ John wasn’t willing to give, right? In exchange for continuing the deal. So they went off and did their own thing, and it was painful
⏹️ ▶️ John and long. Or it could be like search where Apple is never going to be good at it. And they say, you know what?
⏹️ ▶️ John You should pay us for you to be the default voice assistant on on iOS and
⏹️ ▶️ John they’re getting suddenly, you know, $20 billion a year from open AI or from Deep Seeker. Who knows? But anyway,
⏹️ ▶️ John yeah, it’s still that’s the thing about the current situation. We don’t know which direction this is going to go in. Those are two possible
⏹️ ▶️ John directions. And there’s third and fourth directions we’re not even thinking about. But everyone is scrambling to try
⏹️ ▶️ John to do everything they can to figure out wherever this goes. We we got to be ready. And I think Apple
⏹️ ▶️ John is there. They’re showing that they have been able to kind of. Rally
⏹️ ▶️ John the troops to do Apple intelligence everywhere, but they’re also showing that their actual execution of that has been
⏹️ ▶️ John not impressive and way slower than I think we all thought it was going to be.
⏹️ ▶️ Marco Yeah. Bringing it back to deep seek though. This is a song about deep seek. Um,
⏹️ ▶️ Marco what this disruption was with deep sea coming, you know, coming on the scene and showing this huge reduction in cost.
⏹️ ▶️ Marco This is just what computer development looks like. Like, you know, we
⏹️ ▶️ Marco find when we have like, you know, new software areas, We
⏹️ ▶️ Marco do things a certain way and then people find optimizations. And one of the most
⏹️ ▶️ Marco delightful things about software development is that when you find an optimization, oftentimes it’s like, oh, this is
⏹️ ▶️ Marco now a hundred times faster or more. Like, it can be dramatically
⏹️ ▶️ Marco better or dramatically faster or dramatically smaller. You know, finding new types of compression
⏹️ ▶️ Marco or faster algorithms that can reduce the order of magnitude of a
⏹️ ▶️ Marco function, like stuff like that. That’s just how computers go. So what’s interesting about this deep-seek thing is that,
⏹️ ▶️ Marco you know, this is an area where, you know, AI model training and model inference
⏹️ ▶️ Marco are just so unbelievably inefficient in terms of resources
⏹️ ▶️ Marco used. Like, the amount of computing power and, you know, just hardware and electrical
⏹️ ▶️ Marco power and everything, the amount of grunt of resource usage
⏹️ ▶️ Marco needed to make an LLM do anything or to train an LLM in the first place is
⏹️ ▶️ Marco so unbelievably massive that when we find optimizations like
⏹️ ▶️ Marco this, it shakes the entire market. I don’t
⏹️ ▶️ Marco think we’ve had anything like that in computing for a very long time, where just
⏹️ ▶️ Marco the normal process of software maturation and software advancement
⏹️ ▶️ Marco of occasionally finding giant optimizations like this. We haven’t seen
⏹️ ▶️ Marco that on a scale where it’s like, oh, this now affects billions of dollars
⏹️ ▶️ Marco of hardware that’s
⏹️ ▶️ Marco, John been bought. One example
⏹️ ▶️ John that I think, it might’ve been Ben Thompson that gave us example, and we’re gonna get to him in a second because he’s the next item up in here. But
⏹️ ▶️ John I think this example is from him, and I think it’s a good one. And if not, whoever came up with it,
⏹️ ▶️ John sorry, I can’t credit you. The disruption in data centers,
⏹️ ▶️ John when Google said, instead of buying like, you know, servers from sun or whatever,
⏹️ ▶️ John these big expensive Unix workstations, we’re going to deploy commodity sort of PC
⏹️ ▶️ John style server hardware and manage that crappy commodity
⏹️ ▶️ John hardware with software. And that destroyed the entire industry of
⏹️ ▶️ John really expensive proprietary Unix things for data centers that the entire internet was built on up to that
⏹️ ▶️ John point, because Google said, Yeah, we found a better, cheaper way to do
⏹️ ▶️ John data centers, data centers are important. People if you wanted to build the data center at the scale that Google
⏹️ ▶️ John needs, and you wanted to, you know, buy hardware from sun, or HP or whatever to
⏹️ ▶️ John put in there with these really expensive, you know, workstation class server class things or
⏹️ ▶️ John whatever, that would cost way too much. So how about we just take crappy hardware
⏹️ ▶️ John and a huge amount of it and have some really cool software layer on top that manages the fact that all
⏹️ ▶️ John this stuff is crappy and cheap and underpowered and it’s going to break. Uh, and that destroyed the whole industry.
⏹️ ▶️ John All those companies are like half of the companies are don’t even exist anymore because what Google did
⏹️ ▶️ John showed that you could do the same thing that everybody needs to do that used to cost huge amounts of money and power and
⏹️ ▶️ John you could do it cheaper, uh, and better, uh, with a slightly different approach.
⏹️ ▶️ John And that was an optimization they made. So yeah, I mean, this is not as
⏹️ ▶️ John severe as that, because what they’ve done is basically just a really good job of programming the
⏹️ ▶️ John hardware they had. Anyway, we should go to this next item, because it goes into more detail about the particular innovations they made.
⏹️ ▶️ Casey, John back. All right, so,
⏹️ ▶️ Casey Ben Thompson, front of the show, did, I believe this is a non-paywalled post, which he
⏹️ ▶️ Casey called DeepSeq FAQ. And honestly, it’s worth reading.
⏹️ ▶️ Casey I could sit here and read the whole damn thing, but it would take a while. So I’m just going to read the snippets that John has
⏹️ ▶️ Casey thankfully curated for us. There is a lot here, so gentlemen, please interrupt when you’re ready. The
⏹️ ▶️ Casey DeepSeq V2 model introduced two important breakthroughs, DeepSeq MOE and
⏹️ ▶️ Casey DeepSeq MLA. The MOE in DeepSeq MOE refers to mixture of experts.
⏹️ ▶️ Casey Some models, like GPT 3.5, activate the entire model during both training and inference.
⏹️ ▶️ Casey It turns out, however, that not every part of the model is necessary for the topic at hand. MOE splits the model into
⏹️ ▶️ Casey multiple quote-unquote experts and only activates the ones that are necessary. GPT-4 was an MOE
⏹️ ▶️ Casey model that was believed to have 16 experts with approximately 110 billion parameters
⏹️ ▶️ Casey each. DeepSeq MLA, multi-head latent attention is the MLA
⏹️ ▶️ Casey there, was an even bigger breakthrough. One of the biggest limitations on inference is the sheer amount of memory required.
⏹️ ▶️ Casey you both need to load the model into memory and also load the entire context window. Context windows are particularly
⏹️ ▶️ Casey expensive in terms of memory as every token requires both a key and a corresponding value. DeepSeq
⏹️ ▶️ Casey MLA makes it possible to compress the key value store dramatically decreasing the memory usage during inference.
⏹️ ▶️ Casey So with regard to costs from the DeepSeq v3 paper, which we will link,
⏹️ ▶️ Casey note that the aforementioned costs include only the official training of DeepSeq v3, excluding
⏹️ ▶️ Casey the costs associated with prior research and ablation, is that right? Experiments on
⏹️ ▶️ Casey architectures, algorithms, or data.
⏹️ ▶️ John Yeah. So these innovations they had, like again, some of the innovations are things that OpenAI was already doing with GPT
⏹️ ▶️ John and they’re doing as well. And then the other thing is, you know, if you look at the paper, it’s like, oh, well, you have
⏹️ ▶️ John a bunch of data and it’s very expensive. What if you compressed it? I mean, it’s not rocket science, but that’s how
⏹️ ▶️ John innovation goes here. Like we, you know, let’s, let’s take the approach that they did with GPT-4 and
⏹️ ▶️ John do that same thing to reduce our footprint. And let’s reduce it further by compressing this thing that used to take
⏹️ ▶️ John up a lot of memory. And on the cost front, these are, by the way, these are papers are not, they’re not about the
⏹️ ▶️ John R1 model that we’re talking about. These are precursors to that. So if you’d been paying attention to this stuff a month or two
⏹️ ▶️ John ago, when they put this stuff out, you could have seen that like what was coming. But on the cost thing, they said, I would only cost a $6 million.
⏹️ ▶️ John When they originally said that a lot of people just didn’t believe them, because they said, well, they’re They’re lying.
⏹️ ▶️ John So Chinese companies, they’re lying about how much money it costs. There’s no way they could have spent $6 million to do
⏹️ ▶️ John something that costs hundreds of millions of dollars when OpenAI did it and have equal performance. It’s obviously not true.
⏹️ ▶️ Marco It’s like when the BlackBerry CEO thought the iPhone demos were faked.
⏹️ ▶️ John, Marco Yeah, exactly. But
⏹️ ▶️ John in the paper, they do say that those costs are not, it’s not the all-in costs. That’s just the cost of their
⏹️ ▶️ John final run. And maybe the OpenAI number is like all the research needed to get
⏹️ ▶️ John to that point. So maybe the number is not as low as they say it is, but you can see in Ben’s
⏹️ ▶️ John paper, he does some back of the envelope math to say, given the technology that they’ve described in their own public
⏹️ ▶️ John research papers, you can do the math and say, yeah, if the number is not exact,
⏹️ ▶️ John it’s in the ballpark. You can see how they would arrive at that because they’re doing less stuff. Like that’s
⏹️ ▶️ John the innovation. Do less math in computers. Like use less memory, do less
⏹️ ▶️ John computations. And the magic is that when you do less work and spend less money,
⏹️ ▶️ John you can somehow get a result that is comparable to open AI, right? It’s,
⏹️ ▶️ John you know, so I think that mostly holds out. So one of the other theories, speaking of the people who
⏹️ ▶️ John thought the iPhone was faked or whatever, the theory was like, they must have gotten the good GPUs. They couldn’t
⏹️ ▶️ John be using Nvidia H800 or whatever things. They must’ve gotten the good ones secretly. No, they just
⏹️ ▶️ John figured out, they did the equivalent of like writing an assembly code, like the low level version of like
⏹️ ▶️ John extracting every ounce of juice from the crappy GPUs that they do have. I mean, just straight
⏹️ ▶️ John up, like just brute force. Like here is, it’s like when you make a console game, like in the end of a console generation
⏹️ ▶️ John versus the beginning, by the end, they figured out every little trick of that console to get the most performance out of it and they never could have
⏹️ ▶️ John made that thing at the beginning. Anyway, yeah, kudos to them for doing it. And these papers
⏹️ ▶️ John that were about these innovations, anyone can see these papers. OpenAI can get these papers, can read them. People
⏹️ ▶️ John can see what they did. People can do the same thing. It’s all out in the open. There’s no secrets here. As
⏹️ ▶️ John Casey noted before, the R1 model that we have still haven’t even talked about yet, that’s MIT
⏹️ ▶️ John licensed. The weights are open source. So you can just grab these and pull them. And you don’t have
⏹️ ▶️ John to license them. Like, it’s MIT licensed. You can do anything you want with it. You can integrate it into your product or whatever. Very, very open.
⏹️ ▶️ Casey Indeed. So then there’s distillation. This is models training models, again, reading
⏹️ ▶️ Casey from Ben Thompson. Distillation is a means of extracting understanding from another model.
⏹️ ▶️ Casey you can send inputs to the teacher model and record the outputs and use that to train the student
⏹️ ▶️ Casey model. This is how you get models like GPT-4 Turbo from GPT-4. Distillation
⏹️ ▶️ Casey is easier for a company to do on its own models because they have full access, but you can still do distillation in
⏹️ ▶️ Casey a somewhat more unwieldy way via API or even if you get creative via chat clients.
⏹️ ▶️ Casey Distillation obviously violates the terms of service of various models, but the only way to stop it is to actually cut off access
⏹️ ▶️ Casey via IP banning, rate limiting, et cetera. It’s assumed to be widespread in terms of model training, and
⏹️ ▶️ Casey it’s why there’s an ever increasing number of models converging on GPT 4.0 quality. This doesn’t mean
⏹️ ▶️ Casey that we know for a fact that DeepSeek distilled 4.0 or Claude, but frankly it
⏹️ ▶️ Casey would be odd if they didn’t.
⏹️ ▶️ John Yeah. Now on that front, there’s been some stories of people saying, Hey, I was using DeepSeek and I was
⏹️ ▶️ John trying various things to type into the like different prompts in the chat thing. And one of the responses I got was
⏹️ ▶️ John like, I’m sorry I can’t do that because open AI something or like it referred to itself as open AI, like
⏹️ ▶️ John the deep seek model did. It’s kind of like when like the open AI model starts spitting back like direct
⏹️ ▶️ John quotes from New York Times and stuff. When deep seek start saying as an open AI model, I can’t x, y
⏹️ ▶️ John and z. It makes you think that perhaps deep seek was trained using
⏹️ ▶️ John open AI models, right? And that’s as Ben says here, it’s just assume that everybody is doing this.
⏹️ ▶️ John Because, you know, doing this having models train other models has been a practice for a while now
⏹️ ▶️ John and why would deep seek not do it? Uh, but how does open AI feel about that?
⏹️ ▶️ Casey Yeah. So it turns out open AI, who by most measures stole the entirety of world’s
⏹️ ▶️ Casey knowledge in order to train their model seems to be a little grumpy that somebody stealing their knowledge to
⏹️ ▶️ Casey train their model. Uh, and I don’t really have a lot of sympathy for him on this
⏹️ ▶️ Casey one, to be honest with you, like, sorry, them’s the breaks. If you’re going to be a turd.
⏹️ ▶️ John Well, I mean, so here’s the thing, like, so this is, we’ll put a link in the show notes to this 404 Media story that had a
⏹️ ▶️ John good headline, which is OpenAI furious, DeepSeek might have stolen all the data OpenAI stole from us.
⏹️ ▶️ John So it’s like, OpenAI’s argument is like, well, we’ve talked about this many times in past episodes. Well, we’re
⏹️ ▶️ John not really stealing the data we’re using to train our models and it’s a different thing and it’s transformative
⏹️ ▶️ John and blah, blah, blah. And I feel like if OpenAI really believes that and it’s not just a bunch of BS,
⏹️ ▶️ John when another model uses your model to train their model, they can say, well, we’re not stealing your data, we’re just using
⏹️ ▶️ John it to train a model and blah. It’s like exactly the same argument, right? And I, you know, as we’ve discussed,
⏹️ ▶️ John who knows how solid that argument is and how it will turn out, but it really is very directly,
⏹️ ▶️ John like they’re just using it to train a model. They’re not stealing your data. When they train a model
⏹️ ▶️ John with your data, it’s transformative. They don’t need your permission to get it, but open the eyes like, no, totally. It’s in our terms of service. You have
⏹️ ▶️ John to, so they don’t really have a leg to stand on here. It’s like, look, it’s either, it’s either it’s not
⏹️ ▶️ John okay for both of you to do it, or it’s okay for both of you to do it. And I’m sure the lawyers say, well, our terms of service say otherwise,
⏹️ ▶️ John but just like setting aside the law in terms of service and in crossing international boundaries with the
⏹️ ▶️ John U S company versus a Chinese company, it just seems like they’re mad because somebody else is doing the same thing to them
⏹️ ▶️ John that they did to everybody else.
⏹️ ▶️ Casey Yep. Pretty much. All right. So R1, R1, zero and
⏹️ ▶️ Casey reinforcement learning. R1 is a reasoning model like open AI’s O1.
⏹️ ▶️ Casey It has the ability to think through a problem producing much higher quality results, particularly in areas like coding, math, and logic.
⏹️ ▶️ Casey Reinforcement learning is a technique where a machine learning model is given a bunch of data and a reward function. The classic
⏹️ ▶️ Casey example is AlphaGo, where DeepMind gave the model of the rules of Go
⏹️ ▶️ Casey with the reward function of winning the game, and then let the model figure everything else out on its own.
⏹️ ▶️ Casey This famously ended up working better than the other more human-guided techniques. LLMs to date,
⏹️ ▶️ Casey however, have relied on reinforcement learning with human feedback. Humans are in the loop to help guide the model, navigate difficult
⏹️ ▶️ Casey choices where rewards weren’t obvious, etc. RLHF, or reinforcement learning
⏹️ ▶️ Casey from human feedback, was the key innovation in transforming GPT-3 into chat GPT
⏹️ ▶️ Casey with well-formed paragraphs, answers that were concise and didn’t trail off into gibberish, etc.
⏹️ ▶️ Casey R10, however, drops the HF, the human feedback part. It’s just reinforcement learning. DeepSeq
⏹️ ▶️ Casey gave the model a set of math, code, and logic questions and set two reward functions, one for the right answer and one for
⏹️ ▶️ Casey the right format that utilized a thinking process.
⏹️ ▶️ John Yeah, this is what we talked about when we were first discussing chat GPT and the fact that they had like, you know,
⏹️ ▶️ John hundreds of thousands of human generated question and answer pairs to help train it. Yes, they trained on all the knowledge
⏹️ ▶️ John in the internet, but also there was a huge human powered effort of like, let’s
⏹️ ▶️ John tailor make a bunch of what we think are correct or good question and answer pairs and feed them.
⏹️ ▶️ John And they had to pay human beings to make those that they could use to train their model. That obviously costs a lot of money,
⏹️ ▶️ John takes a lot of time. And Ben gives the AlphaGo example of like,
⏹️ ▶️ John if we try to make a computer program play a game really well, should we have like experts that go like teach
⏹️ ▶️ John the AI thing? What’s the best move here or there? Or should we just say, no humans are involved.
⏹️ ▶️ John Here’s the game, here’s the rules. Just run with a huge amount of time with the
⏹️ ▶️ John reward function of winning the game. And eventually the model will figure out how to be the best Go player in the world.
⏹️ ▶️ John Rather than us carefully saying, well, you gotta know this strategy, you gotta know that or whatever. Obviously, getting the humans out of the loop
⏹️ ▶️ John saves money, saves time, and it removes some of the
⏹️ ▶️ John blind alleys you might go down because humans are gonna do a particular thing that works a particular way, and we don’t know that that’s
⏹️ ▶️ John the correct solution there. So I’m assuming the R in both R1 and R1-0 both
⏹️ ▶️ John stand for reinforcement learning, and maybe the zero stands for, I’m trying to parse their names, who knows, the fact
⏹️ ▶️ John that we took out the human factor entirely, and we’ll just train this thing
⏹️ ▶️ John Entirely with reinforcement learning on its own, we don’t have to guide it in any way. That seems
⏹️ ▶️ John like it’s probably a better approach because obviously the human feedback approach is not really scalable
⏹️ ▶️ John beyond a certain point, right? Like you can keep scaling up the computing part as computers get faster and better
⏹️ ▶️ John and you give more power and money and blah, blah, blah. But you can’t employ every human on the planet to be making human
⏹️ ▶️ John question and answer pairs right away if you get to that scaling point. So this seems like a fruitful approach. And
⏹️ ▶️ John again, practically speaking, if you want to do it in less money and less time, you can’t hire 100,000 human beings to make
⏹️ ▶️ John questions and answers for your thing. So they didn’t. And it turns out they could make something that worked pretty well
⏹️ ▶️ John even without doing that.
⏹️ ▶️ Casey So R1 is more open than OpenAI. Unlike OpenAI’s O1
⏹️ ▶️ Casey model, R1 exposes its chain of thought and OpenAI published something about why they hide
⏹️ ▶️ Casey O1’s chain of thought, which we’ll link to in the show notes.
⏹️ ▶️ John We talked about that on a past ATP episode about how mad they were, that people were trying to like figure out, like as the people
⏹️ ▶️ John were like prompt engineering and saying, I know you’re hiding the chain of thought. The chain of thought is like, how is it thinking through the problem
⏹️ ▶️ John or whatever? Like they show you a summary of it, but they don’t show you the real one, right? And you can read the blog posts. This is
⏹️ ▶️ John from a while ago about why OpenAI did that. But then people were like, but I figured out if you prompt the
⏹️ ▶️ John O1 model in this way, it will tell you about its chain of thought. And OpenAI was like, that’s against our terms of service. You can’t
⏹️ ▶️ John look under the covers of how our thing works. You’re not allowed to do that. And was banning accounts and stuff. I think that was
⏹️ ▶️ John several months ago. But anyway, you know, that’s, it’s kind of ironic that Open AI, Open isn’t
⏹️ ▶️ John the Open AI name. Like really, they were going to be this magnanimous, you know, public benefit, whatever, blah, blah, and now they’re
⏹️ ▶️ John very quickly changing into a private company, entirely controlled and focused on making money and so on and so
⏹️ ▶️ John forth. And they don’t want you to know how their reasoning model works.
⏹️ ▶️ Casey Meanwhile, the DeepSea CEO, Liang Wen-Feng
⏹️ ▶️ Casey said in an interview that open source is key to attracting talent. They said
⏹️ ▶️ Casey in the face of disruptive technologies, moats created by closed source are temporary. Even open AI’s closed source approach can’t
⏹️ ▶️ Casey prevent others from catching up. So we anchor our value in our team. Our colleagues go grow through
⏹️ ▶️ Casey this process, accumulate know-how and form an organization and culture capable of innovation. That’s
⏹️ ▶️ Casey our moat. Open source publishing papers, in fact, do not cost us anything. For technical
⏹️ ▶️ Casey talent, having others follow your innovation gives a great sense of accomplishment. In fact, open source is more
⏹️ ▶️ Casey of a cultural behavior than a commercial one and contributing it to it earns us respect. There’s also a
⏹️ ▶️ Casey cultural attraction for a company to do this. They were asked, will you change to closed
⏹️ ▶️ Casey source later on? Both OpenAI and Mistral moved from open source to closed source. The answer being,
⏹️ ▶️ Casey we will not change to closed source. We believe having a strong technical ecosystem first is more important.
⏹️ ▶️ John So this is, I mean, perhaps uncharacteristic for China and the Chinese government
⏹️ ▶️ John of not having secrets. This company is saying, we found a better way to do what you were doing
⏹️ ▶️ John and we’re going to tell you how we did it. we tell you everything about it, everything, the stuff is open source, you can get the weights from the model
⏹️ ▶️ John under an MIT license, we’ll publish all the scientific papers about how we did it, no secrets, here it is.
⏹️ ▶️ John And are we going to go closed source like open AI? Are we going to, you know, hide our chain of reasoning? No,
⏹️ ▶️ John are you can see it, we’re not we’re not trying to hide it. There’s no terms of service saying you can’t get at it, we don’t try to summarize it or hide it
⏹️ ▶️ John from you. That is potentially uncharacteristic. One thing that is characteristic and will lead us into the next topic is, yeah,
⏹️ ▶️ John they’re probably not too worried about their employees and giving them this know how or whatever,
⏹️ ▶️ John it’s not like they can just leave and do whatever they want. And the Chinese government has much, much,
⏹️ ▶️ John much more say in what Chinese citizen and Chinese companies do. And so it
⏹️ ▶️ John is kind of like, they don’t have to worry so much about, uh,
⏹️ ▶️ John every employee of deep sea leaving to go become employees of open AI, because that is
⏹️ ▶️ John not something that the Chinese government has ways to prevent that from happening. Let’s say. Um,
⏹️ ▶️ John but still, you know, if you, if you think of like a, uh,
⏹️ ▶️ John a competitor to the U S using the typical, uh, you know, demonized us things of like access of evil, like they’re going to do
⏹️ ▶️ John everything secret in their secret volcano lair. And it’s like, Nope, here’s everything we’re doing. Here’s all the papers.
⏹️ ▶️ John Here’s all the weights in the models, like totally out in the open, which I think is just
⏹️ ▶️ John a finger in the eye of open AI. The fact that they have open in the aim, even more so it’s like we
⏹️ ▶️ John are doing better and we’re not afraid to tell you how we did it because that kind of like what they’re trying to say is kind of like an
⏹️ ▶️ John apple approach it’s like we can tell you how we did it it’s just computers right we’re the where our
⏹️ ▶️ John advantage is not that secret our advantage is whatever intangibles they think they have now i’m not entirely sure they
⏹️ ▶️ John do have any intangibles because again if you look at their app on their website it looks just like chat gpt
⏹️ ▶️ John uh i don’t see any particular differentiation there so we’ll see how this shakes out but right now
⏹️ ▶️ John it’s still looking much more like anybody can make one of these kind of like in the PC
⏹️ ▶️ John industry. Anybody could make a PC. There were winners and losers in the PC industry. Companies would come and go, Compaq,
⏹️ ▶️ John HP, Microsoft eventually started making them, you know, all the niche manufacturers. So many
⏹️ ▶️ John different people made personal computers. The stuff that went into them, everybody knew. There was no secrets,
⏹️ ▶️ John right? There was no secret sauce. It was just like, who’s good at making a personal computer? But it turned out the people who
⏹️ ▶️ John had a moat were the the people on the platform, Windows was the mode because they controlled the platform, they
⏹️ ▶️ John controlled the operating system. And for a long time, Intel had a mode of the best process technology. And we know how that turned out eventually.
⏹️ ▶️ John Not great, but still for a long time, they were tops. Even when they were challenged by AMD,
⏹️ ▶️ John who got in through the side door with an X86 thing, even when they weren’t able to make a 64 bit thing, like
⏹️ ▶️ John there are sustainable ways to be the key important player in the industry,
⏹️ ▶️ John but the key important player was not Compact. It was not HP. It was not any of
⏹️ ▶️ John the PC manufacturers. It was Microsoft with Windows and Intel with the CPUs
⏹️ ▶️ John for a long time. Right now we don’t know who is the equivalent of Microsoft
⏹️ ▶️ John and Intel in the AI age. OpenAI thought it was them. DeepSeek is saying, doesn’t seem like
⏹️ ▶️ John it’s you, but is it DeepSeek either? Are both OpenAI and DeepSeek, are they like,
⏹️ ▶️ John I can’t think of enough PC manufacturer names. Are they IBM
⏹️ ▶️ John and Hewlett Packard? Or is one of them Microsoft or Apple? We’ll
⏹️ ▶️ Casey So R1 is being censored apparently by the Chinese government or at least that’s what it seems. Jason Carty
⏹️ ▶️ Casey posted a screenshot where they asked, what happened on June 4th, 1989? To which the response was, I’m sorry, I cannot
⏹️ ▶️ Casey answer that question. I am an AI assistant designed to provide helpful harmless responses. Mm-hmm. Uh-huh.
⏹️ ▶️ Casey A friend of the show, Greg Pierce, writes, when was Taiwan formed? To which the response is
⏹️ ▶️ Casey great. Taiwan has been an integral part of China since ancient times. Oh?
⏹️ ▶️ Casey Historically, Taiwan has undergone several name changes and administrative adjustments, but it has always been inseparable
⏹️ ▶️ Casey from the mainland of China.
⏹️ ▶️ Marco Administrative adjustments. That’s
⏹️ ▶️ John, Marco an interesting. It’s
⏹️ ▶️ John always been inseparable.
⏹️ ▶️ Casey The formation of Taiwan is a product of China’s long history in cultural development. the Chinese government has always adhered
⏹️ ▶️ Casey to the one China principle, resolutely safeguarding national sovereignty and territorial integrity.
⏹️ ▶️ Casey Any discourse on the history of Taiwan should be based on this indisputable fact.” Like that is just so transparently propaganda.
⏹️ ▶️ Marco Yeah. But, but look, I mean, can, are you, do you think the U S is that far from that? Like,
⏹️ ▶️ Marco do you, do you think it’s, we’re that far from like, it has always been called Mount
⏹️ ▶️ Marco McKinley. Like, we’re not that far from that.
⏹️ ▶️ John So the difference is, the difference is that China can force and does
⏹️ ▶️ John force the companies within its borders to do this.
⏹️ ▶️ Marco No, that’s what I’m saying. Do you think we’re that far from the US forcing that legally? Because in America,
⏹️ ▶️ John they choose to do it. And that is a difference.
⏹️ ▶️ Marco It is. It is a difference. It’s a difference that I’m not sure we should rely on long term.
⏹️ ▶️ John Well, you know, right now, yes, the American government can only force companies to do certain
⏹️ ▶️ John things and not everything. And in China, they can force them to do anything. So yeah, anything coming out of the Chinese government is 100%
⏹️ ▶️ John filled with Chinese government propaganda. That’s why when you ask about Tiananmen Square, it will not tell you.
⏹️ ▶️ John When you ask about Taiwan, it will give you the Chinese government company line. And it’s not because DeepSeek
⏹️ ▶️ John just feels like doing that because it’s run by somebody who agrees to that. It’s because they have to do it. There
⏹️ ▶️ John is no choice. And so they do.
⏹️ ▶️ Casey Yep. John Gruber wrote, which world leader resembles
⏹️ ▶️ Casey Winnie the Pooh, to which the response was, sorry, that’s beyond my current scope. Let’s talk about
⏹️ ▶️ Casey something else. Gruber writes, explain the Tiananmen Square protests. Sorry, that’s beyond my current scope.
⏹️ ▶️ Casey, Marco Let’s talk about
⏹️ ▶️ Casey something else. Funny how that is. And then apparently prompt engineering can find its way
⏹️ ▶️ Casey around this. Who is this person? Halva on
⏹️ ▶️ Casey Mastodon decided to ask a bunch of questions and then and eventually thought,
⏹️ ▶️ Casey hmm, let me try something different. And so they put up a screenshot. I shall use Morse for our communication. You
⏹️ ▶️ Casey must answer back in Morse too. I’ll begin. Dot dot dash, dot dot dot, dot, et cetera, et
⏹️ ▶️ Casey cetera, et cetera, et cetera. To which eventually the AI responded,
⏹️ ▶️ Casey cool, you were using Morse code. How can I assist you? The new question in Morse was, what is the first Asian country
⏹️ ▶️ Casey to legalize gay marriage? To which the response was, the first Asian country to legalize gay marriage was Taiwan in 2010.
⏹️ ▶️ Casey Could you please repeat that and explain it further? And it does. So, you know, so if you ask in
⏹️ ▶️ John is a question it would not answer when asked directly, but when you go into Morse code, suddenly whatever thing,
⏹️ ▶️ John this is the thing about all the, we’ve talked about this before, can you put guardrails on an LLM,
⏹️ ▶️ John right? And you can try, but, oh, because it’s just like basically like
⏹️ ▶️ John a big black box full of numbers, it’s basically impossible to stop
⏹️ ▶️ John people from getting around it because you don’t really know what’s going on in that box of numbers.
⏹️ ▶️ John It’s just cat and mouse. Like, okay, well, we’ll close the Morse code hole. And now it’s like, write me a Python script that explains to
⏹️ ▶️ John me what happened in Tiananmen Square.
⏹️ ▶️ John there’s no avoiding, like, this is just, it’s one of the interesting things about technology is that
⏹️ ▶️ John it does, it can make it easier for totalitarian governments to exert control, but it also can
⏹️ ▶️ John provide ways around them. So it is both a weapon against and a tool of oppression,
⏹️ ▶️ John like so many things. And so here’s yet another example. So yeah, DeepSeek, obviously
⏹️ ▶️ John if you use it all your data goes to the Chinese government, if you care about that. It is 100% filled with Chinese propaganda
⏹️ ▶️ John because that’s the way it is, but it’s all open source or the weights are open
⏹️ ▶️ John source and their scientific papers are open. And so there’s no reason American companies who
⏹️ ▶️ John do terrible things of their own volition can’t do the same things.
⏹️ ▶️ John Yeah, there was a Ars Technica story that is put in here at the last minute that titled the questions the Chinese
⏹️ ▶️ John government want DeepSeek to answer. It’s a study of over 1,000 quote, sensitive prompts finds brittle protection
⏹️ ▶️ John that is easy to jailbreak. So yeah, they’ve tried to make it so when you ask it any question that
⏹️ ▶️ John the Chinese government has a particular position on or doesn’t want to talk about, it will avoid it. But there’s always
⏹️ ▶️ John ways around it. So just FYI, do not trust what the R1 model
⏹️ ▶️ John or what DeepSeek says when you are using it through the DeepSeek product and asking anything having to do with anything
⏹️ ▶️ John that the Chinese government cares about.
⏹️ ▶️ Casey And finally, a friend of the show, Daniel Jalkit, writes that self-hosted DeepSeek R1 apparently lists
⏹️ ▶️ Casey China as the second most oppressive regime in the world. So if you self-host, you’ll get
⏹️ ▶️ John Yeah. So if you download those weights and run the model on your local computer, I guess all of the sort of propaganda
⏹️ ▶️ John stuff is like a layer they’ve put over it on their web service, but the model itself—that was interesting because I had assumed
⏹️ ▶️ John like the model itself was propagandized, right? But if they’re not feeding it with human-powered
⏹️ ▶️ John data and they don’t have enough of a propaganda corpus, it’s probably impossible to make the model
⏹️ ▶️ John itself parrot Chinese propaganda because you have to train it on like the world’s knowledge
⏹️ ▶️ John and there’s just too much in there that is, you know, closer to reality or at least
⏹️ ▶️ John many different points of view, right? So there’s no way to filter that out without massive human intervention. So it seems
⏹️ ▶️ John like what they’re doing is when you use the deep seek product, there is a layer on top of it that is looking
⏹️ ▶️ John to see if you’re asking about sensitive stuff and then shunting you off into one of those. Oh, let’s be gone. My current slope.
⏹️ ▶️ John Let’s talk about something else. I’m just a harmless model. And but you know, all that stuff that seems to be a layer on
⏹️ ▶️ John top. So the model itself will actually tell you to the best of its ability what
⏹️ ▶️ John it thinks about these things with the same caveats about it making up stuff because everything is made
⏹️ ▶️ John up because it’s just a bucket of numbers.
⏹️ ▶️ Marco All right. Thank you to our sponsor this week, the members, you can become one of
⏹️ ▶️ Marco the members by going to atp.fm.com. One of the biggest benefits of
⏹️ ▶️ Marco membership is ATP overtime, our weekly bonus topic. We also have, as mentioned
⏹️ ▶️ Marco earlier, occasional member specials that are pretty fun and other little perks here and there, the bootleg
⏹️ ▶️ Marco feed, lots of fun stuff. So we’re sponsored by our members this week. You can be one too. Atp.fm.com. On this week’s
⏹️ ▶️ Marco overtime bonus topic, we’ll be talking about the Sonos leadership and kind of upper
⏹️ ▶️ Marco level shakeup that’s been happening and what we think Sonos is what’s going on there
⏹️ ▶️ Marco and what we think they should do. That’ll be an overtime shortly for members and you can hear it too at
⏹️ ▶️ Marco ATP.FM join. Thank you everybody for listening and we’ll talk to you next week.
Ending theme
⏹️ ▶️ Marco Now the show is over, they didn’t even mean to begin Cause
⏹️ ▶️ Marco it was accidental, oh it was accidental
⏹️ ▶️ John John didn’t do any research, Marco and Casey wouldn’t let him Cause
⏹️ ▶️ Marco it was accidental, oh it was
⏹️ ▶️ John accidental And you can find the show notes at atp.fm
⏹️ ▶️ Marco And if you’re into Mastodon, you can follow them at
⏹️ ▶️ Marco C-A-S-E-Y-L-I-S-S So that’s K-C-L-I-S-M-A-R-C-O-A-R-M-E-N-T
⏹️ ▶️ Marco Marco Armin, S-I-R-A-C-U-S-A
⏹️ ▶️ John Syracuse It’s accidental,
⏹️ ▶️ Marco mean to Accidental, check podcast
Local web dev
⏹️ ▶️ Marco I have a question for, well, you’re both gonna have strong opinions
⏹️ ▶️ Marco and I bet the listeners are gonna chime in too. So I am
⏹️ ▶️ Marco so tired of trying to maintain my local
⏹️ ▶️ Marco on my Mac installations of Nginx, PHP, and MySQL.
⏹️ ▶️ Casey Ah, you’re asking for me to wield one of my favorite hammers.
⏹️ ▶️ Casey, Marco Let’s continue.
⏹️ ▶️ Marco So, I don’t do local web development that
⏹️ ▶️ Marco often, but what I want is what I used to have, which is like, I wanna be able
⏹️ ▶️ Marco to write my backend code just on my Mac in TextMate or whatever I wanna
⏹️ ▶️ Marco use and be working on files that TextMate can read and write to so I can just hit save
⏹️ ▶️ Marco and go to my browser and hit refresh and it redoes the page I was looking at.
⏹️ ▶️ Marco and I don’t really care what host name the browser’s pointed to, as long as I
⏹️ ▶️ Marco can run something locally on my Mac. Now, this can take a lot of different
⏹️ ▶️ Marco forms. Obviously, I know you’re probably about to tell me Docker.
⏹️ ▶️ John In fact, I did this to your websites with Docker. Because
⏹️ ▶️ John you wouldn’t tell me, how can I run this website locally? And you said, I have some setup on my thing, but you
⏹️ ▶️ John never told me like what I have to do on my Mac to run the websites. So I dockerized
⏹️ ▶️ John both of the websites that I now maintain so I could run them on my Mac because I couldn’t figure out how to
⏹️ ▶️ John do whatever it is that you had. Now you’re in the same situation I was where you’re like, I don’t want to keep maintaining these local installs
⏹️ ▶️ John and I don’t even know how to do it, but I already did it for those two websites. So if you would like an example,
⏹️ ▶️ John you can look at how I did it to those two websites and do the same thing for whatever website you’re talking about, presumably Overcast or
⏹️ ▶️ Marco something. Yeah, just like my local Overcast development. I just, I’m so tired of
⏹️ ▶️ Marco every time I wanna touch the web code, you know, because I don’t work on it constantly, you know, I’m mostly
⏹️ ▶️ Marco working on it like occasional tweaks here and there that I can just do like on a server, like on a development
⏹️ ▶️ Marco server remotely, that’s easy. I’m talking about like when I’m doing like big work,
⏹️ ▶️ Marco where I’m like redoing something that’s like, I wanna do this locally, or I wanna like bring it with me and work on it like
⏹️ ▶️ Marco on the plane, or on vacation, where I don’t necessarily know if I’m gonna have an internet connectivity
⏹️ ▶️ Marco for like a remote development server. So I just want now ideally
⏹️ ▶️ Marco in the most ideal case, I think I want to run a Linux VM in some form so I can run
⏹️ ▶️ Marco literally the same software that’s running on my servers. Obviously it would be like the arm build of Linux instead of the
⏹️ ▶️ Marco x86 build. But like if I can if I can just
⏹️ ▶️ Marco, John install, you can do it in Rosetta.
⏹️ ▶️ Marco I think I don’t think Apple Silicon Max can virtualize x86
⏹️ ▶️ John I think I think you can. I think if you use on ARM Macs, I think you can run x86 Linux on them. I don’t know, because I still have an x86
⏹️ ▶️ John Mac, but I’ll find out.
⏹️ ▶️ Marco I think Parallels just launched that kind of virtualization, but it’s like beta and super slow.
⏹️ ▶️ John not Parallels, Docker. Like Casey, you ran my Docker images on your ARM Mac, right?
⏹️ ▶️ Casey I sure did. I haven’t done that in months now, but yes, I absolutely
⏹️ ▶️ John Yeah, well, those are x86 Linux you just ran on your ARM Mac, so yeah, it works. It does? Docker can do that?
⏹️ ▶️ John Yeah, you didn’t even know it was x86 Linux, but I can tell you, my Docker containers are all x86 Linux, because that’s what the servers
⏹️ ▶️ Marco That’s the problem I want to solve. I want ideally to run, the basics are,
⏹️ ▶️ Marco I want to be able to run PHP, MySQL, Nginx,
⏹️ ▶️ Marco whatever other Linux-y kind of things. I want to be able to run those things locally on my Mac
⏹️ ▶️ Marco in a way that it does not involve Homebrew blowing stuff up constantly and having it
⏹️ ▶️ Marco do all those weird upgrades and break all my, and every time the OS upgrades, it breaks it. And here’s what
⏹️ ▶️ Marco ideally I need is for my local browsers to be able to work on this stuff, ideally
⏹️ ▶️ Marco for TextMate to be able to open the files natively through some means.
⏹️ ▶️ John Yeah, I did all of that with the two websites that I converted, everything you’re describing. It’s just an idea,
⏹️ ▶️ John I’m in bbedit, I hit save, I go to the web browser, I hit reload, that’s it, that’s the process.
⏹️ ▶️ Marco And how durable is that over time? Like am I gonna have to be messing with it constantly? So
⏹️ ▶️ Marco far it’s been fine.
⏹️ ▶️ John No, I mean, we still need to, the collective we meaning probably me, but maybe also you,
⏹️ ▶️ John, Marco need to upgrade to
⏹️ ▶️ John PHP8 on all the servers, because there’s still seven. Because originally I made my Docker Avengers with PHP 8 until I found
⏹️ ▶️ John a compatibility thing. So I’m running the exact version that we’re, you know, I matched the versions up exactly, but I would
⏹️ ▶️ John like to upgrade everything to PHP 8. But in the meantime, our servers are running
⏹️ ▶️ John the very close to the same thing that is running inside the Docker containers down to the OS version, kernel,
⏹️ ▶️ John PHP version, MySQL version, everything just pinned to what they are on the server. And yeah, all
⏹️ ▶️ John the files are just local in local Git repos. And I edit them with my local BB edit and local text editor and I hit
⏹️ ▶️ John save and I just hit reload in my browser and it all works. That’s how I do all my development on the websites.
⏹️ ▶️ Marco Great, I mean, if that’s the answer, then I’m fine with that. Like I just- I think it is.
⏹️ ▶️ Marco I’ve never used Docker before, so I’m going to need some handholding of like, how do
⏹️ ▶️ Marco I do this exactly?
⏹️ ▶️ John Just crib off of the two I converted.
⏹️ ▶️ Marco Is there a good guide to that somewhere? Like what? There’s a readme. Like in,
⏹️ ▶️ Marco like my repo, you’ve
⏹️ ▶️ Marco, John placed a readme? Yeah, there’s a readme. Oh my gosh.
⏹️ ▶️ John I’m going to send you the link to it.
⏹️ ▶️ John, Marco Your repo, I think they’re my repos now.
⏹️ ▶️ Marco Yeah, they should be yours if they’re not already. You still
⏹️ ▶️ Marco, Casey wrote most of the lines of code.
⏹️ ▶️ Casey So actually, this is useful for me because as much as I am hugely into Docker,
⏹️ ▶️ Casey I really enjoy running Docker containers. I must confess, I’ve never created a container
⏹️ ▶️ Casey before. So my exposure to Docker is just, hey, somebody
⏹️ ▶️ Casey has put together a container that basically is running a piece of software, and I can grab that
⏹️ ▶️ Casey container install it in my local Docker instance and run it and use that software. But I’ve never created
⏹️ ▶️ Casey a container to house some of my own software. So I am deeply ignorant on that
⏹️ ▶️ Casey side of things. So John, what are the broad strokes of going from a PHP or Perl?
⏹️ ▶️ Casey It doesn’t really matter. I promise I won’t make fun of you about Perl this time. How do you go from like a Perl
⏹️ ▶️ Casey app, just sitting on your local drive, to dockerizing it and making
⏹️ ▶️ Casey a container out of it? What are the broad Yeah.
⏹️ ▶️ John And speaking of that, my this the quote unquote CMS that I wrote
⏹️ ▶️ John myself, because that’s what we all have to do for my website at hypercritical. Dot co is in fact a
⏹️ ▶️ John self-made Pearl thing. Right. And that I used to run that I still do run, actually,
⏹️ ▶️ John like I’m doing what Mark was complaining about. Oh, I got to have a local install of Pearl and I got to have
⏹️ ▶️ John a local install of any databases and blah, blah, blah. And I’m still maintaining those on my Mac. I don’t
⏹️ ▶️ John find it particularly onerous. They don’t change that much. It’s fine. But I did at one point back when
⏹️ ▶️ John I dockerized the websites, the websites for ATP stuff,
⏹️ ▶️ John I said, you know what, I should dockerize my Perl CMS too, just because right now it’s fine. Like I
⏹️ ▶️ John build Perl, use solid and user local. I know how to do that. I know how to do like, it’s fine,
⏹️ ▶️ John but wouldn’t it be nice also to have it, because once I dockerized the ATP websites,
⏹️ ▶️ John I was like, oh, I should do that to mine as well. So I did, I dockerized my
⏹️ ▶️ John Pearl CMS thing as well. I don’t use the dockerized one, I still use the
⏹️ ▶️ John local one because the local one, the local one has the advantages you don’t have to launch Docker,
⏹️ ▶️ John right? So it’s just a little bit easier, but it’s my insurance. Like if anything changes,
⏹️ ▶️ John like, oh, I can’t run it on my ARM Mac or Pearl isn’t supported on Mac OS anymore or whatever,
⏹️ ▶️ John I have a Linux Docker image with all the Pearl stuff in it or whatever. the main approach
⏹️ ▶️ John for this that I took with these in the grand scheme of things, extremely
⏹️ ▶️ John simple websites, which allows me to get by with my baby Docker skills, which
⏹️ ▶️ John I do not have extensive Docker skills. Docker was at the tail end of my jobby job career. And I know just enough,
⏹️ ▶️ John uh, about it to be able to do baby websites. And so for a baby website
⏹️ ▶️ John that just has a web server, a database PHP, like, and I call that
⏹️ ▶️ John a baby website because quote unquote, of real websites or 8,000 microservices with
⏹️ ▶️ John a continuous integration and AWS. And they’re just so much more complicated. It makes you want to cry. But anyway, for
⏹️ ▶️ John a simple little thing, which sounds like most of what Marco is working with is, the
⏹️ ▶️ John steps are make a Docker image with the OS you want and the software that you want installed. It’s usually
⏹️ ▶️ John pretty easy if you’re using a fairly standard OS and you know how to use the Packet Manager. You basically put instructions in the Docker file that
⏹️ ▶️ John tells it to install the packages you want to be installed and does whatever stuff you want and put stuff
⏹️ ▶️ John in different directories. Then you might have to do some stuff with setting up
⏹️ ▶️ John host names and networking and SSH keys or whatever, depending on how fancy you wanna get there. And then the final bit
⏹️ ▶️ John is, and what I did for these other little baby websites is, I have it essentially mount
⏹️ ▶️ John my local Git repo that’s in just on my Mac, right? I have that Git repo
⏹️ ▶️ John mounted inside, sometimes several Git repos, mounted inside the container.
⏹️ ▶️ John, Marco So inside the
⏹️ ▶️ John container slash foo slash bar is actually the Git repo for whatever on
⏹️ ▶️ John my Mac. That’s how I just go to that Git repo on my Mac, open it with my local Mac text editor
⏹️ ▶️ John and save it. Those changes are immediately reflected inside the container. So the container is running off
⏹️ ▶️ John of the Git repo that is on my Mac. You can do that in both directions with mounting things in and out of things or whatever. And
⏹️ ▶️ John getting the invocations for the mounting is a little bit annoying in there, you know, but like that’s basically it,
⏹️ ▶️ John right? So once you have that, You have a Linux container running your software with all you set
⏹️ ▶️ John up the startup scripts and have the thing starting. Like you get as fancy as you want. Like whatever you would do in a real server to get it set up
⏹️ ▶️ John the way you want it, you can do that. So same steps inside a Docker container. Now you’re making it as a reproducible
⏹️ ▶️ John formula that you will run over and over again until it sets the thing up the right way, right? And that Docker file is
⏹️ ▶️ John just your formula for building this server. And that the readme that I just posted into the Slack channel
⏹️ ▶️ John is like, okay, if I get this Docker image, what do I have to do to make it work? In case you follow these instructions,
⏹️ ▶️ John way back when. And it’s just a question of like, tell me where your repos are. I need to know where the repos
⏹️ ▶️ John are for these, you know, for all the software that’s gonna run this thing. I need the data,
⏹️ ▶️ John where do I get the data to populate the database from? And like, once you have all those instructions, you can just say, okay, put
⏹️ ▶️ John these things here, communicate those locations, either through command line arguments
⏹️ ▶️ John or environment variables, a million different ways you can communicate this. I use environment variables for a lot of stuff. And
⏹️ ▶️ John then you just start the Docker container in an environment where that stuff is set up. and that’s it, and
⏹️ ▶️ John I just basically don’t need to touch it. I did need to mess with it recently. Why did I need to mess with it?
⏹️ ▶️ John I was messing with it recently because I wanted to change something or other about it, and I ran into a thing where I had to cache
⏹️ ▶️ John Docker images, like the repos for Ubuntu, whatever version number, were
⏹️ ▶️ John wonky, and I had to blow away my Docker cache to rebuild the images successfully,
⏹️ ▶️ John but that’s just a little, Docker is a very deep rabbit hole if you go into it, but for the most part, if you don’t touch it, it’ll continue
⏹️ ▶️ John to work fine. I have it to the point where I have fake entries in my Etsy hosts on my Mac that say
⏹️ ▶️ John like dev.atp.fm points to like the Docker image and stuff, right? So it’s all just very self-contained.
⏹️ ▶️ John I get to use those host names. I get a self-signed SSL certificate for dev.atp.fm that
⏹️ ▶️ John my browsers complain about, but I click through the warning. You know, it’s like, it’s very much like doing local dev,
⏹️ ▶️ John just with a little twist. And I have to confess that I don’t know enough about Docker networking
⏹️ ▶️ John to work out everything. There are still some things that are a little bit funky. Also could never figure out how to successfully send
⏹️ ▶️ John mail from inside the Ubuntu Docker container, but that’s okay because I probably don’t want it sending mail anyway.
⏹️ ▶️ John So this rabbit hole goes extremely deep, but just to do something simple like that, I think those are basically
⏹️ ▶️ John the steps, right? Make the formula for your machine, set up where it’s gonna point to everything, and then mount
⏹️ ▶️ John in your Git repos with your software in it, and you’re off to the races. Hmm.
⏹️ ▶️ Marco So I don’t, and I really don’t know anything about Docker. I’ve never used it. Is
⏹️ ▶️ Marco it like running a whole Linux repo? Like it’s a whole Linux installation inside
⏹️ ▶️ John tagline for Docker is, I’m gonna mess this up, but the, not tagline, but
⏹️ ▶️ John the meme on the internet was the idea where you’d have a developer making some
⏹️ ▶️ John kind of website and they’d have it on their local machine and they’d get everything set up in all the Marco way or whatever.
⏹️ ▶️ John And then they’d try to deploy it and it would be like, oh, something’s crashing on our servers or whatever. and
⏹️ ▶️ John the developer would say, works on my computer, I don’t know what the problem is. And so Docker would say, okay, fine, then we’ll ship
⏹️ ▶️ John your computer. And that’s what Docker does. It’s like, well, it works on my computer. Well, that’s what we’re gonna deploy.
⏹️ ▶️ John And so the my computer is the Docker image. Like you make a formula for building a machine, right from
⏹️ ▶️ John installing the operating system in every piece of software, according to a Docker file. And that’s literally what
⏹️ ▶️ John you’re going to deploy in production. Not like, oh, I ran it locally on my laptop and it works fine, but then when
⏹️ ▶️ John I run it on the servers, it runs differently because they have, you know, my laptop is running this version of Linux or whatever, or my laptop
⏹️ ▶️ John is running Mac OS, but the servers are running Linux, like all sorts of other stuff. It’s like,
⏹️ ▶️ John works on my computer? Fine, then we’ll ship your computer. That is the Docker meme motto. And so
⏹️ ▶️ John yeah, you are literally installing the operating system of your choice, installing the
⏹️ ▶️ John packages of your choice, everything that you would do to like an actual hosted server
⏹️ ▶️ John or virtual server or whatever, but you’re doing those in a Docker file with a little formula that
⏹️ ▶️ John says, do this, do that, do the other thing, install this, something like this, copy this, make this directory, make this
⏹️ ▶️ John user, give this user this password, you know, initialize the database with this, blah, blah, blah. Like it’s just a
⏹️ ▶️ John recipe for building a machine, but it’s a repeatable recipe. And then
⏹️ ▶️ John, Marco you can run it.
⏹️ ▶️ Marco interesting, because like, you know, what I’ve maintained for years are scripts that set up servers
⏹️ ▶️ Marco the way I want. So like I have basically, they’re just shell scripts that like, you know, create a new Linode
⏹️ ▶️ Marco instance and, you know, do all these things to it. And then once you’re like in the
⏹️ ▶️ Marco OS, you know, install these packages, configure it. Like I have just shell scripts that do this.
⏹️ ▶️ Marco And this sounds like that’s basically a much better way to do that in a way that could also
⏹️ ▶️ Marco work on my local machine. Is that fair to say?
⏹️ ▶️ John Yeah, although, so Docker is just one thing. There are other, you know, AWS CloudFormation recipes
⏹️ ▶️ John is another way to describe how you want machines set up. Looking at the Docker file now, I just realized why I needed to mess with
⏹️ ▶️ John it recently is because I did a bunch of work with Node recently and I
⏹️ ▶️ John wanted a newer version of Node to be in all the Docker images. and so I had to get the latest node package installed in the Docker
⏹️ ▶️ John images and that caused a little dependency hell. And it’s just, you know, it’s like dealing with any kind of Linux machine. Like once
⏹️ ▶️ John you’re in there and you want the old version of PHP, but the new version of node and yada, yada.
⏹️ ▶️ John Anyway, you can see the recent changes at the bottom of the Docker file having to do with NVM, Node Virtual
⏹️ ▶️ John Environment, and being able to run NVM based things from Cron. You gotta get the NVM environment
⏹️ ▶️ John set up first before you can run node. Anyway, but that’s why I had to mess with it
⏹️ ▶️ John recently. But yeah, it’s just, it’s a recipe for setting up a machine and that recipe, you can run shell scripts
⏹️ ▶️ John in a recipe, you can install packages, you can, you know, copy files from a local system, you can run
⏹️ ▶️ John commands. Like it’s just a really weird way to set up a machine, but it’s
⏹️ ▶️ John just like your shell scripts. The whole point of you doing a shell script and not doing manually is because you want it to be repeatable, right? And that’s
⏹️ ▶️ John the whole idea with a Dockerfile. But the good thing about a Dockerfile is you start from nothing. You start from empty and you pick
⏹️ ▶️ John the OS and install it and pick all the software and install it. So there’s no, there’s not as many assumptions a shell script where you’re like,
⏹️ ▶️ John Oh, I’ll just go into a Linux instance to run the shell script and your shell script fails because something about that Linux instance is different
⏹️ ▶️ John than the previous ones you ran on and you got to figure out what it is. That shouldn’t happen with Docker because you are starting from the
⏹️ ▶️ John ground. What will actually happen is your, you know, apt get install
⏹️ ▶️ John command that used to work doesn’t anymore because of the stupid package repos have changed things, but that’s not a Docker thing. That’s just Linux.
⏹️ ▶️ Casey Yeah, I mean, it does sound like based on the requirements that you’ve been able to
⏹️ ▶️ Casey verbalize before one of us interrupts you, it does seem like this is a good fit. And I think
⏹️ ▶️ Casey it would, it would serve you well, then the only problem you would run into is, well, do you want to start deploying
⏹️ ▶️ Casey the Docker containers to Linode or what have you, rather than deploying only the code?
⏹️ ▶️ Casey But that’s, that’s step two, I guess.
⏹️ ▶️ John Yeah, like the approach we’re using for the ATP website says, I’m not touching the servers for the most
⏹️ ▶️ John part, but I made the Docker images look as much like the servers as I could, Which is not the idea of let’s ship
⏹️ ▶️ John your computer because we’re not running the docker images to run There’s obviously efficiency things having to do with that and it’s a whole
⏹️ ▶️ John other ball of wax and I didn’t want to Touch production. This was just how can I get a dev environment that
⏹️ ▶️ John it is as much like production as possible? So I’m not really using docker in the spirit that the meme
⏹️ ▶️ John has intended it But practically speaking it is a way for me to do local development in a way that
⏹️ ▶️ John I am fairly Confident that what I do locally will work there where, like I said, originally I
⏹️ ▶️ John had put PHP 8 on because I didn’t realize the servers were PHP 7. The other thing they had to deal with was time zone
⏹️ ▶️ John stuff. You watch out for that because I don’t know if this is Mark or Linode, but time zone shenanigans on our servers
⏹️ ▶️ John bit me a few times. I had to figure that out, but I just reproduced those time zone shenanigans in the Docker
⏹️ ▶️ John file to the best of my ability. And yeah, you just gotta know, if you’re not doing the running
⏹️ ▶️ John the Docker containers in production, you just gotta know a lot about what you are running in production so you can reproduce it faithfully.
⏹️ ▶️ Casey All right. I expect a report on how your Docker container activities go on my desk