A better keyboard than QWERTY

26:35And then I did nothing. Just nothing. The thing is, I type too much too often, and I type really fast, and it just seems like so much work to essentially learn what amounts to a new instrument just to be able to do my job all day. I'm not sure how long I could get away with not filing stories because I can't type fast enough. You know what I mean? So I'm stuck sitting here staring at this stupid QWERTY keyboard on my desk, which apparently everyone agrees is a bad option, but it's kind of just the one we're stuck with. And so now I'm looking around my desk wondering what else is like this. What other tech is this way, just because it is this way?

27:13We talk about AI as upending the way we do everything online, and maybe that's good, even if the AI isn't good and ultimately isn't going to change everything about how we do everything. Because maybe it's good to sit down and rethink how we do everything every once in a while. My image of the ultimate keyboard was just a regular keyboard with cooler switches. So that was the extent of my knowledge. I knew I had this Ergodox, but it's like, that's just a weird, strange thing. But I mean, I think it's when you start going down the rabbit hole, the hole starts to open and you realize, oh, there are so many other things.

27:53All right, we're going to take a break and then we're going to come back and talk about Surface, Windows, and the future of Microsoft. We'll be right back.

28:03Support for this show comes from LinkedIn. When you're a small business owner, your business is on your mind 24-7. So when you're hiring, you need a partner that works just as hard as you do. That hiring partner is LinkedIn Jobs. When you clock out, LinkedIn clocks in. LinkedIn makes it easy to post your job for free, share it with your network, and get qualified candidates that you can manage all in one place. LinkedIn's new features can help you write job descriptions and then quickly get your job in front of the right people with deep candidate insights. Either post your job for free or pay to promote.

28:38Promoted jobs get three times more qualified applicants. At the end of the day, the most important thing to your small business is the quality of candidates. And with LinkedIn, you can feel confident that you're getting the best. Based on LinkedIn data, 72 % of SMBs using LinkedIn say that it helps them find high quality candidates. Find out why more than two and a half million small businesses use LinkedIn for hiring today. Find your next great hire on LinkedIn. Post your job for free at linkedin.com slash track. That's linkedin.com slash track to post your job for free. Terms and conditions apply.

29:14Support for the show comes from Crucible Moments, a podcast from Sequoia Capital. Sometimes the difference between success and failure comes down to one chance encounter, or following a counterintuitive instinct, or ignoring conventional wisdom to make a bold decision. Like when the founders at Palo Alto Networks wanted to redefine cybersecurity for the modern age. Everybody thought we were crazy. Nobody would use the cloud for cybersecurity. And when I hear something like that, I say, of course, I'm going to do it. You know, if everybody believes it shouldn't be done or it cannot be done, then, you know, it's a good reason to do it.

29:48Or when mobile gaming giant Supercell could only rewrite the rules of the industry after failure in the company's formative stages. Many of the best things we've learned have actually come through failures. These are crucible moments, turning points in a company's journey that made them what they are today. Hosted by Sequoia Capitals, Ruloff, Botha, Crucible Moments is back for a new season with stories from Zipline, Stripe, Palo Alto Networks, Supercell, and more. Subscribe to Season 3 of Crucible Moments. New episodes are out now and you can catch up on Seasons 1 and 2 at cruciblemoments.com on YouTube or wherever you get your podcasts.

30:29Listen to Crucible Moments today. Support for this show comes from Zapier. When people talk about AI, they often talk about the future. How will AI affect the economy? What will AI's potential be in the future? Will AI be able to make popular art? But the more important question is, what can AI actually do for your business today? If you're looking for concrete implementation of AI for your business, you can try Zapier. Zapier says they can help you and your team actually reap the benefits of AI, not just talk about it. With Zapier's AI orchestration platform, you can bring the power of AI to your work.

31:05Connect top AI models like ChatGPT and Claude to the apps your team already uses, so you can add AI exactly where you need it. Whether that's automatically enriching leads, coaching your sales rep, resolving IT tickets, or more, you can do it all with Zapier. Zapier is for everyone, tech expert or not. No IT bottlenecks, no complexity. No AI hype, just results. Teams have already automated over 300 million AI tasks using Zapier. Join the 3.4 million companies already automating with Zapier and transform how you work with Zapier in AI. Get started for free by visiting zapier.com slash verge. That's Z-A-P-I-E-R dot com slash verge.

31:56Welcome back. Over the course of the last few weeks, there's been a lot of what I'd call sneaky shakeups happening inside of Microsoft. I mean, nothing to the level of last year when it looked like Microsoft might just hire the entirety of OpenAI. And of course, all of that turned out to be nothing. But there have been a few corporate moves and a couple of product launches and even a different non-acquisition acquisition that all make me think there is something big going on here. in terms of how Microsoft thinks about AI and Windows and Surface in particular. So it's time to do what we always do in these situations, which is get The Verge's Tom Warren on the line to help us figure out what's actually going on here.

32:36Tom Warren, welcome back to the show. Hey, good to be back. I feel like I have to bring you on every time Microsoft does something weird and confusing and you have to make sense of it. I feel like this is not what your job should be, but this is what I need you for. It does feel like that's my job. It's just explaining Microsoft's mess, essentially. Just forever like reading the org chart of Microsoft to make sense. And so actually the org chart is where I want to start. I want to talk about AI PCs and I want to talk about services, but we're going to get to all that in a minute because I actually think what is happening inside of Microsoft actually explains a lot of that stuff.

33:08So let's go back to six-ish months ago, I think, which is when Panos Panay announced he was leaving Microsoft. Describe that moment to me at Microsoft because he's a guy who'd been there forever. He was a big face of Microsoft. He was in charge of a lot of things. like what happened when Panos left? Yeah, I think it was kind of a surprise, right? I think it took us by surprise and I think it probably took Microsoft by surprise as well. Like the thing is, it was a few days like before a Surface event, which was also like timing, you know, that they announced it publicly anyway. And I think like catching up with people at the event, you got that sense that it was a surprise, you know, this wasn't expected to happen.

33:47So yeah, I think that was the big takeaway was it was a surprise. And I think the way that Microsoft kind of responded to it, like Nadella's shapeshifting of Windows, was that they split Windows and Surface. So Panos' responsibilities into two different leaders. And if I'm not mistaken, Microsoft had made a big deal out of uniting those things under Panos, right? That was supposed to be this coherent, holistic way of thinking about devices, right? And that wasn't that long ago. No, it wasn't. And it was around, actually, I think it was just before the pandemic started. Oh, wow. Okay. Which I guess at this point is a while ago, but it doesn't seem like that.

34:24It was a while ago. It doesn't feel like it, right? Okay. But it was interesting timing because it was a few months after Panos and the rest of the Surface team had unveiled the dual screen Surface stuff. And that was going to be the future of Windows and all that sort of stuff. so it seemed like they you know putting everything under panos was gonna align surface hardware and windows software so that seemed like that was what was going to happen that's what panos been working on and then they just split it out in a really weird way so obviously panos left for amazon they gave the surface hardware stuff to pavan davaluri he's been at microsoft for over 20 years he's pretty key surface engineer he's definitely a key player there it was pretty instrumental to the qualcomm and amd relationships where they've done the customs sq1 processes and stuff for these so like a natural guy to put in charge of this after upon a safe okay yeah definitely for the hardware side and then they split off the windows sort of engineering efforts and all those teams over to to mikhail parakin who had been at microsoft for a number of years as well but was essentially his official title was formerly now but uh ceo of like web and advertising.

35:35I feel like he's a name I started to hear with some of the AI stuff over the last year. So is that right? Yeah. So he was looking after, you'd hear him a lot when he was tweeting about the Bing chat stuff. Oh yeah. Okay. Edge maybe, all that sort of stuff. So it was difficult for me to wrap my head around this because it was like, it didn't seem like the natural person to take over Windows, just given his role. But he's obviously a talented engineer and leader, but it It seems strange for it to go that way, given where Windows and Surface have been. So that happened about six months ago. He's been looking after that and then not so much anymore.

36:11Yeah, so that brings us to now. And basically that has been, at least as far as I can tell, sort of pretty cleanly undone. Pretty much, yeah. So the teams that, I'd say most of the teams that went over to Mikhail's organization have gone back, essentially, if you want to call it that, to Pavan Davaluri, who was obviously looking after the Surface hardware. So he's now looking after Windows, most of the Windows engineering and Surface hardware. The key thing is that it all kind of ties into a big move that Nadella made. So this is why it's actually happening. And he's obviously hired the former co-founder of Google DeepMind, Mustafa Suleiman.

36:52And he hired him into a position as CEO of Microsoft AI. So a new group inside Microsoft, or you might say the same group that Mikhail was looking after and has kind of been pushed out slightly. He's given a lot to Microsoft and particularly on their ads and search side. They've become a lot more relevant in the last few years on that side than they've been before. You can argue they've been pushing Bing in weird ways, pushing Edge in weird ways that haven't been too great for consumers. You're saying you don't love giant pop-ups telling you to use Bing chat? I love it when I'm in the middle of playing a game or like in my Excel spreadsheet and there's this giant, please use Copilot or whatever.

37:32Yeah, perfect. Yeah, it's the clippiest thing Microsoft has done since Clippy and I'm not a fan. But okay, so if I understand sort of the org structure now, you have basically the sort of Windows and devices side of things, which is basically like hardware that people use. And that's all under Pavan Davaluri now. He has like most of the Parnos job is now his. Yeah. And he's not like he's still not on like the senior leadership team. So he still reports up to someone. But essentially, that's what he's looking after. And I think I get the impression that most of that work that he's going to be looking after is how they optimize Windows for the new silicon that's coming.

38:13And, you know, the NPUs, all that sort of stuff, because that seems to be pretty key. And we'll see you talk about that soon, but that's why I think they're bringing it back to that side because they know they need that engineering effort. Got it. Okay. I do want to talk about that in just a sec, because I think how all of this comes back together in the product is the part of this that I understand the least. And we're going to talk about that. But the other side of this, we haven't really talked about all of the inflection AI, non-acquisition acquisition stuff going on. So basically, as I understand it, Microsoft hired Mustafa Suleiman, who is the co-founder of DeepMind, and then went and founded this company called Inflection AI, paid a ton of money for him and a bunch of other executives to come join the company, didn't call it an acquisition.

38:57A lot of people are saying that's because it was worried about the antitrust ramifications of calling it an acquisition. But essentially, they acqui-hired or like whatever the opposite of an acqui-hire is, like hire-sitioned a bunch of inflection AI people to come do AI for Microsoft. Which again, to me, says Microsoft doesn't feel like it is where it needs to be in terms of like winning the AI race. And this is the company that 12 months ago was just as confident as you could possibly be about the future of AI and Bing being the thing and making Google dance and that they were really going to win.

39:31This says to me that Microsoft is much less feel in itself maybe than it was a year ago. What's your read of this situation? I think that's spot on. Yeah, so a year ago, obviously, and I've written about this recently, they were obviously all in on Bing and Bing chat, all that sort of stuff. and internally they were talking about picking the bing brand because obviously there's i think particularly among tech people anyway and then the general consumer can look at bing and probably be like don't even know what it is maybe but along the along the tech sort of um people is is kind of laughed at right it's frowned upon the bing brand yeah and they were talking internally at that time like you know it's it's 500 million or whatever it's a marketing spend to pick a new name so they were like we're going to go with bing that's the way forward it didn't really shift the needle right like nadella wanted google to dance they wanted to get all these you know percentage points off of google i don't i think they probably only got about one or two percent if that otherwise they were they were aiming for for more for there to be an impact you know my impression is that most people who use bing through microsoft think they're using chat gpt i can't prove that but like when i talk to people who are using bing as ai stuff they think they're using chat gpt which is a bad sign for microsoft yeah that's probably reasonable and they obviously rebranded the bing chat stuff to copilot right and they've done a super bowl ad all that sort of stuff so that's that's all linked in with this and like when i've been speaking to people about working under mikhail it sounds like it was run as as kind of like a startup atmosphere there where you're constantly work like you're trying to do as much out of as little resources as you have so it seems like it was kind of a bit crunchy over there and i think there was some micromanaging going on there as well there's some of that sort of stuff that's been happening which to be fair probably happens in a bunch of organizations anyway but it seemed like it just i think from the fact that nadella has gone and hired someone into this ceo position it clearly wasn't working right like whatever they were doing in that team wasn't working and that's the key point that i think bloomberg reported that nadella was frustrated at that team or frustrated that the timeline's there or something because he obviously wants to go fast and microsoft has been going very fast and co-pilot in office um and in that world where where they can push it and actually there's revenue opportunities there and you see it in the stock price and all that sort of stuff so they want to really push it on the consumer side as well and obviously it hasn't skyrocketed to relevance so but i will say though that i think it's the first time i've known people to be downloading the bing mobile app or actually talking about bing right that's true yeah so it's it's had some sort of impact but obviously not not enough and it does seem like i will say credit where credit's due for microsoft in terms of like baking co-pilot into the office apps and stuff and and sort of giving people like real actual business uses for ai it's done a better job i think than anybody right chat gpt is sort of neater uh and gemini is weird and whatever and as a as consumer brands those are really the two that people talk about but in terms of like the whatever 30 a month you're trying to get it departments to pay for i actually think microsoft is at least my own perception ahead of everybody else on that front so credit where credit's due for that it's just that's not a victory for bing no exactly like the good thing about copilot in in that office space is the team's integration where you can literally like you know how you're like do i really need to go to that meeting or not can i catch up like in the notes and you don't even have to worry now you just click on a link and it tells you if someone even spoke about you and you can just click like and it will just instantly go to that part it's like oh that's it's pretty yeah it's pretty wild what they've done for the team's integration and and also like your outlook um like i tested out the copilot pro stuff which is for consumers actually um the team stuff isn't available for consumers but i tested that out and the most impressive part of that is you go into outlook and there's an email is summarized the email at the top of it so which is really useful when you have a back and forth with like customer support rep and you're like i can't even remember what i was doing in this email like there's there's loads of little interesting touch points um to it it's still a little bit still and it needs to mature and stuff yeah but yeah they have they have nailed some of it on the office side but on the windows side they've been promising a bunch of stuff and it hasn't i'm not sure what this aipc stuff is yet Yeah, me neither.

43:56So let's talk about that. Because we came on this show a few weeks ago when Apple announced the new MacBook Airs. And we made a lot of fun of Apple for saying that these are really great AI PCs, which amounted to essentially like you can do some photo editing and also use the internet. Thus, these are AI PCs. and Microsoft and Intel and others are out here making all this noise also about AI PCs and they have the co-pilot key and it feels like if they want to be consumer winners in this space it seems like Microsoft is betting a lot on really having like a hardware instantiation of AI be the thing yeah but like just back all the way up for me like when Microsoft says AI PC the hell is it talking about that's what i want to know yeah that's the key question at the moment yeah like i think aipc is definitely a bit of a marketing term right now because microsoft saying all this stuff and it's like the definition which we actually learned this week from intel is that it needs to have the latest cpus and gpus and an mpu right so the neural processing unit which is a chip that we've seen on a bunch of arm powered windows laptops for years now and essentially it's been there and accelerates tasks like background blowing on a video call and like noise cancelling when your dog's barking behind you on a call so there's obviously other instances and software developers are working on stuff that will utilize it but for all intents and purposes that's the most consumer friendly option you have for that mpu right now so what else is the mpu going to do what other ai tasks are going to leverage that in windows that's still the mystery so although microsoft's coming out saying here's our surface pro 10 for business yes that's the full name of the product and it's our first surface ai pc it still doesn't really explain like why like because the ai stuff that they have been pushing is copilot which is cloud-based so it doesn't need an mpu locally to accelerate anything so there's just that that's a missing part of the puzzle and i think that's all going to make sense on may 20th when they when they do this surface and windows event yeah what do we know about the surfaces so far microsoft just released a couple of businessy surfaces yeah which seemed like that reading your story about them you seem to see those as sort of a hint at what we might see from the consumer facing surfaces but also those might be different and i don't know like what what's your sense of kind of how much we know about this next generation of surfaces that's coming yeah so so they announced the surface pro 10 for business and the Surface Laptop 6 for business.

46:32And they're both running Intel Core Ultra chips, which do have MPUs. And I'd say beyond that, there's some webcam changes, but the designs are pretty much the same. It's pretty much a very plain release, but some good upgrades for business customers. And the hardware's good. Like I like the Surface and the Surface laptop. So I think to the extent that it's like desperate for a hardware redesign, like whatever. But I think that what you just described to me in no way makes these things qualify as AI PCs. So that's still where I'm stuck, right? It's like, what is the AI PC story inside of this thing?

47:05So on May 20th, they're going to do the same devices, but for consumers. There'll be some tweaks to the designs. The Surface Pro 10 is supposed to get an OLED display, which is nice. All sorts of stuff you expect from the consumer side. But more importantly, they're not going to be Intel. So they're not going to be Core Ultra PCs. they're going to be Qualcomm, Snapdragon, X Elite, or Elite X, I can't remember which way it goes. So ARM-based, which is not unusual for Microsoft to do ARM-based surfaces. They've done Surface Pro X in the past, but it's unusual for them to only do that for consumers.

47:40Yeah, as the main option, that's a pretty big deal. I mean, they could, I guess, offer it later. It could be some sort of exclusive. But as far as I know, they're launching exclusively with Qualcomm, which is, I think, that speaks to some sort of confidence in these qualcomm chips because the qualcomm's been promising this performance to match macbooks and stuff for years and it's never really delivered right so the year of windows on arm has been every year for like a decade at this point yeah but it seems like microsoft is really confident from what i understand internally that they're going to even be able to beat m3 performance which is wow which is interesting yeah because even if they can beat that would be a huge deal Yeah, like it seems like they're unusually confident, which I don't usually see.

48:24I mean, forget M3 performance. If it can get to like Apple's M1 performance on these things, like that's a huge leap forward, both for Qualcomm in general, which can then go sell these things to other people also, but also for the Surface line. Like that would make the Surface probably more interesting to more people than it has been in a really long time. Yeah, I think it's going to be super interesting, obviously, because these chips are obviously going to go into regular laptops as well. And it does feel like with the rumors of, was it AMD and NVIDIA potentially doing Windows on ARM next year or the year after?

48:58I can't remember the exact year. It does feel like there's something going on here. And Intel keeps inviting me to briefings about the Core Ultra. So they obviously know that something's going on. So I feel like there is something going on here, but I'm not fully committed to trusting that this is going to be M3 performance. Yeah. And in terms of the AI stuff, right? Like I think the one part of the AI story that I do buy is that basically efficient performance goes a really long way. Yes, definitely. Granted, every computer with good performance and good battery life is a better PC for AI. Yeah.

49:30Is that an AI PC? No, that's just like a pretty good laptop in 2024. But I digress. But as you we've seen Microsoft on the consumer side, I feel like try a bunch of different ways to put AI in front of people, right? There's Bing Chat. They've made all this big deal about Edge. They're trying to put it in the start menu. What's your sense of what, if anything, is working there? What would you bet on being the front face of AI on Windows going forward? Yeah, I think Copilot is definitely the front face. So Intel obviously had these specs and there was the chip side, but then there was the, it has to have access to Copilot software, which is like, sure, it's Windows 11, whatever.

50:08But it has to have the Copilot key as well for it to be an AI PC in Microsoft's definition, which is like, okay, interesting. So it has to do AI and it has to have an AI button. That's what I'm hearing. And then it's an AI PC. I don't know if they're going to have stickers on these things, but I mean, you never know. That would be amazing. instead of intel inside it should just start saying ai inside like that's the stuff right there that's it but copilot is definitely their their portal now how they'll get away from it just being a chatbot or seen as just a chatbot on the side in windows which is basically what it is right now that's going to be interesting to see how it integrates for the apps they do have something coming called i think it's called ai explorer and it sounds real creepy but it's going to essentially record everything you do on your pc and then say you were i don't know researching a holiday last week you could say to it i was looking at holidays last week but i can't remember where it was and then it'd be like oh you were looking like holidays to europe last week do you want to check out the tabs that you had oh sure and then you click and it brings up all the tabs you're working on documents whatever so it's like a timeline of how you're using your pc and the state it was in and all that sort of stuff don't know exactly how it's going to work in reality but that's that's what they've been that's that's what they're building at the moment essentially so that's going to be the the killer the killer feature for because it's going to leverage the mpu um obviously to to quickly pull up this this stuff and and reference it which sounds very similar to the windows 10 timeline feature they tried to do which was like a similar sort of thing but sort of going across mobile apps um but it needed app support so it was kind of like dead in the water without developers actually supporting it this doesn't need app support they can just do it because they you know it runs on top of windows it knows what's on the screen yeah i mean privacy minefield but yeah well yeah i i remember i wrote about this company uh rewind maybe a year ago that is doing something very similar and there is this like a separate app that you download that then gets permission to record your screen and all that stuff and in talking to people about that everybody immediately goes one of two ways where it's like oh awesome that runs locally on my computer and doesn't go anywhere else kick ass what an incredible productivity tool that's going to be and then a bunch of other people who are like absolutely not under no circumstances never ever ever ever ever yeah and i literally i talked to a bunch of people around that story and found no one who was in the middle of those two points yeah i think that's definitely what this is gonna be like yeah it's obviously gonna run locally and all that but it's still it doesn't matter if it's recording you it's still weird right yeah i agree but i will say if that becomes kind of part of the interface of a computer which i actually think done well is pretty interesting the idea of like because i mean you think about your browser history, right?

52:50Like the fact that you can't sort of query your browser history to be like, what was that page I was looking at about X, Y, and Z is silly. Like you should be able to do that. The information's right there. It has it anyway, right? So to be able to take information that like, of course it already has and do something with it, that's a pretty compelling use of AI to me. Again, I wonder what that will look like and whether it will totally collapse this supposed good performance and battery life of my computer. Yeah. And advertising companies basically have access to your search history i mean not not in the literal sense but by tracking you all around the web so and they can make relevant queries so it would be really useful to be able to look back in useful ai powered ways so yeah i see that i see the big benefits of it it's just how they sell it and the privacy aspects and everything else do you think all the other windows manufacturers are going to be into the co-pilot idea and co-pilot key the same way that microsoft is yeah i think probably because usually they follow.

53:48I guess they don't have much choice in a lot of places. Microsoft will say you can put some of your stuff where you can make a tiny bit of revenue on everything. They have all these big deals with OEMs to incentivize all this sort of stuff. So they'll have like marketing deals. It would just basically be one of those things where it's like if you're an OEM, you kind of have to go with it. Because we obviously saw in the past support for Windows and ARM from a bunch of OEMs. Some of them pulled it pretty quick, But they all kind of, most of them did one or two models just in support of it. So I think there will be support for it.

54:21I think most of them will probably ship the copilot key eventually as well, even if not all of them have yet. But yeah, I think it would be pretty broadly supported. Okay. And for Microsoft, it seems like this event in May actually turns out to be a pretty big deal. Even if it is, you know, relative sort of spec updates of things. Like, I don't know, my last question for you was basically going to be like, what do we think year two of Microsoft's all in on AI bet is going to look like? And I kind of feel like what it sounds like is we're going to know the answer to that much more clearly when we see this new hardware.

54:55That like that is going to be the way to see what all this looks like. Yeah, I think the key is the Snapdragon chip, like how that performs. If it lands well, then you can imagine there's going to be next gen AI PCs, which they will see talking about internally. I think these are considered next gen whereas perhaps the stuff that's in the market right now of NPUs is you know current gen or whatever but yeah I think year two will be I hope it's like more fleshing out windows and how it's actually using all this this NPU power and leveraging it all locally on the device because it's still that's the main feature I think there's going to also be a video streaming improvement as well and that's the other kind of main feature that there's two features that they're kind of selling for the next-gen AI PCs, marketing slogan.

55:42It's interesting to hear you talk about AI as the sort of startup inside of Microsoft. It actually makes a lot of the last 12 months make a lot of sense, both in terms of this sort of flailing around all over Windows, trying to figure out what to do, but also some of the stuff like the growth hacking inside of Edge and inside of Windows. That's the kind of stuff that big, mature companies do less of, but startups do lots of yes yeah and that culture clash i think it's like maybe why it felt so bad as a user to get some of that stuff over the last year it's like microsoft this is gross like leave me alone i'm just trying to use your web browser for god's sake but it's interesting because the windows engineers have obviously gone back to the surface side but they've kept bing edge and copilot on that microsoft ai side so it does suggest that they'll be doing all of the ai work within those those three brands so we might not get as i don't know as pushy hopefully as pushy stuff in the windows experience itself so but edge is probably still going to be messy yeah i did want to mention one thing that the interesting thing i think they're going to say and i well i kind of know from sources is you know how microsoft used to say a pc in every home was like the windows slogan right it's going to be a co-pilot for every person like that's how they're going to sell this now yeah and and they're hoping to get to like 50 of new laptops like by the end of 2026 will be you know have all these all the npu hardware on them interesting and that's very telling yeah when you get to that scale then it's quite telling of their their ambitions and where they might go with us so fair enough all right well we're we're gonna have to check back in in may i'm so skeptical of these qualcomm chips tom i cannot even tell you but boy i hope they're awesome i'm so skeptical impossible AIPC slogan, marketing nonsense.

57:31Yeah. If anyone knows what an AIPC is, please email us. Virgcast at the Virg.com. I want to know. I want to know too. All right. We got to take one more break and we'll be right back with the Virgcast Hotline. Thank you, Sam. All right. Thank you.

57:47Support for this show comes from LinkedIn. When you're a small business owner, your business is on your mind 24 seven. So when you're hiring, you need a partner that works just as hard as you do. That hiring partner is LinkedIn Jobs. When you clock out, LinkedIn clocks in. LinkedIn makes it easy to post your job for free, share it with your network, and get qualified candidates that you can manage all in one place. LinkedIn's new features can help you write job descriptions and then quickly get your job in front of the right people with deep candidate insights. Either post your job for free or pay to promote.

58:21Promoted jobs get three times more qualified applicants. At the end of the day, the most important thing to your small business is the quality of candidates. And with LinkedIn, you can feel confident that you're getting the best. Based on LinkedIn data, 72 % of SMBs using LinkedIn say that it helps them find high quality candidates. Find out why more than two and a half million small businesses use LinkedIn for hiring today. Find your next great hire on LinkedIn. Post your job for free at linkedin.com slash track. That's linkedin.com slash track to post your job for free. Terms and conditions apply.

58:58Support for the show comes from Framer. Are you still jumping between a host of different tools just to update your website? Framer unifies design, CMS, and publishing on one canvas. No handoff, no hassle. Everything you need to design and publish in one place. Framer already built the fastest way to publish beautiful, production-ready websites, and it's now redefining how we design for the web. With the recent launch of Design Pages, a free canvas-based design tool, Framer is more than a site builder. It's a true all-in-one design platform from social assets to campaign visuals to vectors and icons, all the way to a live site.

59:37Framer is where ideas go live, start to finish. Ready to design, iterate, and publish all-in-one tool? Start creating for free at framer.com slash design and use code VERGE for a free month of Framer Pro. That's framer.com slash design and use promo code VERGE. Framer.com slash design. Promo code VERGE. Rules and restrictions may apply.

1:00:08300 sensors. Over a million data points per second. How does F1 update their fans with every stat in real time? AWS is how. From fastest laps to strategy calls, AWS puts fans in the pit. It's not just racing. It's data-driven innovation at 200 miles per hour. AWS is how leading businesses power next-level innovation.

1:00:41All right, we're back. Let's get to the hotline. As always, the number is 866-VERGE-11. The email is vergecast at theverge.com. We love all of your questions. We try to answer at least one on the show every week. Thank you to everybody who has been calling in. The questions have been so weird the last few weeks. I love it. I wish we could do more of them. We're going to do more of them. We've definitely got to do a hotline episode soon because we've got a bunch of really fun ones to get into. For now, we have a question from Josh. Hey, David. This is Josh calling from northern Washington state.

1:01:11I recently watched a World War II documentary on Netflix, which is pretty meh. Now, my question is, I like World War II documentaries. So how should I rate this on Netflix? Should I give it a thumbs up, the two thumbs up, or the thumbs down? Because I want to see more World War II documentaries. So how does the algorithm take it? If I say thumbs down, will it not recommend any more documentaries for me? or if I give it two thumbs up, would it say, oh, you like mediocre documentaries? I'm going to recommend more mediocre documentaries for you. I'm really confused on how to train the algorithm when it comes to mediocre shows and all of these algorithms that, you know, really want to show me things rather than letting me choose things.

1:01:55Really appreciate your help with this. Thanks. If you've been listening to this show for a while, you know that recommendation systems are like an obsession of mine. So in order to answer this question, I went back to an interview I did last year with Pat Fleming, who is a senior director of product management at Netflix. Basically, he runs the member experience team. So everything that you do when you use the app, that is his responsibility. And one of the things that is his responsibility is all of this recommendation stuff. You can listen to a lot of our interview in an episode from last fall.

1:02:28I'll put that in the show notes. But one of the things we talked about specifically was basically how to tell Netflix what you do and don't want to watch. He gave me a bunch of information on kind of what Netflix cares about, what it's paying attention to, what you can do to signal it, all this good stuff. So the first one is, let me just play you a clip from that interview in which he described a few of the things that Netflix is looking for. The biggest one is what you watch. What you watch and don't matters most. But here is Pat describing some of what Netflix cares about. You could think of titles as being broken down into these constituent components, this metadata.

1:03:05Is it a movie or a film? What's its originating country? How long is it? What language is it available in? Genre, perhaps some tagging about attributes of the title. And those things all help to sort of round out what a title is like. And then you match that against member tastes and preferences and take that sports docs, drama, baking shows. if you watch any two of those things and lots of people watch all three of those things, then it's a pretty good chance that that third genre might be interesting to you. And so some point in time, you'll discover some genre that what wouldn't otherwise have been in your consciousness and try it out on Netflix.

1:03:38One of the other things Pat and I talked a lot about was these thumbs. I've always been fascinated by the thumbs versus stars thing. Here's actually how Pat described why Netflix goes with the thumbs thing, two thumbs up, one thumbs up or thumbs down instead of like a zero to five stars ranking. When you really love it, give it two thumbs up. When you love it, give it a thumbs up. And if you don't like it, that's okay. Let us know. Give it a thumbs down. So that's one very concrete way. I actually like that explanation a lot. And it really makes sense to me. I think about things like Yelp, where everything seems to be 3.5 stars and you don't know whether that's good or bad.

1:04:13Any restaurant on Yelp that's over four stars, I assume is great. And anything that's under like 3.8, I assume is awful. That doesn't make any sense. But that's kind of where we are. And then you get in an Uber and anything other than five stars essentially means I would like for you to be fired, which also doesn't make any sense. So we're in this odd place where different levels of stars mean different things depending on where you are, and there's no real rubric for it. So I think in this case, the thumbs up thumbs down thing is much easier to parse. But again, for our friend Josh here, not quite as easy to parse.

1:04:45One more thing about the thumbs. I think this is actually kind of relevant, which is that a thumbs down is very important. So what I would say in this case is unless you truly loathe this documentary and don't ever want to see anything like it, do not thumbs down. Because what I heard from Pat is that they'd love to have more information about why you gave something a thumbs down, but they really don't. The magic would be the longer that you spend, the better that final recommendation would be because we would then know so much more about the things you could have been interested in and might not have been interested in.

1:05:18So one thing that's a little more challenging is extracting, say, negative feedback from you. So it would be nice for us to know with a little bit more precision when you skip over something, whether you spend a little bit of time with a trailer and you say, oh, that's not for me tonight. Sometimes that's a little more challenging for us to know precisely in the moment. And that's an area where we want to spend a little more time. so basically where pat landed on that one is if you're going to do a thumbs down they are going to take that as a very strong signal that you don't want to see this anymore and it's possible that over time netflix will get better at understanding what you mean by that whether like pat said you never want to see it again or you just don't want to see it tonight but i would use that pretty sparingly and as a kind of gavel that just says get this out of my face so that's a thumbs down wouldn't do that for this one so we're down to one thumbs up or two thumbs up for this one.

1:06:06Here is how Pat described the difference, which I would argue is both helpful and unhelpful. Thumbs is a simple way for you to provide feedback on whether you like something or didn't like something. The trick on stars is sometimes you might rate something very highly on the stars because that's how something you think something ought to be perceived. So it's critically acclaimed. And so you feel like it deserves many stars as opposed to how did you actually feel about it. And that's what's more meaningful to improve your recommendations is the reflection of how you actually felt about something and whether it was good for you.

1:06:40And so thumbs has the benefit of one being simpler and to hopefully being a better reflection of what you actually love or didn't love about what you're watching on Netflix. Okay, right. Helpful and unhelpful. But I would say for you, Josh, I think we are squarely in one thumbs up territory. To me, I think of it as one thumbs up is like, I had a nice time, or I'm glad I watched this, or I'm psyched that this exists and was being recommended to me. It's kind of like a, anything that you see that you'd make the noise of just like, yeah, it was good. Like that to me is one thumbs up, right? It's not, I'm staking my reputation on the quality of this.

1:07:22This is the greatest thing that I've ever seen. I would watch this every single day if it made sense for me to do so. It's just like, yeah, I had a nice time. It's how I feel about like 95 % of movies I've ever watched on planes, right? Not the worst way to spend two hours. We'll do it again. Probably not going to write about it in my diary at the end of the day. Whereas two thumbs up is like, yes, please, Netflix, ruin my recommendations with this one. I want to see everything that you think is associated with this specific title, the people who made it the thing it's about when it came out what it looks like the colors in the background in the third scene like yes Netflix yes and I don't think we're there on this one so I think we are squarely in one thumbs up territory I'd be curious to know what everybody else thinks how you would rate this but I feel like I'm saying yeah show me more stuff kind of like this but I want you to know this was not the best version of this one thumb up before I go though I went back through the interview with Pat and tried to pull out a bunch of the things that he told me about how to signal to Netflix what you do and don't like.

1:08:25So let me just run through a few of them. One of the ones that he said is if you go to the upcoming stuff, the stuff that hasn't come out on Netflix yet, if you watch a trailer for something, but especially if you hit the remind me button to get notified when something comes out, that's a super strong signal that you like that stuff. It means you went way out of your way to actually go and get it and declare your intention to watch it to Netflix. That's a super strong signal. So is watching stuff to the end. Netflix cares a lot about where you fall out of a show. So the more of it you watch, the better.

1:08:57And I would say even in the case of like a documentary you ought to like, but isn't your favorite, if you just sort of leave it on and leave the room until it ends, that's a pretty good signal that you want to see more of this. Whereas if you shut it off after three minutes, that might be a strong signal that you don't. All of this is tricky push and pull stuff, but how far you watch is a strong signal. what you scroll past is also a strong signal. This kind of goes with the thumbs down that Pat was talking about, where if I scroll past 10 thumbnails of things, it's very hard for Netflix to figure out why I don't want to watch those things.

1:09:30But it knows that I don't want to watch those things. And it's getting stronger and stronger at recommending stuff after that, right? It can say not only here's what you watched and liked, but okay, you don't want to watch this, you don't want to watch this, you don't want to watch this. Oh, that must mean you're in the mood for a movie or a cooking show or whatever it has to be. And so that stuff is coming more slowly, but actually what you don't watch is a real signal. So is what you search for. Even when you search for stuff that's not on Netflix, if you go in and type in the title of something, it'll both give you the recommendations for things similar to that that you might watch on Netflix, which again, if you go watch, strong signal that you like that.

1:10:06But also it just shows Netflix what you care about. It's tracking that stuff. It wants to know what you want to watch, even the stuff that's not on Netflix. So that is a strong signal. One other one is just the taste profile that you have. I don't know if you do this too, but I always just leap through the thing at the very beginning of any music service or streaming service or whatever, where it's like, pick three titles that you like. But it turns out that's actually pretty important and goes a long way towards sort of kickstarting the personalization that Netflix and all these other services do.

1:10:35So if you're new to Netflix, or if you want to go back in and do it again, that is a thing you can do. And it is apparently more important than I have given it credit for. It's called your taste profile and Netflix cares a lot about it. But again, the single most important thing you can do, according to Pat, is just watch stuff. It is much more attuned to what you click on and what you actually watch than anything else. So I'm kind of in this position where every time I scroll on Netflix or click on anything, I'm petrified of what I'm teaching the algorithm. I think we all feel this way about YouTube and TikTok and all these other things where it's like, oh no, did I linger too long?

1:11:08Is the app going to think I like this and show too much of it to me? I think Netflix is working on that a little bit. But again, the fact that you went in and watched this whole documentary is a stronger signal to Netflix than how you rate it. So if you want more of these documentaries, I wouldn't worry too much about it. Just keep watching the documentaries. That's my main advice. One thumb up for Netflix. Two thumbs up for cool documentaries.

1:11:35All right. That is it for the Vergecast today. Thank you to everybody who was on the show. Thank you to everybody who called the hotline and sent us emails. And thank you as always for listening. There's lots more on everything we talked about and my old conversation with Pat in the show notes on theverge.com. Read the website. It's a good website. There is so much stuff going on this week. It's nuts. As always, if you have thoughts, questions, feelings, keyboard layouts I should know about or anything else, email us vergecast to theverge.com. Call the hotline 866-VERGE-11. Looking forward to hearing from you.

1:12:04This show is produced by Andrew Marino, Liam James, and Will Poore. The Verge Cast is a Verge production and part of the Vox Media Podcast Network. Neel, I, Alex, and I, after a series of vacations and illnesses, we'll be back on Friday, and we have a lot of news to catch up on. We'll see you then. Rock and roll.

1:12:30And Doug, here we have the Lemo Emu in its natural habitat, helping people customize their car insurance and save hundreds with Liberty Mutual. Fascinating. It's accompanied by his natural ally, Doug. Uh, Lemu? Is that guy with the binoculars watching us? Cut the camera. They see us. Only pay for what you need at libertymutual.com. Liberty, Liberty, Liberty, Liberty. Savings vary. Underwritten by Liberty Mutual Insurance Company and affiliates excludes Massachusetts.