ATP

00:00:00 ◼   ► If you wanna really screw anything up big time,

00:00:03 ◼   ► really, if you want to just really make some of the biggest

00:00:06 ◼   ► public mistakes of your life, now is the time to do it,

00:00:10 ◼   ► because at least we're not the OpenAI board.

00:00:12 ◼   ► (Geoffrey laughs)

00:00:14 ◼   ► - You're jumping ahead, my friend.

00:00:15 ◼   ► Jumping ahead. - What a mess.

00:00:17 ◼   ► Oh my God.

00:00:18 ◼   ► Wow, I mean, you talk about people having bad weeks.

00:00:24 ◼   ► Like, sometimes, oh yeah, they're having a bad week.

00:00:26 ◼   ► Boy are they having a bad week.

00:00:28 ◼   ► - At least it's relatively low stakes, I feel like.

00:00:31 ◼   ► (Geoffrey laughs)

00:00:32 ◼   ► It's fun because it's only like, whatever,

00:00:34 ◼   ► 80 billion dollars on the line, but it's, you know,

00:00:36 ◼   ► in the end, it's people with computers and stuff and things,

00:00:38 ◼   ► and it's not like life or death,

00:00:40 ◼   ► and most of the people involved,

00:00:42 ◼   ► you feel okay laughing at them a little bit

00:00:44 ◼   ► because they're such big, distant figures,

00:00:46 ◼   ► they don't seem like real people to you.

00:00:48 ◼   ► But rest assured, they are,

00:00:49 ◼   ► and that's probably why they're in this mess.

00:00:52 ◼   ► (electronic beeping)

00:00:53 ◼   ► - Final warning.

00:00:55 ◼   ► This is the moment where you, a dear, beloved listener,

00:00:59 ◼   ► who has not yet put in your order,

00:01:01 ◼   ► you sit or stand there and you think,

00:01:04 ◼   ► I've got this, no worries.

00:01:06 ◼   ► I'm gonna take care of that as soon as I get to the office,

00:01:08 ◼   ► or I'll take care of it as soon as I get home.

00:01:11 ◼   ► And then, you, dear, beloved listener, you forget.

00:01:16 ◼   ► And then what happens?

00:01:18 ◼   ► Then the rules state, you have to go tweeting,

00:01:21 ◼   ► even though I won't see it, or tweeting at me,

00:01:23 ◼   ► saying, I'm the one, I'm that person this time.

00:01:28 ◼   ► Don't be that person, don't do that to yourself.

00:01:30 ◼   ► Go to ATP.fm/store.

00:01:32 ◼   ► This is your final warning, this is all we got, this is it.

00:01:35 ◼   ► The sale ends this weekend, Sunday, November 26th.

00:01:40 ◼   ► That's it, that's all the time you get.

00:01:43 ◼   ► So, go now, don't hesitate, pull over the car,

00:01:46 ◼   ► use a signal, because you're an adult and courteous,

00:01:50 ◼   ► use a signal.

00:01:51 ◼   ► If you're walking, do what you need to do

00:01:53 ◼   ► to indicate to those behind you,

00:01:55 ◼   ► you will be pulling to the side of the sidewalk.

00:01:57 ◼   ► If you're biking, God help you,

00:01:59 ◼   ► but do what you gotta do as a bikist, as Marco would say,

00:02:02 ◼   ► to get to the side of the road

00:02:04 ◼   ► and not get run over by a car.

00:02:06 ◼   ► - That makes people so mad, by the way.

00:02:08 ◼   ► - I know, I know it does.

00:02:09 ◼   ► - They don't get that it's a joke, it makes them so mad.

00:02:12 ◼   ► - I know, we know that bikist is not a thing.

00:02:15 ◼   ► Anyways, all kidding aside, go to ATP.fm/store,

00:02:18 ◼   ► check out our merch, check out our wares,

00:02:20 ◼   ► and remember, if you are a member,

00:02:23 ◼   ► you get 15% off the ATP store

00:02:27 ◼   ► during this time-limited sale

00:02:29 ◼   ► and all other time-limited sales, now is the time.

00:02:32 ◼   ► I know we kinda glossed over it last week,

00:02:34 ◼   ► but Jon, would you mind just doing

00:02:36 ◼   ► a brief overview of everything,

00:02:37 ◼   ► just one last time to really sell it,

00:02:40 ◼   ► just nail it and send it home.

00:02:42 ◼   ► Let's get a few more sales, Jon, what do we have on offer?

00:02:45 ◼   ► - We're not gonna go through all the merch again,

00:02:46 ◼   ► they know what's there, people know.

00:02:48 ◼   ► - All right, I'll go through all the merch,

00:02:49 ◼   ► why you gotta be such a party pooper?

00:02:51 ◼   ► All right, I'll do a speed run.

00:02:52 ◼   ► We got ATP Pixels, we've got ATP Space Black,

00:02:56 ◼   ► we've got the M3 shirt, the M3 Pro shirt, the M3 Max shirt,

00:02:59 ◼   ► we've got ATP Six Colors, which is six colors of fabric,

00:03:02 ◼   ► actually more than six colors of fabric,

00:03:03 ◼   ► but the text is always white,

00:03:05 ◼   ► really should've workshop that name, but here we are.

00:03:07 ◼   ► ATP Logo shirt, the OG, we've got the ATP hoodie,

00:03:10 ◼   ► ATP Polo, don't let me down, Polo ponies, Polo people,

00:03:14 ◼   ► don't let me down, this is my jam,

00:03:16 ◼   ► get an ATP Polo, please and thank you.

00:03:18 ◼   ► We've also got the ATP mug in now white

00:03:21 ◼   ► and a bluey purpley sort of thing,

00:03:23 ◼   ► Jon will correct the actual--

00:03:24 ◼   ► - It's cobalt, please, Casey, cobalt.

00:03:26 ◼   ► - Of course, I said that, it's definitely cobalt,

00:03:28 ◼   ► definitely not a bluey purpley thing.

00:03:30 ◼   ► - It's recycled cobalt, don't worry.

00:03:31 ◼   ► - Sure, and the ATP pint glass,

00:03:33 ◼   ► so go now, ATP.fm/store, and remember, if you're a member,

00:03:38 ◼   ► go to your member page, landing, strip,

00:03:42 ◼   ► whatever you call it thing,

00:03:43 ◼   ► and go get your bespoke discount code for 15% off,

00:03:47 ◼   ► ATP.fm/store, this is it, people,

00:03:50 ◼   ► the 26th, that's all you got.

00:03:52 ◼   ► - Yeah, the only thing I'll add is,

00:03:53 ◼   ► remember that a lot of the products come in

00:03:55 ◼   ► long sleeve T-shirt versions and sweatshirt versions.

00:03:59 ◼   ► I think this is kind of the first time we've done those

00:04:01 ◼   ► in a lot of these products, so check,

00:04:02 ◼   ► even though a lot of times it will show a picture

00:04:04 ◼   ► of a T-shirt, if you watch long enough,

00:04:05 ◼   ► the picture will rotate and you'll see

00:04:06 ◼   ► some long sleeve varieties, and it's the cold winter months

00:04:10 ◼   ► in our parts anyway, so think about those long sleeve

00:04:15 ◼   ► options if you're interested, and this sale does span

00:04:19 ◼   ► Black Friday, we did this intentionally,

00:04:21 ◼   ► so it ends on the 26th, Black Friday is a couple days

00:04:23 ◼   ► before that, you're gonna be thinking about shopping

00:04:27 ◼   ► for the holidays, probably, or you'll be bombarded

00:04:30 ◼   ► by messages about shopping for the holidays,

00:04:33 ◼   ► don't forget, this is your last chance,

00:04:35 ◼   ► I know we did this sale kind of early,

00:04:36 ◼   ► but we're trying as best we can to have a chance

00:04:38 ◼   ► of getting these things to you in time for the holidays,

00:04:41 ◼   ► so, you know, nerdy gifts for yourself or for others,

00:04:44 ◼   ► don't forget, when Black Friday arrives,

00:04:46 ◼   ► that's another reminder for you to hear Casey's voice

00:04:49 ◼   ► in your head, just saying, oh wait, was I supposed

00:04:50 ◼   ► to buy some nerd thing?

00:04:51 ◼   ► Yes, ATP.fm/store.

00:04:53 ◼   ► - Indeed, there's also Ute sizes as well,

00:04:56 ◼   ► which we added a little bit late, apologies about that,

00:04:58 ◼   ► so you can check out for your kids, you can get them stuff

00:05:00 ◼   ► for friends, kids, you know, whatever you wanna do,

00:05:03 ◼   ► ATP.fm/store, I'm telling you, this is good stuff,

00:05:06 ◼   ► that you should definitely check it out,

00:05:08 ◼   ► and time's running out, my friends, time is running out.

00:05:12 ◼   ► I feel like I had something else that I wanted to add,

00:05:13 ◼   ► but I forgot what it was, so, that's all right.

00:05:15 ◼   ► We'll just make me feel, we'll make me sound smart in post,

00:05:17 ◼   ► right Marco, right, right, right, okay, good.

00:05:18 ◼   ► - Garbage in, garbage out.

00:05:20 ◼   ► - Oh, brutal, brutal, well played, but very mean.

00:05:25 ◼   ► Keyboardware, let's talk about,

00:05:28 ◼   ► God, you really derailed me with that.

00:05:29 ◼   ► Keyboardware, let's talk about what happens to keyboards

00:05:32 ◼   ► because of your damn dirty fingers.

00:05:34 ◼   ► Ryan Holmes writes, Casey, I think what you're seeing

00:05:37 ◼   ► as finger grease is likely what keyboard enthusiasts

00:05:40 ◼   ► call quote unquote shine that happens to ABS plastic.

00:05:43 ◼   ► It's relatively soft plastic that is susceptible

00:05:45 ◼   ► to finger oils and heat.

00:05:47 ◼   ► To John's point, some people's finger oils shine them

00:05:49 ◼   ► more than others.

00:05:50 ◼   ► When you use a magic eraser to remove the shine,

00:05:52 ◼   ► you are literally abrading off a layer of plastic.

00:05:55 ◼   ► Okay, so, interrupting briefly, I don't debate

00:05:57 ◼   ► that that's very possibly true, but I'm not like sawing

00:06:01 ◼   ► the keyboard with a magic eraser, like,

00:06:03 ◼   ► just a quick glance with it is more than enough,

00:06:06 ◼   ► and I only do this a handful of times a year, like I think--

00:06:08 ◼   ► - But that is true, though, like that is how the abrasive

00:06:11 ◼   ► in a magic eraser works, you are taking a layer off.

00:06:14 ◼   ► It's just a very small layer

00:06:15 ◼   ► 'cause it's a very fine abrasive.

00:06:16 ◼   ► - But you're also leaving bits of the magic eraser

00:06:18 ◼   ► on your keyboard and going down into the,

00:06:20 ◼   ► I'm still against the magic eraser.

00:06:22 ◼   ► - Yeah, it's not a great option to have to use.

00:06:25 ◼   ► Ideally, you don't need it.

00:06:27 ◼   ► 'Cause you're right, there is some risk of getting

00:06:30 ◼   ► little dots of it stuck in there and jamming up the keys.

00:06:34 ◼   ► Obviously, now that we have a little more robust keyboards

00:06:36 ◼   ► than we used to, that risk is lower, but it is still a risk.

00:06:40 ◼   ► So yeah, you are better off not doing that

00:06:41 ◼   ► if you can help it.

00:06:42 ◼   ► - That's fair.

00:06:43 ◼   ► Continuing back to what Ryan Holmes was writing,

00:06:45 ◼   ► ABS is nice for keycaps because they are easy to double shot,

00:06:48 ◼   ► meaning the letters can be embedded into the key

00:06:50 ◼   ► and won't wear off.

00:06:51 ◼   ► It's a separate layer of plastic molded inside.

00:06:53 ◼   ► This also enables any color combo.

00:06:56 ◼   ► A more durable option is PBT, polybutylene terra--

00:07:00 ◼   ► Oh my gosh, that's all it costs, that's all it wants.

00:07:02 ◼   ► Terra-fith-al-ate, which takes much higher temperatures

00:07:06 ◼   ► to deform, it does not quote unquote shine.

00:07:09 ◼   ► It often has a deeper sound and can be double shot,

00:07:10 ◼   ► but double shotting it is not common

00:07:12 ◼   ► and can involve a blend with ABS.

00:07:14 ◼   ► The more common way to print legends on PBT

00:07:17 ◼   ► is dye sublimation, which involves heat and dye.

00:07:21 ◼   ► The dye embeds into the top layer of the key

00:07:23 ◼   ► so the legends will essentially never wear off.

00:07:25 ◼   ► There is so much lingo here, my word.

00:07:28 ◼   ► I know every interest and profession

00:07:31 ◼   ► has their own vocabulary, but my gosh.

00:07:33 ◼   ► Anyways, it's difficult to get them as crisp

00:07:35 ◼   ► as double shot, however, and they will not

00:07:37 ◼   ► let LEDs glow through.

00:07:39 ◼   ► PBT is also difficult for light legends on dark caps

00:07:42 ◼   ► since the whole cap will need to be dyed

00:07:43 ◼   ► except the legend, which is masked out.

00:07:45 ◼   ► This amount of dyeing involves longer exposure

00:07:47 ◼   ► of the plastic to heat and often deforms larger keys,

00:07:49 ◼   ► most notably space bars.

00:07:50 ◼   ► Cheaper OEM keyboards use other kinds of surface printing

00:07:53 ◼   ► that wears off relatively easily.

00:07:55 ◼   ► So that was way more information than certainly I needed

00:07:57 ◼   ► or wanted to know about keyboards.

00:08:00 ◼   ► - Some background on Apple's keyboards,

00:08:01 ◼   ► so Apple's using ABS, which is the softer plastic.

00:08:04 ◼   ► Apple has used PBT in the past on older laptops

00:08:07 ◼   ► before your time, but back in the day,

00:08:10 ◼   ► the Power Brooks used to come with keyboards

00:08:12 ◼   ► that it was kind of hilarious.

00:08:13 ◼   ► They looked kind of like, I mean,

00:08:15 ◼   ► you've all seen the Apple Extended 2, right,

00:08:17 ◼   ► even if you've never actually touched one,

00:08:18 ◼   ► what those key caps look like.

00:08:19 ◼   ► Imagine those, but imagine you took them into a 3D program

00:08:23 ◼   ► and you squished them so they're not as high, right?

00:08:26 ◼   ► And that's what they made the Power Brook key caps look like.

00:08:28 ◼   ► From the top, it looked like,

00:08:29 ◼   ► "Oh, that's just a regular key."

00:08:30 ◼   ► But then you'd look from the side

00:08:31 ◼   ► and you'd see they were squished out.

00:08:32 ◼   ► Like they had the same kind of like slanted edges,

00:08:35 ◼   ► but just much, much smaller.

00:08:36 ◼   ► I'm pretty sure that back in those days,

00:08:38 ◼   ► the key caps were PBT.

00:08:40 ◼   ► Of course, they weren't backlit,

00:08:41 ◼   ► so that's the whole big thing.

00:08:43 ◼   ► I think all of Apple's keyboards now

00:08:44 ◼   ► on their laptops are backlit.

00:08:45 ◼   ► You can't backlight the PBT

00:08:47 ◼   ► 'cause it doesn't let light through it.

00:08:49 ◼   ► That's why they were talking about having ABS mixed in

00:08:51 ◼   ► because you could let the light go through the ABS part,

00:08:53 ◼   ► like just the letter that's on them or whatever.

00:08:55 ◼   ► But anyway, Apple's key caps on all their modern keyboards

00:08:58 ◼   ► all appear to be the very much softer ABS.

00:09:03 ◼   ► - Anonymous writes, "The backlit keyboard key caps

00:09:05 ◼   ► "on Apple laptops are injection molded in clear plastic.

00:09:08 ◼   ► "They are painted with white paint, then black paint.

00:09:11 ◼   ► "The glyphs are then laser etched using a method

00:09:13 ◼   ► "that removes the black paint, but not the white paint.

00:09:15 ◼   ► "Lastly, a protective clear coat is applied."

00:09:19 ◼   ► Andrew writes, "Your conversation about wear on keyboards

00:09:21 ◼   ► "reminded me of a recent article

00:09:22 ◼   ► "by a British cycling journalist."

00:09:24 ◼   ► Would that be a bikist, Marco?

00:09:25 ◼   ► Is that correct?

00:09:26 ◼   ► A bikist journalist.

00:09:27 ◼   ► - A bikalist.

00:09:28 ◼   ► - (laughs)

00:09:28 ◼   ► Oh my, we're gonna get so much hate mail.

00:09:30 ◼   ► "On the recent demise of his 10-year-old MacBook,

00:09:32 ◼   ► "which had survived an impressive amount of abuse."

00:09:35 ◼   ► You might like to see how worn his most used keys were.

00:09:38 ◼   ► We will put a link in the show notes

00:09:39 ◼   ► and maybe Marco will make this chapter art, maybe, maybe.

00:09:43 ◼   ► Let me tell you, actually, you know what?

00:09:44 ◼   ► I would not advise you to make this chapter art

00:09:46 ◼   ► because this keyboard is fricking horrifying.

00:09:50 ◼   ► Not only is this the ridiculous Tetris key return key,

00:09:53 ◼   ► which I know the British people love,

00:09:55 ◼   ► leaving that monstrosity aside,

00:09:57 ◼   ► this is an abomination.

00:09:59 ◼   ► Look, this is disgusting, this thing.

00:10:01 ◼   ► I can't handle it.

00:10:02 ◼   ► I need to scroll down because this is--

00:10:04 ◼   ► - This is one of many photos that we got.

00:10:05 ◼   ► This is the worst one, obviously.

00:10:06 ◼   ► This is one of many, many, many photos that we got,

00:10:09 ◼   ► most of which were not 10 years old, to be clear,

00:10:11 ◼   ► most of which were much younger laptops.

00:10:14 ◼   ► And if you, you know, to visualize what this picture

00:10:16 ◼   ► looked like, picture an Apple keyboard

00:10:18 ◼   ► with all of its black key caps,

00:10:20 ◼   ► and then have little circles worn through them

00:10:22 ◼   ► at various places, seeing the mechanism beneath,

00:10:26 ◼   ► because as the earlier feedback item said,

00:10:29 ◼   ► Apple's keys are made from clear plastic,

00:10:32 ◼   ► and then they have black paint and white paint over them.

00:10:34 ◼   ► And so this, imagine the keys with the black

00:10:37 ◼   ► and white paint rubbed off,

00:10:39 ◼   ► just showing you the clear plastic.

00:10:41 ◼   ► It's like these little window panes of various sizes,

00:10:43 ◼   ► and you can see where they wore through the black first,

00:10:46 ◼   ► and then the white second, right?

00:10:49 ◼   ► And you see kind of like a terrace pattern

00:10:51 ◼   ► where it's like black, then there's a white ring,

00:10:53 ◼   ► and then there's a clear spot.

00:10:54 ◼   ► And you can tell where this person hit the space bar,

00:10:56 ◼   ► like the one very specific spot all the time.

00:11:00 ◼   ► Yeah, tons of people had this, and you know, again,

00:11:03 ◼   ► the leading theory is pH levels,

00:11:07 ◼   ► either more acidic or more basic, you know,

00:11:10 ◼   ► sweat from their fingers or whatever,

00:11:12 ◼   ► essentially wearing through the paint,

00:11:13 ◼   ► because these are painted clear plastic.

00:11:17 ◼   ► And if they were a PBT, where it was the color

00:11:20 ◼   ► of the plastic instead of being painted,

00:11:22 ◼   ► this wouldn't happen,

00:11:23 ◼   ► but then you couldn't have the backlights

00:11:24 ◼   ► unless you double shot them.

00:11:26 ◼   ► But then there's the, you know,

00:11:27 ◼   ► they're saying like to put the dye on the entire key,

00:11:30 ◼   ► something along here, like the space bar might warp.

00:11:33 ◼   ► So I would say, like, as we mentioned in the last show,

00:11:35 ◼   ► keyboard durability on laptops,

00:11:39 ◼   ► aesthetic durability, not like, you know,

00:11:43 ◼   ► making the keys continue to work,

00:11:44 ◼   ► still seems like an area

00:11:45 ◼   ► where Apple could use some improvement.

00:11:48 ◼   ► I think what they have now is a reasonable-ish compromise,

00:11:50 ◼   ► 'cause they're lightweight, they're fairly sturdy,

00:11:54 ◼   ► they hold up pretty well most of the time,

00:11:56 ◼   ► but if you're one of those people who wears away key caps,

00:11:59 ◼   ► it's a bummer for you,

00:12:01 ◼   ► because you're going to have a worse experience

00:12:03 ◼   ► with this keyboard.

00:12:04 ◼   ► It'll be nice if the keys still work,

00:12:06 ◼   ► and I guess as long as you remember what those letters were

00:12:08 ◼   ► before you wore them off with your fingers, you're fine.

00:12:10 ◼   ► So functionality is job one,

00:12:12 ◼   ► but aesthetic should be somewhere in the top five,

00:12:14 ◼   ► and the current keyboards are failing that

00:12:17 ◼   ► for some subset of Apple's customers.

00:12:20 ◼   ► - This next section is John tries to make Casey feel bad

00:12:23 ◼   ► about liking 14-inch laptops, so I'm just gonna--

00:12:27 ◼   ► - It's not just John.

00:12:28 ◼   ► I mean, if only one of us would've said last episode,

00:12:32 ◼   ► you know, with this amount of power draw,

00:12:35 ◼   ► so much increase from the M1 generation,

00:12:37 ◼   ► maybe the 14-inch might be a bad idea

00:12:40 ◼   ► if you get the max chip,

00:12:42 ◼   ► because it seems like that's a lot of heat

00:12:44 ◼   ► for the 14-inch to quietly and gracefully get rid of

00:12:48 ◼   ► compared to the 16-inch.

00:12:49 ◼   ► - So John, what happened?

00:12:51 ◼   ► - I mean, to be clear, this was true

00:12:53 ◼   ► of the M2 and M1 generations as well,

00:12:55 ◼   ► which is why Marco gave that advice.

00:12:58 ◼   ► But anyway, we'll put a link in the show notes

00:13:01 ◼   ► to a YouTube video where they tested

00:13:02 ◼   ► the M3 Max 14-inch versus 16-inch.

00:13:06 ◼   ► And I think I said on the last show that like,

00:13:07 ◼   ► well, you know, maybe the cooling

00:13:09 ◼   ► should be roughly equivalent,

00:13:11 ◼   ► and the thing that is stuck on top of the M3 Max,

00:13:13 ◼   ► like the little whatever heat spreader thing

00:13:15 ◼   ► that's pulling the heat off, yeah, that part's the same.

00:13:17 ◼   ► Every other part of it is not the same though.

00:13:20 ◼   ► First of all, the heat pipe itself

00:13:21 ◼   ► that's leading away from the SOC,

00:13:23 ◼   ► that heat pipe is wider on the 16-inch.

00:13:25 ◼   ► So right away, there's a difference in the cooling system.

00:13:27 ◼   ► But the big difference is the fans in the 16-inch

00:13:30 ◼   ► are so much bigger fans in the 14-inch.

00:13:33 ◼   ► Like they're really using that extra space to good effect.

00:13:37 ◼   ► It is a dramatic difference in the size of the fans

00:13:39 ◼   ► and obviously the amount of air they can move.

00:13:41 ◼   ► Then the fans are blowing that air over,

00:13:44 ◼   ► you know, these little fin things

00:13:45 ◼   ► that are attached to the heat pipe and there's more of them.

00:13:48 ◼   ► Anyway, in the testing, two big things came out.

00:13:51 ◼   ► First, that 14-inch is going to run the fans

00:13:54 ◼   ► at much higher RPM.

00:13:55 ◼   ► So they did like a Cinebench benchmark

00:13:58 ◼   ► and the 14-inch was running the fans at 7200 RPM

00:14:02 ◼   ► and the 16-inch at 3800.

00:14:03 ◼   ► So it was like double the fan speed.

00:14:05 ◼   ► And obviously, if the fans are running faster,

00:14:07 ◼   ► they're gonna be noisier.

00:14:08 ◼   ► And since the 14-inch fans are smaller,

00:14:10 ◼   ► they're gonna be a little bit more annoying too

00:14:12 ◼   ► because they're higher pitched.

00:14:14 ◼   ► That's not great.

00:14:15 ◼   ► They stabilize fan speed.

00:14:17 ◼   ► I love when they fancy up their benchmarks here.

00:14:20 ◼   ► So the average speed in the Cinebench 2024 GPU test

00:14:25 ◼   ► was 7200 RPMs for the 14-inch, 1700 RPM for the 16-inch.

00:14:30 ◼   ► That's a big difference.

00:14:32 ◼   ► That's as in like, you can hear the fans screaming

00:14:34 ◼   ► the entire time in the 14-inch and on 16-inch,

00:14:37 ◼   ► you probably wouldn't hear anything.

00:14:39 ◼   ► And now Casey may be saying, I don't care about that.

00:14:41 ◼   ► I don't run 3D benchmarks.

00:14:43 ◼   ► I don't care about the average fan speed

00:14:46 ◼   ► over a 10-minute run of some type of thing.

00:14:49 ◼   ► Then they did an Xcode test

00:14:50 ◼   ► and that's really, really hidden Casey where it hurts

00:14:53 ◼   ► 'cause that's the thing that he does do.

00:14:55 ◼   ► And this is not about the fan speed

00:14:57 ◼   ► or about temperatures or anything like that.

00:14:59 ◼   ► This is the fact that all those things

00:15:01 ◼   ► with the fan being higher on the 14-inch

00:15:03 ◼   ► translate to the 14-inch being in more thermal distress

00:15:07 ◼   ► and thermal distress means thermal throttling.

00:15:10 ◼   ► And so this is where the rubber meets the road.

00:15:12 ◼   ► The 14-inch is a little bit slower than the 16-inch

00:15:17 ◼   ► because it thermal throttles when doing Xcode compiles.

00:15:20 ◼   ► Now this is probably a big Xcode compile.

00:15:22 ◼   ► Maybe it's bigger than any of Casey's apps.

00:15:24 ◼   ► Maybe it doesn't affect him or whatever.

00:15:25 ◼   ► And the difference isn't huge.

00:15:26 ◼   ► - Yeah, 'cause this is, it took like almost a minute and a half

00:15:28 ◼   ► to do this compile.

00:15:29 ◼   ► So like if you're doing a very long sustained CPU

00:15:33 ◼   ► or GPU drain, you're gonna hit throttling on the 14-inch.

00:15:38 ◼   ► No question.

00:15:39 ◼   ► Like that's what this is clearly showing.

00:15:40 ◼   ► Like if you are pegging the CPU or GPU for a minute or more,

00:15:45 ◼   ► you are almost certainly gonna see throttling on the 14-inch

00:15:48 ◼   ► that is not present on the 16-inch.

00:15:50 ◼   ► - Yeah, and the Xcode benchmark was 82 seconds

00:15:52 ◼   ► on the 14-inch and 72 on the 16-inch.

00:15:55 ◼   ► So not a huge difference.

00:15:55 ◼   ► Not like the fan speed difference.

00:15:57 ◼   ► I feel like the big quality of life difference

00:15:58 ◼   ► is the fan speeds.

00:15:59 ◼   ► - I mean, that's like an M2 to M3 difference

00:16:01 ◼   ► in performance though.

00:16:02 ◼   ► But again, this is a long sustained one.

00:16:05 ◼   ► Like if my build takes 11 seconds,

00:16:08 ◼   ► it's probably in that span,

00:16:10 ◼   ► I probably wouldn't see the difference

00:16:11 ◼   ► between a 14 and a 16.

00:16:13 ◼   ► But if you're doing builds that take a minute,

00:16:14 ◼   ► you probably will.

00:16:15 ◼   ► - Yeah, so I mean, as the case with all these laptops,

00:16:18 ◼   ► right, it's just a question of how much the cooling system

00:16:23 ◼   ► can fight off the inevitable heat saturation

00:16:27 ◼   ► due to long-range jobs.

00:16:28 ◼   ► It's rare that a laptop, a very high-powered,

00:16:31 ◼   ► top-of-the-line, stuff-to-the-gills laptop,

00:16:34 ◼   ► can sustain its full power for a very long period of time

00:16:36 ◼   ► because the cooling systems usually aren't adequate

00:16:39 ◼   ► to get all that heat out of there to keep up with it.

00:16:41 ◼   ► Especially if you have some contrived,

00:16:43 ◼   ► like CPU maxed out at the same time as GPU maxed out,

00:16:47 ◼   ► the same time as a neural engine maxed out,

00:16:48 ◼   ► the same time as a video encoder matched out.

00:16:50 ◼   ► Like they're not gonna be able to deal with that.

00:16:51 ◼   ► But almost no real job is like that.

00:16:53 ◼   ► That's why the GPU benchmarks are so brutal here

00:16:55 ◼   ► because this was just a benchmark of the GPU

00:16:58 ◼   ► and the fan speeds were 7,000 RPM versus 1,700.

00:17:02 ◼   ► That's not even close.

00:17:03 ◼   ► That's just, they don't,

00:17:05 ◼   ► even just stressing the GPU is enough

00:17:08 ◼   ► to really, really see a difference.

00:17:09 ◼   ► And that's relevant to people playing games.

00:17:11 ◼   ► If you're gonna play games on your laptop,

00:17:13 ◼   ► you're like, I want one of these M series laptops,

00:17:15 ◼   ► so I don't need to hear the fans screaming when I play games.

00:17:19 ◼   ► Get a 16-inch.

00:17:20 ◼   ► You have a fighting chance of making that real

00:17:22 ◼   ► because if you wanna find something

00:17:25 ◼   ► that's gonna really stress the GPU,

00:17:26 ◼   ► yeah, playing a game for a long period of time will do it.

00:17:29 ◼   ► I don't think this means that a case

00:17:31 ◼   ► you should get a 16-inch.

00:17:33 ◼   ► I think the 16-inch is too big as well.

00:17:34 ◼   ► I think the whole point of being portable

00:17:35 ◼   ► is to have something small.

00:17:37 ◼   ► But if you are concerned about sustained performance,

00:17:39 ◼   ► A, don't get a laptop, but B.

00:17:42 ◼   ► - Hey, the 16-inch sustains it just fine.

00:17:44 ◼   ► - Well, no, but the 16-inch eventually heat soaks as well,

00:17:46 ◼   ► especially if you do a CPU plus GPU combo.

00:17:48 ◼   ► These benchmarks are a little bit shorter.

00:17:51 ◼   ► Yeah, and also the other thing they were measuring,

00:17:53 ◼   ► which isn't in any of these graphs in the shots here,

00:17:55 ◼   ► is clock speed.

00:17:55 ◼   ► Like they're just seeing,

00:17:57 ◼   ► are they hitting their rated top clock speed?

00:17:59 ◼   ► And both of these laptops occasionally,

00:18:02 ◼   ► depending on the benchmark,

00:18:03 ◼   ► would never actually achieve their rated top clock speed

00:18:07 ◼   ► or would do it very briefly in the beginning of the test

00:18:09 ◼   ► and then never see it again, right?

00:18:11 ◼   ► So there's always room for improvement, right?

00:18:13 ◼   ► 16-inch versus 14,

00:18:14 ◼   ► doesn't mean the 16-inch never thermal throttles,

00:18:16 ◼   ► it just means it does it less than the 14,

00:18:18 ◼   ► which is why it's coming out ahead in these tests.

00:18:20 ◼   ► - I would say these tests prove my hunch last week

00:18:25 ◼   ► that if you're going for the max chip,

00:18:28 ◼   ► maybe don't get the 14-inch.

00:18:30 ◼   ► Like if you're getting the 14-inch

00:18:32 ◼   ► for like an all-around balance of portability and size,

00:18:35 ◼   ► maybe you're going for the M3 Pro chip,

00:18:38 ◼   ► that's a different story.

00:18:39 ◼   ► It can probably handle that a lot better.

00:18:40 ◼   ► But if you're getting the M3 Max

00:18:43 ◼   ► and you intend to push it at all,

00:18:45 ◼   ► which you probably are if you're getting the Max,

00:18:48 ◼   ► you should be aware of this.

00:18:49 ◼   ► Like this might not change your mind,

00:18:51 ◼   ► but this is something that you should really be aware of,

00:18:52 ◼   ► that the M3 Max seems to be substantially crushing

00:18:57 ◼   ► the 14-inch's thermal system more than the M1 generation did.

00:19:01 ◼   ► The M2 was probably somewhere in the middle

00:19:02 ◼   ► 'cause its power was probably, it was between the two.

00:19:06 ◼   ► But they've ramped up the peak power draw of these chips

00:19:10 ◼   ► compared to previous generations.

00:19:12 ◼   ► The M1 draws less power than the M2.

00:19:15 ◼   ► The M2 draws less power than the M3 Max in all those cases.

00:19:18 ◼   ► And so be aware of that,

00:19:20 ◼   ► that they have raised the power envelope

00:19:23 ◼   ► without increasing the thermal dissipation capacity

00:19:27 ◼   ► significantly.

00:19:28 ◼   ► So if you're coming from a 14-inch M1 Max

00:19:31 ◼   ► and you're thinking, well, it handled my M1 Max

00:19:33 ◼   ► just fine in the 14,

00:19:35 ◼   ► the same can't necessarily apply to the M3 Max.

00:19:38 ◼   ► Like you have to reevaluate

00:19:39 ◼   ► because they're not dropping replacements

00:19:42 ◼   ► in terms of power and thermals.

00:19:43 ◼   ► The M3 Max uses substantially more power

00:19:46 ◼   ► and makes substantially more heat under load

00:19:49 ◼   ► than the M1 Max did.

00:19:50 ◼   ► So that might change your calculus

00:19:53 ◼   ► of whether you want a 14 or a 16.

00:19:56 ◼   ► - I think the cooling system is upgraded a tiny bit

00:19:59 ◼   ► in the M3 generation versus the M2.

00:20:01 ◼   ► I think the heat spreader thingy on the SOC

00:20:03 ◼   ► is a little bit larger in this generation.

00:20:04 ◼   ► And I'm not sure about the fans

00:20:06 ◼   ► if they are exactly the same,

00:20:08 ◼   ► but I don't think it's entirely true

00:20:12 ◼   ► they just use exactly the same cooling system.

00:20:14 ◼   ► Otherwise it would never be able to do it

00:20:15 ◼   ► because the power difference and the 16-inch

00:20:17 ◼   ► comparing the M2 16-inch versus M3 Max 16-inch,

00:20:21 ◼   ► it's a pretty big difference.

00:20:22 ◼   ► The M3 Max is pushing like 50 watts

00:20:24 ◼   ► and the M2 Max was only ever getting like 36 or so.

00:20:27 ◼   ► So there's a lot of extra heat.

00:20:29 ◼   ► - Yeah, and the M1 generation I think was in the 20s.

00:20:31 ◼   ► It's substantially more power and heat here.

00:20:34 ◼   ► And also that also results in a larger difference

00:20:38 ◼   ► between the two models in battery life

00:20:40 ◼   ► than there used to be.

00:20:41 ◼   ► There was always a decent difference

00:20:42 ◼   ► 'cause the 16-inch battery is just much bigger.

00:20:44 ◼   ► And yeah, the screen uses more power

00:20:46 ◼   ► than the 14 'cause it's bigger,

00:20:47 ◼   ► but not that much more power.

00:20:49 ◼   ► So the 16 is still the battery monster.

00:20:51 ◼   ► But in this case, keep in mind,

00:20:54 ◼   ► again, the CPU is drawing more power

00:20:57 ◼   ► than the previous generations

00:20:58 ◼   ► at the same marketing name level.

00:21:00 ◼   ► And also, I don't know how much of a factor this is,

00:21:04 ◼   ► but if those fans are spinning faster,

00:21:07 ◼   ► fans also use power.

00:21:08 ◼   ► So the result after these tests,

00:21:11 ◼   ► the result in this video,

00:21:13 ◼   ► the 14-inch battery was substantially more drained

00:21:17 ◼   ► than the 16-inch.

00:21:17 ◼   ► It was not a small difference.

00:21:19 ◼   ► So again, just don't make assumptions

00:21:22 ◼   ► based on whatever worked for you in the M1 generation.

00:21:25 ◼   ► Don't assume the same thing will work for you

00:21:27 ◼   ► in the M3 generation

00:21:28 ◼   ► because the power levels are very different.

00:21:31 ◼   ► - I understand everything that you guys are saying,

00:21:33 ◼   ► but I don't feel like I'm bumping

00:21:37 ◼   ► into thermal throttling issues anytime

00:21:39 ◼   ► other than when I'm doing transcosos with FFmpeg,

00:21:42 ◼   ► at which case, okay, yeah, I'm reaping what I sowed

00:21:45 ◼   ► or whatever the turn of phrase is.

00:21:46 ◼   ► You know what I'm trying to say.

00:21:48 ◼   ► Okay, that's fine.

00:21:49 ◼   ► The other thing is I didn't want to go down

00:21:52 ◼   ► from 64 gigs of memory.

00:21:53 ◼   ► I probably could have.

00:21:54 ◼   ► And been fine, but I didn't want to.

00:21:57 ◼   ► And in order to do that in this generation,

00:21:59 ◼   ► you have to have an M3 max.

00:22:00 ◼   ► Like that's the end of the meeting.

00:22:01 ◼   ► So in fact, you not only need an M3 max,

00:22:03 ◼   ► you need the baller M3 max.

00:22:05 ◼   ► That's the only, I'm looking at the configurator right now.

00:22:07 ◼   ► It's the only option in the 14.

00:22:09 ◼   ► So if I want 64 gigs, well, guess what?

00:22:12 ◼   ► I'm getting an M3 max.

00:22:13 ◼   ► In fact, the Pro appears to top out at 36 gigs,

00:22:16 ◼   ► which even for me is probably fine,

00:22:19 ◼   ► but I didn't want that.

00:22:21 ◼   ► I wanted 64.

00:22:22 ◼   ► And honestly, I have zero regrets about this machine.

00:22:25 ◼   ► I've been using it at home.

00:22:26 ◼   ► I've been using it away from home.

00:22:28 ◼   ► I freaking love this computer.

00:22:29 ◼   ► And I mean, I freaking love my M1 max as well,

00:22:31 ◼   ► don't get me wrong, but this computer is great.

00:22:33 ◼   ► I haven't heard the fans except one time

00:22:35 ◼   ► when I was doing a transcode.

00:22:36 ◼   ► I have no reason to believe battery life is suffering

00:22:39 ◼   ► compared to the last one.

00:22:40 ◼   ► It's been flawless so far.

00:22:43 ◼   ► I don't argue any of what you guys are saying.

00:22:45 ◼   ► I'm not trying to say that you're wrong or lying

00:22:46 ◼   ► or anything like that.

00:22:47 ◼   ► It's just in real world use,

00:22:49 ◼   ► I am not personally noticing,

00:22:52 ◼   ► not to say it's not affecting me,

00:22:53 ◼   ► but I'm not noticing the results

00:22:55 ◼   ► of any of these problems in the 14.

00:22:57 ◼   ► And for me, and I totally understand, Marco,

00:23:00 ◼   ► why you reached a different conclusion,

00:23:01 ◼   ► but for me, I don't want a freaking aircraft carrier

00:23:04 ◼   ► in my backpack.

00:23:05 ◼   ► And I travel with this machine often enough

00:23:07 ◼   ► that that matters to me.

00:23:09 ◼   ► And so I made the choice I made,

00:23:11 ◼   ► and I think it's the right one for me.

00:23:12 ◼   ► You made the choice you made.

00:23:13 ◼   ► I think it's the right one for you.

00:23:14 ◼   ► And everyone's happy.

00:23:16 ◼   ► Why won't you just let me be happy?

00:23:17 ◼   ► - No, and look, I think it's important to point out

00:23:21 ◼   ► if the 14 inch still worked for you, that's great,

00:23:24 ◼   ► but people who are buying it should know this

00:23:26 ◼   ► ahead of time, that it's going to have these trade-offs,

00:23:29 ◼   ► that the M1 generation of it had fewer trade-offs

00:23:32 ◼   ► compared to the M3 generation.

00:23:34 ◼   ► And if what you are seeking

00:23:36 ◼   ► is the highest performing laptop,

00:23:40 ◼   ► in the M1 generation, you could say,

00:23:41 ◼   ► well, these are basically the same.

00:23:43 ◼   ► Now you can't really say that.

00:23:45 ◼   ► Now the 16 inches is noticeably higher performing

00:23:48 ◼   ► with a lot of workloads.

00:23:49 ◼   ► So it is worth knowing.

00:23:50 ◼   ► Also, again, if stuff like fan noise

00:23:52 ◼   ► is really critical to you, then again,

00:23:55 ◼   ► that's something that's worth knowing ahead of time.

00:23:57 ◼   ► So that's why.

00:23:58 ◼   ► And for many people, they're gonna still get

00:24:00 ◼   ► the 14 inch like you and be totally happy with,

00:24:02 ◼   ► and that's great.

00:24:03 ◼   ► But again, you gotta know going into it,

00:24:06 ◼   ► the assumptions that you might have made

00:24:07 ◼   ► from previous generations of the M series laptops

00:24:10 ◼   ► are a little bit different now.

00:24:12 ◼   ► - I do wonder if there's not a low tech solution to this,

00:24:15 ◼   ► with the memory target, a little RGB fan cooler thing

00:24:18 ◼   ► on the back of your phone to get sustained frame rates.

00:24:21 ◼   ► - Just get some giant plate.

00:24:22 ◼   ► - Yeah, like I know they make laptop cooler type things,

00:24:25 ◼   ► but I do wonder.

00:24:26 ◼   ► - They're really quiet and elegant.

00:24:28 ◼   ► - Yeah, right.

00:24:29 ◼   ► But I do wonder if you say you get a 14 inch

00:24:31 ◼   ► and you wanna use it in desktop mode

00:24:32 ◼   ► and you don't want it to thermal throttle,

00:24:33 ◼   ► just how much extra cooling would it take?

00:24:35 ◼   ► Probably not that much.

00:24:36 ◼   ► Like, and probably could be applied externally.

00:24:38 ◼   ► I don't, you know, again, the caveats about condensation

00:24:40 ◼   ► and don't destroy your computer and blah, blah, blah,

00:24:42 ◼   ► whatever, but like based on how well it works

00:24:44 ◼   ► on the little phone thing to sort of sustain frame rates

00:24:47 ◼   ► when the phone could not do it without the cooler,

00:24:49 ◼   ► I do wonder if just a little bit of extra help

00:24:51 ◼   ► on a 14 inch would help it not thermal throttle at all,

00:24:54 ◼   ► because it's, you know, it's,

00:24:55 ◼   ► the fans are almost doing the job

00:24:57 ◼   ► 'cause yeah, the fan speed difference is huge.

00:24:59 ◼   ► Those fans are working overtime, right?

00:25:00 ◼   ► But the benchmark difference is not that big.

00:25:03 ◼   ► So it's like the 14 inches cooling system is overmatched,

00:25:07 ◼   ► but just by a little bit.

00:25:08 ◼   ► So I wonder if you just gave it a little bit of help,

00:25:10 ◼   ► which is unlike lots of other computer things.

00:25:11 ◼   ► Like, oh, I didn't get enough cores.

00:25:13 ◼   ► Well, there's not really anything you can do

00:25:14 ◼   ► to add more cores to your computer,

00:25:15 ◼   ► but if thermal throttling is your problem,

00:25:18 ◼   ► pointing a fan at the back of your laptop

00:25:20 ◼   ► when it's like sitting upright, you know,

00:25:22 ◼   ► like in a bookshelf type thing might be enough.

00:25:24 ◼   ► - One potential limitation there is,

00:25:27 ◼   ► remember we learned this from people who wrote in

00:25:29 ◼   ► like a couple of years ago, this came up

00:25:32 ◼   ► where we were talking about like the MacBook Air

00:25:33 ◼   ► thermal design without any fan,

00:25:35 ◼   ► that like there are certain standards

00:25:37 ◼   ► that different, you know, governments and things set

00:25:40 ◼   ► about how hot the outside of a laptop can get

00:25:44 ◼   ► for safety reasons.

00:25:45 ◼   ► And so the way these are designed,

00:25:48 ◼   ► usually the processor is not like heat pipe

00:25:53 ◼   ► directly to the exterior case.

00:25:55 ◼   ► Like there's usually some kind of air gap there

00:25:57 ◼   ► so that the processor is intentionally not

00:26:01 ◼   ► transferring its heat much to the metal case.

00:26:05 ◼   ► And because of that, it actually makes it difficult

00:26:08 ◼   ► to make external coolers that are very effective

00:26:11 ◼   ► because you're not getting all that heat away

00:26:14 ◼   ► from the chip through the case very much.

00:26:16 ◼   ► It's mostly dumping it into the air.

00:26:18 ◼   ► So it's a little bit harder.

00:26:20 ◼   ► So you couldn't, for instance, just stick like

00:26:22 ◼   ► a Peltier Element cooler on the back of a laptop

00:26:24 ◼   ► and have it be--

00:26:25 ◼   ► - Oh, sure you could.

00:26:26 ◼   ► You just need a temperature differential.

00:26:29 ◼   ► And one exists.

00:26:29 ◼   ► I mean, feel the bottom of a laptop

00:26:31 ◼   ► when it's hard working.

00:26:32 ◼   ► I know what you're saying, like not all the heat

00:26:33 ◼   ► is being dumped out there, but there's still

00:26:34 ◼   ► a temperature differential.

00:26:35 ◼   ► Like that's a test that someone should try.

00:26:37 ◼   ► Like just consider that.

00:26:38 ◼   ► I guess the reverse of it is like,

00:26:41 ◼   ► if you're doing some sustained workload with your laptop,

00:26:43 ◼   ► don't throw it on your bed on top of your comforter.

00:26:45 ◼   ► Like don't, you know what I mean?

00:26:47 ◼   ► Like, you see how kids cheap laptops,

00:26:49 ◼   ► they have no idea that these are like

00:26:50 ◼   ► living, breathing things, right?

00:26:51 ◼   ► It's a little empathy for the machine here.

00:26:53 ◼   ► Don't take your laptop and just put it on a big,

00:26:55 ◼   ► soft pillow surrounded by blankets.

00:26:57 ◼   ► Like it needs, you need to get it with the lid closed,

00:27:00 ◼   ► of course, right?

00:27:01 ◼   ► - Naturally.

00:27:02 ◼   ► - Give it a fighting chance.

00:27:04 ◼   ► - Yeah, and just as a final note,

00:27:06 ◼   ► 'cause I'd like to move on,

00:27:07 ◼   ► I feel like there's a lot of FUD going around right now.

00:27:11 ◼   ► These, I really don't subscribe,

00:27:15 ◼   ► I don't have any anecdotal data

00:27:17 ◼   ► that indicates that this problem

00:27:18 ◼   ► is any worse with this generation

00:27:20 ◼   ► than it was with my two year old laptop.

00:27:21 ◼   ► Like yes, again, I'm not trying to argue

00:27:23 ◼   ► that the facts that you're presenting to me are wrong.

00:27:26 ◼   ► All I'm trying to say is that in day to day use,

00:27:29 ◼   ► I straight up do not hear the fans.

00:27:31 ◼   ► So yes, in the occasion that I can hear the fans,

00:27:34 ◼   ► yes, they probably are faster,

00:27:36 ◼   ► they probably are more annoying, they're probably louder.

00:27:39 ◼   ► I don't debate that,

00:27:40 ◼   ► but I feel like we're making a problem out of nothing here.

00:27:43 ◼   ► This machine is phenomenal.

00:27:45 ◼   ► If you want a desktop replacement

00:27:48 ◼   ► that you occasionally move, 100% get the 16, no argument.

00:27:51 ◼   ► But if you want a computer that you move occasionally,

00:27:53 ◼   ► and you favor that, you know,

00:27:56 ◼   ► the ease of moving it over anything else,

00:27:58 ◼   ► which is where I am, get the 14.

00:28:00 ◼   ► Don't let these two knuckleheads

00:28:02 ◼   ► try to convince you otherwise.

00:28:03 ◼   ► It's a perfectly good computer.

00:28:05 ◼   ► There's no need to, there's no need to--

00:28:06 ◼   ► - I'm not trying to convince, I also agree.

00:28:07 ◼   ► I think the 16 is just too big for a laptop,

00:28:09 ◼   ► and really you should just get a desktop anyway.

00:28:11 ◼   ► - Oh my God, it's so good, it's so good.

00:28:13 ◼   ► The 16 is so good.

00:28:15 ◼   ► There's basically no trade off, except it is large.

00:28:20 ◼   ► But again, when you look at, I mean, we are so spoiled now.

00:28:24 ◼   ► The 16 inch MacBook Pro is lighter

00:28:28 ◼   ► than most 15 inches of most of our computing using careers.

00:28:34 ◼   ► And it is 16 only because they made the bezel smaller.

00:28:38 ◼   ► So the footprint is roughly 15 inch class

00:28:41 ◼   ► that we've had forever.

00:28:42 ◼   ► The weight is similar or lighter

00:28:45 ◼   ► than most 15 inch class laptops we've ever used.

00:28:48 ◼   ► It is glorious.

00:28:49 ◼   ► I'm not gonna argue that the 14 inch

00:28:51 ◼   ► isn't noticeably smaller, and lighter it is.

00:28:54 ◼   ► It is noticeably smaller and lighter.

00:28:56 ◼   ► I would maybe argue that if that's really your top priority,

00:29:01 ◼   ► maybe consider the Air.

00:29:03 ◼   ► But the 16 inch, again, it's all relative.

00:29:07 ◼   ► Yes, it is larger and heavier than yours,

00:29:10 ◼   ► but compared to anything ever in the history of computing,

00:29:14 ◼   ► these are all very thin and light, and it's fine.

00:29:17 ◼   ► And even like, again, when you're looking at,

00:29:19 ◼   ► suppose you're putting it in a backpack

00:29:21 ◼   ► and bringing it to wherever.

00:29:23 ◼   ► The weight of your backpack empty is probably a few pounds.

00:29:28 ◼   ► Like, it depends on what kind of bag you have.

00:29:29 ◼   ► The weight of your total carry

00:29:32 ◼   ► of what most people bring around in their backpacks every day

00:29:35 ◼   ► may be 10 or 15 pounds.

00:29:37 ◼   ► And so the difference in laptop weight

00:29:39 ◼   ► of like a one pound difference between two things--

00:29:42 ◼   ► - That's fair.

00:29:43 ◼   ► - Is not really that significant.

00:29:44 ◼   ► It feels significant when it's in your hand,

00:29:47 ◼   ► but like however you are carrying it,

00:29:49 ◼   ► generally speaking for most people,

00:29:51 ◼   ► the difference between a 16 inch MacBook Pro

00:29:53 ◼   ► and a 14 inch MacBook Pro, being like a pound or whatever,

00:29:56 ◼   ► is not a massive difference

00:29:58 ◼   ► in whatever they're carrying it in total.

00:30:01 ◼   ► So again, if it matters to you

00:30:03 ◼   ► and if you really still wanna minimize it, great.

00:30:05 ◼   ► I get that, please do.

00:30:07 ◼   ► But these are trade-offs that you should be aware of.

00:30:10 ◼   ► Despite what you said, Casey,

00:30:12 ◼   ► they weren't as severe of trade-offs in the M1 generation

00:30:15 ◼   ► because the total power usage of the M1 chips was way lower.

00:30:20 ◼   ► Like, rather the peak power usage, like under load.

00:30:23 ◼   ► The peak power usage was way lower in the M1 generation

00:30:25 ◼   ► than it is in the M3 generation.

00:30:26 ◼   ► So that is a difference that is worth noting.

00:30:29 ◼   ► And before, again, with the M1 generation,

00:30:34 ◼   ► you could say these computers

00:30:35 ◼   ► are basically the exact same computer,

00:30:37 ◼   ► just bigger and smaller.

00:30:38 ◼   ► And now that's a little harder to say.

00:30:40 ◼   ► There's a much bigger asterisk on that now.

00:30:42 ◼   ► - So let's talk about chip packaging.

00:30:44 ◼   ► Johnny Srouji had an interview in July

00:30:47 ◼   ► with his alma mater who is a college that's,

00:30:50 ◼   ► the name is escaping me, it was somewhere in Israel.

00:30:53 ◼   ► And he went back there and did an interview

00:30:55 ◼   ► with somebody there and talked a bit about chip packaging.

00:30:59 ◼   ► John, do you wanna either intro this or read it to me

00:31:02 ◼   ► or tell me how you wanna handle this?

00:31:04 ◼   ► - Yeah, so this is the interview

00:31:05 ◼   ► that I referenced on a past show

00:31:07 ◼   ► where I figured we're talking about,

00:31:09 ◼   ► we're talking about three-nanometer processes

00:31:10 ◼   ► or stuff like that.

00:31:11 ◼   ► But anyway, I've said that Apple executives

00:31:13 ◼   ► won't reveal anything about future products, obviously,

00:31:15 ◼   ► but anytime one of them talks,

00:31:16 ◼   ► they'll usually say something that lets you know

00:31:19 ◼   ► some vague future direction.

00:31:21 ◼   ► And in the interview, Johnny Srouji said

00:31:24 ◼   ► that he thought that packaging was a potential area

00:31:27 ◼   ► of innovation, and I wanted to follow up on that today

00:31:30 ◼   ► to first provide the excerpt from the interview

00:31:34 ◼   ► because the whole blah, blah, blah is interesting

00:31:36 ◼   ► is I think that was like Tim Cook saying

00:31:37 ◼   ► the wrist is interesting or some BS like that.

00:31:39 ◼   ► And whatever the quote from Jobs

00:31:42 ◼   ► when he was talking about cell phones was similar,

00:31:44 ◼   ► that's how we all knew Apple was making a phone.

00:31:46 ◼   ► So I just wanted to get the actual words

00:31:48 ◼   ► that he said about packaging

00:31:50 ◼   ► because there was also recently an interview

00:31:54 ◼   ► of Intel CEO, Pat Gelsinger

00:31:57 ◼   ► with Ben Thompson of Stratechery,

00:32:00 ◼   ► or Strateecery, however you wanna pronounce it,

00:32:02 ◼   ► where the Intel CEO talked about, you guessed it,

00:32:07 ◼   ► packaging and how important it is to Intel

00:32:10 ◼   ► and for the future.

00:32:11 ◼   ► And so I feel like this is a confluence of events

00:32:13 ◼   ► that's worth digging into deeper,

00:32:14 ◼   ► even if this, as we said in the past episode,

00:32:17 ◼   ► of this generation, it just looks like the M3

00:32:18 ◼   ► is gonna be, M3 Ultra is just gonna be two M3 Maxes

00:32:22 ◼   ► stuck together just like it was before,

00:32:23 ◼   ► which is fine and great and it'll be awesome,

00:32:25 ◼   ► but keep your eye on the future for packaging stuff.

00:32:27 ◼   ► So yeah, Casey, you just wanna go through

00:32:29 ◼   ► the interview segment thing.

00:32:30 ◼   ► - Sure, so how about I'll play the role of the interviewer

00:32:33 ◼   ► and you can play the role of Johnny Cerusci.

00:32:35 ◼   ► - Sure, give me the longer one.

00:32:36 ◼   ► - Yeah, well, all right, fine, you know what?

00:32:37 ◼   ► I'll take one. - I don't know if we can play.

00:32:38 ◼   ► We can play acted.

00:32:39 ◼   ► If we get Marco to be mixed in, he could be the announcer.

00:32:42 ◼   ► - Oh gosh, all right, so you be the interviewer,

00:32:44 ◼   ► I will be Johnny.

00:32:45 ◼   ► Let's do it that way. - Well, you don't need

00:32:46 ◼   ► two people to read them.

00:32:47 ◼   ► We don't need to, you can just be one person.

00:32:48 ◼   ► - Play with me in the space, John, come on.

00:32:50 ◼   ► Are you really just gonna leave me hanging like this?

00:32:53 ◼   ► - Jesus, wow. - I forgot, is it my line?

00:32:55 ◼   ► - Yes, it's you. - Line?

00:32:56 ◼   ► - Line. - We're actually doing this?

00:32:59 ◼   ► - Yes, yes, you are the interviewer, John.

00:33:01 ◼   ► - All right, this is the interviewer.

00:33:03 ◼   ► What are the next challenges in the design

00:33:05 ◼   ► and architecture of processors that Apple should tackle

00:33:07 ◼   ► to get to the real next generation of processing?

00:33:09 ◼   ► This is where the interviewer starts asking questions

00:33:11 ◼   ► like they shouldn't ask or they know

00:33:12 ◼   ► that I'm gonna get the answer.

00:33:13 ◼   ► What's the next challenge?

00:33:16 ◼   ► What should they tackle to get to the real next generation?

00:33:18 ◼   ► I'm not asking a question about the future,

00:33:20 ◼   ► but I totally am.

00:33:21 ◼   ► Anyway, here's what Johnny Srdi said.

00:33:23 ◼   ► - It is getting more and more challenging.

00:33:24 ◼   ► Those of you who follow CMOS technology,

00:33:26 ◼   ► whether it's five nanometer or beyond,

00:33:28 ◼   ► that's getting harder and harder, which I think is great,

00:33:31 ◼   ► by the way, because if it gets harder,

00:33:33 ◼   ► that means Apple will try better and better, so that's good.

00:33:37 ◼   ► I'm not going to describe our future roadmap,

00:33:40 ◼   ► but there are many challenges.

00:33:41 ◼   ► For example, when you take CMOS technology,

00:33:43 ◼   ► I think one of the things that is going to be important

00:33:44 ◼   ► is packaging without getting into details.

00:33:47 ◼   ► So the way you package the chip is going to be important,

00:33:50 ◼   ► or maybe you architect the chip in a different way.

00:33:52 ◼   ► - And I know you think like, that's saying nothing.

00:33:55 ◼   ► What are you talking about?

00:33:55 ◼   ► This was the quote?

00:33:56 ◼   ► Yeah, they don't say a lot,

00:33:58 ◼   ► but the fact that they say anything about anything,

00:34:01 ◼   ► oh, packaging, just that word, packaging,

00:34:04 ◼   ► what does that mean?

00:34:05 ◼   ► I mean, again, Apple already does interesting things

00:34:07 ◼   ► with packaging.

00:34:08 ◼   ► The silicon interposer is an interesting thing

00:34:09 ◼   ► done with packaging, and we've talked a lot

00:34:11 ◼   ► about packaging technology in the past,

00:34:12 ◼   ► but this is the tiny tidbit they left.

00:34:15 ◼   ► Maybe you architect your chip a different way.

00:34:17 ◼   ► Packaging is interesting.

00:34:19 ◼   ► I do like these typical Apple bravado,

00:34:23 ◼   ► where they said, oh, the world of silicon,

00:34:27 ◼   ► it's getting more and more difficult,

00:34:28 ◼   ► and that's great for Apple,

00:34:29 ◼   ► because when things get hard, we will excel.

00:34:32 ◼   ► That'll make us try harder, 'cause we're the best,

00:34:34 ◼   ► and basically bring it on, because the harder it gets,

00:34:36 ◼   ► the more you'll see that Apple is the best,

00:34:38 ◼   ► which is great for him to say.

00:34:40 ◼   ► We'll see if it actually turns out to be true

00:34:42 ◼   ► how the cell modem's going, Johnny.

00:34:44 ◼   ► I thought that was some top-tier Apple executive

00:34:51 ◼   ► bragging in this interview, but anyway,

00:34:52 ◼   ► that's what he said about packaging.

00:34:53 ◼   ► He just dropped the word packaging,

00:34:55 ◼   ► so now here is the Intel CEO, Pat Gelsinger,

00:34:58 ◼   ► also talking about packaging.

00:34:59 ◼   ► - So I will be playing the role of Pat Gelsinger

00:35:01 ◼   ► in this one, since that's the longer one.

00:35:04 ◼   ► Gelsinger says, this idea of chiplets, I think,

00:35:07 ◼   ► is the new way that all chips get designed,

00:35:09 ◼   ► the idea of advanced packaging, multiple chips

00:35:11 ◼   ► into the advanced package, and whether that's an MCP,

00:35:14 ◼   ► or multi-chip packaging, whether that's a two-and-a-half

00:35:16 ◼   ► or 3D construction, I do think that becomes the standard.

00:35:19 ◼   ► - And this is Ben Thompson.

00:35:20 ◼   ► The interviewer was saying,

00:35:21 ◼   ► why is advanced packaging the future?

00:35:23 ◼   ► You can see, you can ask Intel questions about the future.

00:35:26 ◼   ► I know this has been a big focus for Intel.

00:35:28 ◼   ► It's something that you wanna talk about,

00:35:29 ◼   ► and from everything I know,

00:35:30 ◼   ► your technology is leading the way.

00:35:31 ◼   ► Why is that so important,

00:35:32 ◼   ► in addition to the traditional Moore's Law?

00:35:34 ◼   ► Why do we need to start stacking these chiplets?

00:35:37 ◼   ► Give me the top reasons.

00:35:38 ◼   ► So this, Ben is asking a good question here,

00:35:40 ◼   ► which is like, of all the things, the problems Intel has,

00:35:43 ◼   ► and you can read the full interview,

00:35:44 ◼   ► he's interviewed the new Intel CEO a few times,

00:35:48 ◼   ► Intel's got some problems.

00:35:49 ◼   ► They fell behind, they're getting beat

00:35:51 ◼   ► in the fab business by TSMC.

00:35:52 ◼   ► Why in the world is Intel carrying anything about packaging?

00:35:55 ◼   ► And more importantly, why is Intel putting tons of money

00:35:58 ◼   ► into innovations in packaging?

00:36:00 ◼   ► And so he's asking like, what's the deal?

00:36:02 ◼   ► Give me the top reasons for the whole focus on packaging.

00:36:06 ◼   ► - One is, now you're able to take

00:36:07 ◼   ► the performance-sensitive transistors,

00:36:09 ◼   ► and move them to the leading edge node,

00:36:11 ◼   ► but leverage some other technologies for other things.

00:36:14 ◼   ► So you get to mix and match technologies more efficiently,

00:36:16 ◼   ► effectively this way.

00:36:18 ◼   ► Jon, would you like to jump in,

00:36:19 ◼   ► and explain what I, Pat Gelsinger, just said?

00:36:21 ◼   ► - Yeah, so this is part of Intel's problem,

00:36:23 ◼   ► is they, you know, their fabs fell behind.

00:36:27 ◼   ► They couldn't get off of whatever,

00:36:28 ◼   ► or were they stuck on 14 nanometer for ages, right?

00:36:31 ◼   ► And TSMC went ahead of them,

00:36:33 ◼   ► and has been ahead since then.

00:36:35 ◼   ► And so they're like, look, we know at Intel

00:36:38 ◼   ► we have some fab problems.

00:36:39 ◼   ► In fact, Intel is paying TSMC to fab a bunch of its stuff.

00:36:43 ◼   ► They're saying, one of the things that we can do

00:36:45 ◼   ► by breaking the chips up into smaller individual pieces,

00:36:48 ◼   ► is get the most important parts to be fabbed by TSMC,

00:36:53 ◼   ► not by us, on like the good process node,

00:36:57 ◼   ► and then use other lesser fabs, like Intel's fabs,

00:37:01 ◼   ► to fab the parts of the chip where it doesn't matter as much.

00:37:04 ◼   ► So if you're Intel, you don't wanna pay TSMC to,

00:37:07 ◼   ► and Intel, you can read all about it in Ben's coverage,

00:37:09 ◼   ► but Intel is trying to split where they're like,

00:37:11 ◼   ► we're gonna have a fabbing part of the company,

00:37:13 ◼   ► and then we're gonna have the chip design part

00:37:14 ◼   ► of the company, and we're gonna pretend

00:37:15 ◼   ► like they're separate.

00:37:16 ◼   ► So the fabbing part is way behind and needs to catch up,

00:37:19 ◼   ► but the part that owns like x86 and makes the chips,

00:37:21 ◼   ► we think that's good.

00:37:22 ◼   ► So, you know, they're using TSMC,

00:37:24 ◼   ► but they don't wanna have US TSMC for everything.

00:37:26 ◼   ► That would get very expensive.

00:37:27 ◼   ► Intel's big advantage is they fab their own chips,

00:37:30 ◼   ► and they design their own chips,

00:37:31 ◼   ► their vertical integration, everything.

00:37:33 ◼   ► So like, this is why, one reason why packaging is important,

00:37:37 ◼   ► Intel can pay the minimum amount required

00:37:39 ◼   ► to the company with the good fabs

00:37:41 ◼   ► to make the part of the chip where it's most important

00:37:44 ◼   ► for it to be on the top process level,

00:37:47 ◼   ► and then they can use cheaper stuff for the other parts.

00:37:50 ◼   ► - Second, we can actually compose the chiplets

00:37:52 ◼   ► to more appropriate die sizes

00:37:54 ◼   ► to maximize defect density as well.

00:37:57 ◼   ► If you have a monster server die,

00:37:58 ◼   ► you're going to be dictated to be N minus two, N minus three,

00:38:02 ◼   ► just because of the monster die size.

00:38:04 ◼   ► I get to carve up that server chip.

00:38:05 ◼   ► I get to move the advanced nodes per computing more rapidly

00:38:09 ◼   ► and not be subject to some of the issues,

00:38:11 ◼   ► defect density early in the life

00:38:12 ◼   ► of a new process technology.

00:38:15 ◼   ► - So this is about Intel, setting aside SOCs,

00:38:18 ◼   ► where it's a whole bunch of stuff on a big giant,

00:38:20 ◼   ► it's a system on a chip.

00:38:21 ◼   ► They're giant server chips,

00:38:23 ◼   ► like the huge, like, 56-core Xeon things,

00:38:25 ◼   ► they're just a single, gigantic chip,

00:38:28 ◼   ► and you have to fab that in a single,

00:38:30 ◼   ► it's fabbed in a single die.

00:38:31 ◼   ► And if some part of that screws up,

00:38:33 ◼   ► like, the more area you start covering on the silicon wafer,

00:38:35 ◼   ► the more stuff you put there, the more chances

00:38:37 ◼   ► there's going to be some error,

00:38:38 ◼   ► and you might have to throw out the whole chip.

00:38:40 ◼   ► And so this is another advantage to packaging,

00:38:43 ◼   ► is like, if you just make it out of smaller parts,

00:38:46 ◼   ► A, if it's bad, you're only throwing away

00:38:48 ◼   ► like a smaller area of your wafer,

00:38:50 ◼   ► you don't have to throw away this huge square,

00:38:52 ◼   ► you throw out the little square,

00:38:52 ◼   ► and B, your defect density can be lower.

00:38:55 ◼   ► He says maximizing defect density,

00:38:56 ◼   ► but whatever, he meant minimizing.

00:38:58 ◼   ► That you'll get more of those littler things that you fab

00:39:02 ◼   ► will come off the line and be completely error-free,

00:39:04 ◼   ► whereas trying to just get a single error-free

00:39:06 ◼   ► humongous Xeon is difficult.

00:39:09 ◼   ► SRAMs in particular, SRAM scaling will become

00:39:11 ◼   ► a bigger and bigger issue going forward.

00:39:13 ◼   ► So I actually don't get benefit by moving a lot of my cache

00:39:16 ◼   ► to the next generation node,

00:39:18 ◼   ► like I do for logic, power, and performance.

00:39:20 ◼   ► I actually want to have a 3D construct

00:39:21 ◼   ► where I have lots of cache and a base die,

00:39:24 ◼   ► and put the advanced computing on top of it

00:39:25 ◼   ► into a 3D sandwich.

00:39:27 ◼   ► And now you get the best of a cache architecture

00:39:29 ◼   ► and the best of the next generation of Moore's law.

00:39:31 ◼   ► So it actually creates a much more effective

00:39:33 ◼   ► architectural model in the future.

00:39:35 ◼   ► Additionally, generally, you're struggling

00:39:36 ◼   ► with the power, performance, and speed of light

00:39:38 ◼   ► between chips.

00:39:40 ◼   ► - Yes, this whole technique,

00:39:41 ◼   ► this is a particular technology that Intel has advanced with

00:39:43 ◼   ► which lets them take power and feed it

00:39:46 ◼   ► through the layers of the sandwich

00:39:48 ◼   ► in a more efficient way to get to precisely

00:39:50 ◼   ► where they want it.

00:39:51 ◼   ► I think that's what they're referring to here.

00:39:52 ◼   ► The sandwiching isn't just because of that,

00:39:54 ◼   ► but Intel has an innovative way to send power

00:39:57 ◼   ► up to the bottom of the chip

00:39:58 ◼   ► instead of sending it through the top,

00:39:59 ◼   ► because when you send it through the top,

00:40:00 ◼   ► you gotta weave it around a bunch of obstacles

00:40:02 ◼   ► where you send it through the bottom.

00:40:03 ◼   ► There's nothing in the way

00:40:04 ◼   ► because you make these layered sandwiches or whatever.

00:40:06 ◼   ► That's what that's talking about.

00:40:07 ◼   ► And SRAM is like, look,

00:40:08 ◼   ► the SRAM is what we were talking about before

00:40:10 ◼   ► with the GPU RAM that was being shared better

00:40:14 ◼   ► with dynamic caching.

00:40:15 ◼   ► That's it's faster than the RAM that's out under RAM chips,

00:40:20 ◼   ► like your main memory,

00:40:21 ◼   ► but it's not the same as registers.

00:40:22 ◼   ► It's kind of in between there.

00:40:23 ◼   ► It's like very expensive RAM

00:40:25 ◼   ► or each a little bit of RAM

00:40:27 ◼   ► takes a large number of transistors,

00:40:28 ◼   ► but it's much faster than regular RAM and it's in the CPU.

00:40:32 ◼   ► And I say, I'm like, you don't get any benefit

00:40:33 ◼   ► of doing that on three nanometers.

00:40:34 ◼   ► So not only do you not want to waste money doing it

00:40:36 ◼   ► because it would be more expensive,

00:40:37 ◼   ► you don't get benefit from doing it anyway.

00:40:39 ◼   ► So keep that on a different lower process

00:40:41 ◼   ► and then sandwich it all together.

00:40:43 ◼   ► - Any other questions for me, Ben?

00:40:45 ◼   ► - Yeah, I'm Ben again.

00:40:45 ◼   ► All right, so how do you solve that with chiplets

00:40:48 ◼   ► when they're no longer on the same die?

00:40:50 ◼   ► - In the chiplet construct,

00:40:51 ◼   ► we're going to be able to put tens of thousands

00:40:53 ◼   ► of bond connections between different chiplets

00:40:55 ◼   ► inside of an advanced package.

00:40:57 ◼   ► So you're going to be able to have very high bandwidth,

00:40:58 ◼   ► low latency, low power consumption interfaces

00:41:00 ◼   ► between chiplets.

00:41:01 ◼   ► It also becomes very economical for design cycles.

00:41:04 ◼   ► Hey, I can design a chiplet with this IO

00:41:06 ◼   ► and use it for multiple generations.

00:41:08 ◼   ► - And you can see how all the stuff that I just described

00:41:11 ◼   ► would be appealing to Apple.

00:41:13 ◼   ► Apple's already making SOCs,

00:41:15 ◼   ► which are essentially single dies system on the chips

00:41:19 ◼   ► that have a bunch of stuff in them.

00:41:20 ◼   ► And they're getting pretty big.

00:41:21 ◼   ► That's part of the reason the ultras have been two maxes

00:41:25 ◼   ► weaved together with the Silicon Interposer,

00:41:27 ◼   ► rather than just being a single large chip

00:41:29 ◼   ► the size of an ultra,

00:41:31 ◼   ► because that would be even more expensive to manufacture.

00:41:33 ◼   ► It's also probably part of the reason

00:41:35 ◼   ► we didn't have the quad, right?

00:41:36 ◼   ► So Apple would love the idea

00:41:38 ◼   ► that you could take subunits of this,

00:41:41 ◼   ► like the thing that does IO or whatever,

00:41:43 ◼   ► and reuse that between generations,

00:41:46 ◼   ► because between generations, like, yo,

00:41:48 ◼   ► it's what over run now,

00:41:49 ◼   ► Thunderbolt 3, Thunderbolt 4,

00:41:51 ◼   ► that lasts a couple of generations.

00:41:52 ◼   ► I think both the M2 and the M3

00:41:54 ◼   ► have the same speed Thunderbolt interface.

00:41:56 ◼   ► So if that was in a separate chiplet,

00:41:57 ◼   ► you can just reuse that one from before on the future chip.

00:42:01 ◼   ► And that could still be seven nanometer, for example.

00:42:03 ◼   ► I don't know if this is the right part of the chip

00:42:05 ◼   ► to be reused, but you can design it once

00:42:07 ◼   ► and reuse that little sub chiplet in future generations,

00:42:11 ◼   ► as long as you can do an interconnect between all of these

00:42:13 ◼   ► that has all the attributes,

00:42:14 ◼   ► the subscribe low latency, blah, blah, blah.

00:42:16 ◼   ► So far, Apple's packaging innovation has been the Interposer,

00:42:19 ◼   ► which is technically impressive,

00:42:20 ◼   ► and it makes the ultra possible.

00:42:22 ◼   ► But this is more like a more general purpose thing

00:42:24 ◼   ► of like lots of different little, you know,

00:42:27 ◼   ► a city of chips,

00:42:28 ◼   ► lots of different little islands on your chip

00:42:30 ◼   ► to even to just make what Apple does currently

00:42:32 ◼   ► in a single SoC,

00:42:32 ◼   ► and that can make things less expensive

00:42:34 ◼   ► and improve your yields as well.

00:42:37 ◼   ► - And then you want to tell me

00:42:38 ◼   ► about the Universal Chiplet Interconnect Express.

00:42:40 ◼   ► Is that a train, Jon?

00:42:42 ◼   ► - UCI and then lowercase E.

00:42:44 ◼   ► So this is one of these multi company consortiums

00:42:49 ◼   ► to do exactly what we were just describing.

00:42:51 ◼   ► So this is a press release from March 2nd, 2022.

00:42:54 ◼   ► It says Intel,

00:42:55 ◼   ► along with advanced semiconductor engineering,

00:42:57 ◼   ► AMD, ARM, Google Cloud, Meta, Microsoft, Qualcomm,

00:43:00 ◼   ► Samsung, and TSMC have announced the establishment

00:43:03 ◼   ► of an industry consortium

00:43:04 ◼   ► to promote an open die-to-die interconnect standard

00:43:07 ◼   ► called Universal Chiplet Interconnect Express or UCIE.

00:43:10 ◼   ► The chiplet ecosystem created by UCIE is a critical step

00:43:12 ◼   ► in the creation of a unified standard

00:43:14 ◼   ► for interoperable chiplets,

00:43:16 ◼   ► which will ultimately allow for next generation

00:43:17 ◼   ► of technological innovations.

00:43:19 ◼   ► Apple's not on that list, but TSMC is,

00:43:21 ◼   ► and Apple has a good relationship with them.

00:43:24 ◼   ► So if we're looking towards a future

00:43:26 ◼   ► where it becomes untenable to continue making larger

00:43:30 ◼   ► and larger dies, you know,

00:43:32 ◼   ► with everything on them and these SOCs,

00:43:34 ◼   ► a good industry standard system for making chiplets

00:43:39 ◼   ► at whatever the most economical and useful size is

00:43:43 ◼   ► and weaving them together to essentially form an SOC

00:43:46 ◼   ► out of it seems like a good idea.

00:43:49 ◼   ► Doing it all on one die probably still has advantages,

00:43:52 ◼   ► but price may not be one of those advantages.

00:43:55 ◼   ► And again, Intel is super interested

00:43:57 ◼   ► because they essentially can't do that

00:43:59 ◼   ► without having TSMC do all of their fabbing for them

00:44:01 ◼   ► if they want the top tier technology.

00:44:04 ◼   ► Apple doesn't currently have that problem,

00:44:06 ◼   ► but the fact that Johnny Cerugi mentioned stuff

00:44:08 ◼   ► about packaging and the fact that this whole chiplets thing,

00:44:12 ◼   ► like other companies are already doing this,

00:44:13 ◼   ► AMD has been big into the chiplets thing for a while,

00:44:15 ◼   ► Intel is putting a lot of money and research into this

00:44:18 ◼   ► and their chips that they're gonna be coming out with

00:44:21 ◼   ► use some of this technology,

00:44:23 ◼   ► I think it's unavoidable that Apple will start using it too.

00:44:25 ◼   ► Why am I interested in all this?

00:44:27 ◼   ► You know why.

00:44:28 ◼   ► It's to keep the dream alive of a Mac Pro chip

00:44:31 ◼   ► that is not the same as one you can get in an Mac studio.

00:44:34 ◼   ► And what's gonna make that possible?

00:44:36 ◼   ► Apparently with the approach they use

00:44:37 ◼   ► for the past two generations, not economically feasible.

00:44:41 ◼   ► If it's ever gonna be economically feasible,

00:44:43 ◼   ► we need a change.

00:44:44 ◼   ► This change in packaging could bring that about.

00:44:47 ◼   ► Stay tuned, I don't know, 20, 27, who knows?

00:44:50 ◼   ► - Don't mind.

00:44:52 ◼   ► All right, Jon, in keeping with the idea

00:44:56 ◼   ► of keeping the dream alive,

00:44:58 ◼   ► tell me about error network change.

00:45:00 ◼   ► - This is a nightmare, it's not a dream.

00:45:02 ◼   ► - Touche.

00:45:03 ◼   ► - Error network change continues.

00:45:05 ◼   ► Some more research from the internet about this.

00:45:09 ◼   ► So here is a Stack Overflow question.

00:45:12 ◼   ► This person asks, whenever my iPhone and Mac OS

00:45:15 ◼   ► are on the same WiFi, Chrome and Mac OS

00:45:17 ◼   ► often reports the error network changed.

00:45:20 ◼   ► I found that whenever my iPhone and Mac OS

00:45:22 ◼   ► on the same WiFi, a record will often appear

00:45:24 ◼   ► in the routing table in Mac OS

00:45:26 ◼   ► and disappear after a few seconds.

00:45:28 ◼   ► When this route record appears,

00:45:29 ◼   ► my Chrome will most likely have

00:45:31 ◼   ► an error network change error.

00:45:32 ◼   ► I turned off the iPhone's WiFi,

00:45:34 ◼   ► the routing in a record in Mac OS disappeared,

00:45:36 ◼   ► and Chrome no longer had the error.

00:45:38 ◼   ► So this is, everyone is always looking for like,

00:45:40 ◼   ► what causes this?

00:45:41 ◼   ► And so they start looking at anything,

00:45:42 ◼   ► like what changes my network?

00:45:44 ◼   ► What is it about my network that has changed?

00:45:46 ◼   ► And they'll spot something,

00:45:47 ◼   ► and they'll find a thing to attribute it to,

00:45:49 ◼   ► and they'll stop that thing by saying,

00:45:50 ◼   ► it looks like it's happening because my phone is here,

00:45:52 ◼   ► and if I turn off my WiFi, I fix the problem.

00:45:54 ◼   ► Every single one of these that has been reported,

00:45:57 ◼   ► I have experimented with,

00:45:58 ◼   ► and it doesn't actually eliminate the problem,

00:46:02 ◼   ► it just moves it around.

00:46:03 ◼   ► Because the problem is, the network changed

00:46:05 ◼   ► and Chrome flips out about it.

00:46:07 ◼   ► Is the problem that the network is changing?

00:46:08 ◼   ► Is the problem that's Chrome's flipping out about it?

00:46:11 ◼   ► Is it both?

00:46:13 ◼   ► Either way, if you stop the thing

00:46:15 ◼   ► that changes in your network,

00:46:16 ◼   ► doesn't mean that something else

00:46:17 ◼   ► won't also change your network and cause the error.

00:46:20 ◼   ► But I'm glad that somebody found something that helps them.

00:46:22 ◼   ► This next one is about tailscale.

00:46:25 ◼   ► Gil Penderson says, "The tailscale VPN used to trigger this,

00:46:28 ◼   ► but they fixed it with," and there's a link to a patch.

00:46:30 ◼   ► So this is tailscale patching their own code

00:46:33 ◼   ► to avoid this error,

00:46:33 ◼   ► because they were seeing this error happening,

00:46:35 ◼   ► and they were saying, hey, this thing that we're doing,

00:46:38 ◼   ► Mac OS is flipping out about it.

00:46:40 ◼   ► So basically it says iOS/Mac OS,

00:46:44 ◼   ► so it's apparently not just a Mac thing,

00:46:46 ◼   ► will reconfigure their routing

00:46:47 ◼   ► anytime anything minor changes in the netmap.

00:46:50 ◼   ► And so they're changing their code to not do that,

00:46:52 ◼   ► because they don't wanna anger iOS/Mac OS.

00:46:54 ◼   ► This is a change that committed in 2020.

00:46:57 ◼   ► So is our iOS and Mac OS changing their routing more?

00:47:02 ◼   ► Is there something that was part of the operating system

00:47:05 ◼   ► that's causing it?

00:47:07 ◼   ► Lots of theories about this.

00:47:08 ◼   ► One that's not even in the notes here is about,

00:47:11 ◼   ► if you have a dev device,

00:47:12 ◼   ► like if you have an iPhone

00:47:13 ◼   ► that is configured as a dev device,

00:47:15 ◼   ► there's apparently a new way

00:47:16 ◼   ► that Macs communicate to the phones

00:47:19 ◼   ► for purposes of development,

00:47:21 ◼   ► that also uses a network connection

00:47:23 ◼   ► that's brought up and torn down,

00:47:24 ◼   ► and apparently causes network to be changed

00:47:26 ◼   ► as far as Chrome concerned, and it flips out.

00:47:28 ◼   ► So anyway, here's an anonymous bit

00:47:29 ◼   ► that's getting us closer to the root of the problem.

00:47:32 ◼   ► And this is an explanation of it

00:47:34 ◼   ► from the perspective of Chrome, Chromium, whatever.

00:47:38 ◼   ► Because again, this is not just a Mac or iOS problem.

00:47:40 ◼   ► This happens to be on Linux.

00:47:42 ◼   ► This is a, part of the problem is clearly

00:47:45 ◼   ► with the Chrome/Chromium browser engine,

00:47:50 ◼   ► but the operating system has to participate in it as well,

00:47:52 ◼   ► because they're the ones that control the network

00:47:53 ◼   ► that is changing.

00:47:54 ◼   ► Anyway, anonymous writes,

00:47:56 ◼   ► "The network layer from Chromium is available

00:47:58 ◼   ► as a standalone cross-platform library called Cronet,

00:48:01 ◼   ► which is open source

00:48:02 ◼   ► and used in other non-Google applications.

00:48:04 ◼   ► I work at Google on a major Android app

00:48:05 ◼   ► that uses Cronet extensively,

00:48:07 ◼   ► so I've got experience with their network changed."

00:48:09 ◼   ► This issue appears to be with QUIC, Q-U-I-C,

00:48:12 ◼   ► which is the UDP-based web protocol thing.

00:48:15 ◼   ► We'll put a link to the Wikipedia page.

00:48:16 ◼   ► You can read about it.

00:48:17 ◼   ► And/or HTTP/3, which is a new version of HTTP,

00:48:20 ◼   ► which is sensitive to changing networks.

00:48:22 ◼   ► Whenever the networks on the device change,

00:48:24 ◼   ► the connection needs to be reset.

00:48:25 ◼   ► This isn't exactly a bug.

00:48:26 ◼   ► It's a necessary step.

00:48:28 ◼   ► Not doing this would mean connections

00:48:29 ◼   ► wouldn't work properly.

00:48:30 ◼   ► Unfortunately, this causes error network change

00:48:33 ◼   ► when the device's network changed.

00:48:35 ◼   ► As previously mentioned,

00:48:36 ◼   ► this can be for non-obvious reasons,

00:48:37 ◼   ► such as Docker desktop customizing the networks repeatedly,

00:48:40 ◼   ► VPN apps misbehaving,

00:48:41 ◼   ► cell signal dropping in and out, et cetera.

00:48:44 ◼   ► There are ways in which this can be fixed.

00:48:46 ◼   ► First, not using HTTP/QUIC is the easiest.

00:48:50 ◼   ► Safari and other Apple stuff doesn't use this yet,

00:48:52 ◼   ► so it isn't currently experiencing these issues.

00:48:54 ◼   ► However, this will likely change

00:48:55 ◼   ► as Apple rolls out HTTP/3 support.

00:48:57 ◼   ► Apart from this issue, it's a much better technology,

00:48:59 ◼   ► particularly for mobile devices.

00:49:00 ◼   ► So this is the first offered explanation of,

00:49:03 ◼   ► hey, why doesn't Safari have this problem?

00:49:04 ◼   ► The theory is, oh, Safari doesn't use QUIC or HTTP/3,

00:49:08 ◼   ► and so it's not going to see any of these errors.

00:49:11 ◼   ► Difficult to test because,

00:49:13 ◼   ► as far as I was able to determine

00:49:15 ◼   ► in conversing with this person,

00:49:16 ◼   ► there is no way to disable HTTP/3 and QUIC in Chrome

00:49:20 ◼   ► to disable it entirely.

00:49:21 ◼   ► There's some flags in Chrome flags

00:49:22 ◼   ► where you can turn some things off,

00:49:23 ◼   ► but I think you can't turn it off entirely.

00:49:25 ◼   ► So it's hard to do an apples-to-apples comparison.

00:49:28 ◼   ► But you pin your hopes on the smallest thing.

00:49:30 ◼   ► If Apple does plan on rolling out HTTP/3 or QUIC,

00:49:33 ◼   ► their browser is going to break like Chrome is.

00:49:35 ◼   ► And hopefully they'll be like, oh, geez,

00:49:37 ◼   ► this browser has become useless

00:49:39 ◼   ► because half of my HTTP connections fail.

00:49:41 ◼   ► Let's see if we can fix this,

00:49:43 ◼   ► as opposed to now where apparently no one is doing anything.

00:49:45 ◼   ► Next possible fix.

00:49:46 ◼   ► OSes and networking apps will need to be more careful

00:49:49 ◼   ► about network changes.

00:49:50 ◼   ► Apple will probably fix this in a Sonoma point release.

00:49:52 ◼   ► Other platforms I've seen this on are working on fixes.

00:49:56 ◼   ► I wish I had this person's optimism.

00:49:59 ◼   ► Will it be fixed in a Sonoma point release?

00:50:01 ◼   ► Does Apple even know about this?

00:50:03 ◼   ► I haven't bothered reporting this

00:50:05 ◼   ► because they're just going to say it's a Chrome error,

00:50:07 ◼   ► and I'm not sure they're wrong.

00:50:09 ◼   ► Like, I honestly don't know how much of the blame

00:50:11 ◼   ► to a portion to Chrome versus Mac is.

00:50:13 ◼   ► Anecdotal evidence is that this is way worse in Sonoma.

00:50:17 ◼   ► There are a bunch of new network-related demons

00:50:19 ◼   ► for various things that people have assigned blame to.

00:50:21 ◼   ► Oh, it's because of the Utahn thing.

00:50:23 ◼   ► It's because of talking to a developer iPhone device.

00:50:26 ◼   ► You know, it's because of Docker.

00:50:28 ◼   ► It's because of a VPN.

00:50:29 ◼   ► It's because whenever your iPhone and your Mac

00:50:33 ◼   ► are on the same Wi-Fi network, it happens.

00:50:35 ◼   ► Like, I don't know.

00:50:37 ◼   ► But I don't know how to communicate this

00:50:41 ◼   ► to anybody involved other than to say

00:50:43 ◼   ► Chrome no longer works on Mac in a reliable way,

00:50:45 ◼   ► like in a fundamental, no longer useful

00:50:48 ◼   ► as a browser kind of way because, you know,

00:50:51 ◼   ► some large representative requests just fail,

00:50:54 ◼   ► and it's not great.

00:50:55 ◼   ► Finally, the last thing.

00:50:57 ◼   ► There is an experimentation going on in Cronet

00:50:59 ◼   ► to improve things by migrating connections

00:51:01 ◼   ► across network interfaces.

00:51:03 ◼   ► This is the migrate sessions on network change V2 flag

00:51:06 ◼   ► or feature or whatever.

00:51:07 ◼   ► We'll put a link to the thing in the show notes.

00:51:10 ◼   ► This isn't just a bug fix, though.

00:51:11 ◼   ► It has trade-offs, so it likely needs to be done

00:51:13 ◼   ► in combination with improvements from OSes

00:51:15 ◼   ► and networking apps.

00:51:16 ◼   ► So the idea behind this thing is,

00:51:18 ◼   ► hey, if there's some connection

00:51:20 ◼   ► in this Cronet networking library,

00:51:22 ◼   ► and the network changes, and we have to, you know,

00:51:25 ◼   ► not use that connection anymore because it won't work,

00:51:28 ◼   ► whatever was using that connection,

00:51:30 ◼   ► why don't we smoothly migrate it over

00:51:32 ◼   ► to the new working connection,

00:51:33 ◼   ► and then the application that is using Cronet,

00:51:36 ◼   ► which would be the Chromium browser engine,

00:51:38 ◼   ► doesn't need to worry that that happened.

00:51:40 ◼   ► It will just continue to work.

00:51:41 ◼   ► That would be great, but it's not just like,

00:51:43 ◼   ► it's not just a bug fix

00:51:44 ◼   ► because there are performance implications of that.

00:51:46 ◼   ► Migrating, it is not free.

00:51:47 ◼   ► It takes time, and it's not the same

00:51:50 ◼   ► as just not having the network change.

00:51:52 ◼   ► So I don't know what the solution is here,

00:51:54 ◼   ► but it's extremely frustrating.

00:51:56 ◼   ► We did not have an Air Network Change shirt

00:51:58 ◼   ► as part of this sale,

00:51:59 ◼   ► but I'm seriously thinking about it for future ones.

00:52:01 ◼   ► This continues to happen.

00:52:04 ◼   ► And then the final bit, this is really the icing on the cake.

00:52:06 ◼   ► This is not seemingly related to Air Network Change,

00:52:11 ◼   ► but kind of is.

00:52:12 ◼   ► This was a toot by someone named Thomas who said,

00:52:15 ◼   ► "I just learned what the user interface

00:52:17 ◼   ► in the SpaceX capsules run.

00:52:19 ◼   ► The capsules that provide life support

00:52:21 ◼   ► for people traveling into space

00:52:22 ◼   ► have to be absolutely reliable.

00:52:24 ◼   ► The user interface that controls an explosion.

00:52:27 ◼   ► It runs a home-compiled version of Chromium,

00:52:29 ◼   ► and the UI is written in JavaScript."

00:52:31 ◼   ► And this person wrote this to say,

00:52:32 ◼   ► "Can you believe they're using web technologies

00:52:34 ◼   ► and a thing that has to be reliable?"

00:52:36 ◼   ► But that's not the thing that sticks out.

00:52:38 ◼   ► The thing that sticks out is the word "chromium."

00:52:40 ◼   ► Do you think if they have Air Network Changed

00:52:42 ◼   ► in a SpaceX rocket, that might be a problem?

00:52:47 ◼   ► And who's responsible for fixing that bug?

00:52:49 ◼   ► I'm sure it's running Linux or whatever,

00:52:50 ◼   ► but, again, people running Chrome on Linux

00:52:52 ◼   ► or Chromium-based things on Linux

00:52:54 ◼   ► are also experiencing this problem.

00:52:57 ◼   ► Sometimes I feel like I'm the only person with this problem,

00:52:59 ◼   ► but every time I talk about it on the show,

00:53:01 ◼   ► people send me toots on Mastodon.

00:53:03 ◼   ► Here I am, I'm getting it, I'm getting it on Air Work Changed.

00:53:05 ◼   ► I got people on Linux, people on Macs.

00:53:07 ◼   ► So far no one on iOS,

00:53:08 ◼   ► 'cause maybe those errors are hidden from you

00:53:10 ◼   ► on most iOS apps, but...

00:53:12 ◼   ► I mean, I still think my Winter Dragon bug

00:53:15 ◼   ► is more important 'cause I could just use Safari.

00:53:17 ◼   ► But for the people in the SpaceX capsules

00:53:21 ◼   ► that have a UI that's running Chromium,

00:53:22 ◼   ► maybe talk to your bosses about this.

00:53:27 ◼   ► We are brought to you this episode by Squarespace,

00:53:29 ◼   ► the all-in-one website platform

00:53:31 ◼   ► for entrepreneurs to stand out and succeed online.

00:53:34 ◼   ► Whether you're just starting out

00:53:35 ◼   ► or managing a growing brand,

00:53:37 ◼   ► Squarespace makes it easy to create a beautiful website,

00:53:40 ◼   ► engage with your audience, and sell anything,

00:53:42 ◼   ► from your products to your content to your time,

00:53:44 ◼   ► all in one place and all on your terms.

00:53:47 ◼   ► Squarespace has an amazing platform to build websites.

00:53:51 ◼   ► It has never been easier for anyone

00:53:53 ◼   ► to unlock unbreakable creativity

00:53:56 ◼   ► with their awesome all-new Fluid Engine.

00:53:58 ◼   ► You start with a best-in-class website template

00:54:00 ◼   ► and then you customize every design detail you want

00:54:03 ◼   ► with reimagined drag-and-drop tech

00:54:05 ◼   ► that works on desktop and mobile.

00:54:06 ◼   ► You can really stretch your imagination online

00:54:09 ◼   ► with Fluid Engine, built in with all new Squarespace sites.

00:54:12 ◼   ► And Squarespace has amazing online storefront support.

00:54:15 ◼   ► So you can sell your products on an online store.

00:54:18 ◼   ► And not just physical goods,

00:54:20 ◼   ► even digital goods or service products,

00:54:22 ◼   ► or even things like time slots.

00:54:24 ◼   ► Squarespace has all the tools you need

00:54:26 ◼   ► to start selling online,

00:54:28 ◼   ► all with flexible payment options,

00:54:30 ◼   ► seamless checkouts that really convert very well,

00:54:33 ◼   ► all sorts of options for people to pay,

00:54:34 ◼   ► including of course credit cards,

00:54:36 ◼   ► but also PayPal, Apple Pay, Buy Now, Pay Later,

00:54:39 ◼   ► all sorts of great support there.

00:54:41 ◼   ► All of this is backed by powerful analytics,

00:54:43 ◼   ► so you can grow your business with insights

00:54:46 ◼   ► about things like where your site visits

00:54:47 ◼   ► and sales are coming from,

00:54:48 ◼   ► which marketing channels are most effective.

00:54:50 ◼   ► You can improve your website

00:54:51 ◼   ► and build a marketing strategy based on your top keywords,

00:54:54 ◼   ► or most popular products and content,

00:54:55 ◼   ► or whatever you want.

00:54:57 ◼   ► You can try everything on Squarespace at squarespace.com

00:55:00 ◼   ► and start a free trial there.

00:55:03 ◼   ► You can see it all in trial mode,

00:55:04 ◼   ► and you can see how well it's gonna work for you.

00:55:06 ◼   ► When you're ready to launch,

00:55:07 ◼   ► go to squarespace.com/atp for 10% off

00:55:10 ◼   ► your first purchase of a website or domain.

00:55:13 ◼   ► So once again, squarespace.com to start that free trial.

00:55:15 ◼   ► When you're ready to buy,

00:55:16 ◼   ► squarespace.com/atp for 10% off.

00:55:19 ◼   ► Thank you so much to Squarespace for sponsoring our show.

00:55:23 ◼   ► (upbeat music)

00:55:26 ◼   ► As we record this, it is Monday, what's it say,

00:55:28 ◼   ► the 20th, is that right?

00:55:30 ◼   ► Yes, Monday the 20th.

00:55:31 ◼   ► In the last three days, there have been approximately,

00:55:35 ◼   ► carry the seven, 7,304 different CEOs for OpenAI.

00:55:40 ◼   ► - Somehow still only one board.

00:55:42 ◼   ► - Right, yeah.

00:55:43 ◼   ► - Well, we've been recording for a while, so.

00:55:45 ◼   ► - Yeah, right.

00:55:46 ◼   ► - That's also fair.

00:55:47 ◼   ► So I don't really have too much to say about this,

00:55:50 ◼   ► but I guess--

00:55:51 ◼   ► - Just read the headlines in the notes.

00:55:52 ◼   ► - All right, so these are the headlines,

00:55:55 ◼   ► and we'll link these in the show notes.

00:55:56 ◼   ► - I think this is like 48 hours span, 24, 48.

00:56:00 ◼   ► - Yeah, it's been almost no time.

00:56:02 ◼   ► - So these are the headlines as per the Verge.

00:56:05 ◼   ► Sam Altman fired a CEO of OpenAI.

00:56:08 ◼   ► That was on the 17th.

00:56:10 ◼   ► - And by the way, out of nowhere.

00:56:12 ◼   ► - Yeah, so let's get through the headlines,

00:56:14 ◼   ► and then let, yeah, I think it's worth unpacking all this.

00:56:17 ◼   ► Then next, Twitch co-founder, Emmet Sheer,

00:56:19 ◼   ► is the new CEO of OpenAI, that was apparently today.

00:56:23 ◼   ► Microsoft hires former OpenAI CEO, Sam Altman, also today.

00:56:27 ◼   ► Hundreds of OpenAI employees starting to resign

00:56:30 ◼   ► and join Microsoft today.

00:56:32 ◼   ► Is Sam Altman joining Microsoft?

00:56:34 ◼   ► Satya Dadella doesn't seem to know.

00:56:37 ◼   ► - I think there's hours between those things.

00:56:39 ◼   ► - Yeah.

00:56:40 ◼   ► - How can you have, like, yes, it's a comedy of errors,

00:56:44 ◼   ► but then every time there's a headline,

00:56:46 ◼   ► it's like, this is a thing that happened,

00:56:47 ◼   ► and it's like, but has it?

00:56:49 ◼   ► - Yeah, exactly.

00:56:51 ◼   ► - I skip the ones where it's like, oh, he was fired,

00:56:54 ◼   ► but there are negotiations for him to come back.

00:56:55 ◼   ► But no, he's not coming back.

00:56:56 ◼   ► But yes, he is coming back,

00:56:57 ◼   ► but they're talking about coming back.

00:56:58 ◼   ► But now he's at Microsoft,

00:56:59 ◼   ► but has he been hired by Microsoft?

00:57:01 ◼   ► And this is not just like,

00:57:01 ◼   ► oh, we're wondering what's going on.

00:57:04 ◼   ► That's why the headline says,

00:57:05 ◼   ► is Sam Altman joining Microsoft?

00:57:07 ◼   ► Satya Dadella, who by the way is the CEO of Microsoft,

00:57:09 ◼   ► doesn't seem to know.

00:57:10 ◼   ► Satya Dadella announced,

00:57:12 ◼   ► hey, Sam Altman's coming to Microsoft.

00:57:14 ◼   ► But then when asked about it, like, hours later,

00:57:16 ◼   ► it was like, well, not really sure.

00:57:19 ◼   ► I don't check back in a few hours.

00:57:22 ◼   ► Boy.

00:57:23 ◼   ► - Yeah, so let me try to give

00:57:24 ◼   ► the quick executive summary of this.

00:57:26 ◼   ► Now, I'm surely gonna get a little bit wrong,

00:57:28 ◼   ► but the general gist of it is,

00:57:30 ◼   ► I think it was late Friday, Eastern time,

00:57:33 ◼   ► there was a blog post from the OpenAI board, basically,

00:57:37 ◼   ► that said that Sam Altman, who was the former CEO,

00:57:42 ◼   ► maybe by the time you're listening to this,

00:57:44 ◼   ► might be the CEO again, who knows?

00:57:45 ◼   ► But former CEO had basically not,

00:57:48 ◼   ► what was the phrase they used?

00:57:50 ◼   ► Not been, forthcoming wasn't the word they used,

00:57:54 ◼   ► but it was something along those lines.

00:57:55 ◼   ► - It was pretty severe.

00:57:58 ◼   ► - It was vague, but like, as, you know,

00:58:00 ◼   ► normally when a CEO is fired,

00:58:04 ◼   ► this is the privilege of being in the executive ranks.

00:58:06 ◼   ► When a regular person is fired, you just get fired.

00:58:09 ◼   ► But if you're a CEO, no matter how terrible

00:58:12 ◼   ► whatever you did was, no matter how badly you screwed up,

00:58:16 ◼   ► you're always like, you know, moving on to spend more time

00:58:19 ◼   ► with your family or want to pursue other interests,

00:58:23 ◼   ► or like they have some euphemism about how you're,

00:58:25 ◼   ► it's not like we're firing him.

00:58:26 ◼   ► It's a mutual decision we've all come to,

00:58:28 ◼   ► and he's decided to leave, and he's, you know,

00:58:31 ◼   ► pursuing his passions in this other realm,

00:58:33 ◼   ► and we thank him for all his blah, blah, blah,

00:58:34 ◼   ► and this wasn't like that.

00:58:36 ◼   ► This was as close as you're ever gonna get

00:58:38 ◼   ► to saying we fired him 'cause we don't like him.

00:58:40 ◼   ► - Yeah, that's how I remember being with Forrestal,

00:58:42 ◼   ► was basically like, it was clear that he got fired,

00:58:45 ◼   ► but everyone put on the happy face.

00:58:47 ◼   ► We would like Scott to spend more time with his family.

00:58:51 ◼   ► Yes, I would like to spend more time with my family.

00:58:55 ◼   ► - And usually, and usually in those type of statements,

00:58:57 ◼   ► there's something from the person who was fired,

00:58:59 ◼   ► like quote, you know, whatever, quotes from them, right?

00:59:02 ◼   ► They'll say like, I'm happy that I spent all my time

00:59:04 ◼   ► with this company, I'm glad to be moving on,

00:59:06 ◼   ► and blah, blah, blah, like it's a participatory process.

00:59:08 ◼   ► But what happened with Sam Altman,

00:59:11 ◼   ► the story I heard on Pocket Osean 2 is that

00:59:13 ◼   ► he was on stage doing something,

00:59:15 ◼   ► like in his official capacity as CEO,

00:59:17 ◼   ► and he said, okay, well, you know, thanks everybody,

00:59:20 ◼   ► I have to go, I have a meeting.

00:59:21 ◼   ► In that meeting, he got on, he got on a call,

00:59:25 ◼   ► this is, I was listening to the Verges podcast about this,

00:59:27 ◼   ► he got onto the call using Google Meet,

00:59:28 ◼   ► and one of the people on the show said,

00:59:30 ◼   ► even for $10 billion, no one will use Teams.

00:59:32 ◼   ► Microsoft invested $10 billion or whatever in OpenAI.

00:59:37 ◼   ► So anyway, he goes to his meeting, he sits down,

00:59:39 ◼   ► and it's, you know, an online meeting, right?

00:59:42 ◼   ► And all they did in the online meeting was read him

00:59:45 ◼   ► the blog post he just referred to, okay,

00:59:46 ◼   ► so they just put in, that's what you do in these meetings,

00:59:48 ◼   ► you have a script, you stick to it so you don't get sued,

00:59:50 ◼   ► you just read the words, so if you're wondering,

00:59:52 ◼   ► like, oh, they must have had an extensive meeting

00:59:53 ◼   ► in which they explained to him why he was fired,

00:59:56 ◼   ► when we say this came out of nowhere,

00:59:57 ◼   ► didn't come out of nowhere to us on the outside,

00:59:59 ◼   ► 'cause what do we know?

01:00:00 ◼   ► Came out of nowhere to Sam Altman,

01:00:02 ◼   ► who just got done doing a CEO thing and said,

01:00:04 ◼   ► oh, I gotta go to a meeting,

01:00:05 ◼   ► and they just read him a thing and said, yeah, you're fired.

01:00:07 ◼   ► - And by the way, the thing he was doing,

01:00:09 ◼   ► like having their whole developer conference,

01:00:11 ◼   ► was widely lauded by the industry as like an amazing thing,

01:00:15 ◼   ► everyone thought he did an amazing job,

01:00:16 ◼   ► this is a great path beyond--

01:00:18 ◼   ► - Everyone's excited about open AI.

01:00:19 ◼   ► - Yeah, everyone's super excited about open AI and him,

01:00:22 ◼   ► like it was, that's why when this news dropped

01:00:25 ◼   ► that the board fired him,

01:00:26 ◼   ► everyone in the entire tech business was like, what?

01:00:30 ◼   ► Are you serious, is this a joke, like what happened?

01:00:32 ◼   ► - Yeah, so here's from the blog post

01:00:34 ◼   ► I was able to dig it up while you guys were talking,

01:00:36 ◼   ► this is the blog post from the open AI board,

01:00:38 ◼   ► Mr. Altman's departure follows a deliberate review process

01:00:41 ◼   ► by the board which concluded that he was not consistently

01:00:44 ◼   ► candid in his communications with the board,

01:00:46 ◼   ► hindering its ability to exercise its responsibilities.

01:00:48 ◼   ► The board no longer has confidence in his ability

01:00:50 ◼   ► to continue leading open AI.

01:00:52 ◼   ► - Which by the way, like that's a really serious accusation,

01:00:56 ◼   ► like that is like, potentially misleading the board

01:01:00 ◼   ► is in many cases a crime,

01:01:02 ◼   ► like that is a really serious accusation,

01:01:05 ◼   ► that's why like this isn't just,

01:01:07 ◼   ► oh, we had creative differences,

01:01:09 ◼   ► we don't like him anymore,

01:01:10 ◼   ► that's a very serious thing to throw around,

01:01:13 ◼   ► what they threw around.

01:01:14 ◼   ► - And by the way, as far as I understand it,

01:01:17 ◼   ► the board doesn't have to have a reason,

01:01:19 ◼   ► like they have complete control,

01:01:21 ◼   ► there's no like power struggle or whatever,

01:01:23 ◼   ► the structure of this company is such that the board

01:01:26 ◼   ► can decide, oh, we don't want you to be CEO,

01:01:28 ◼   ► I think for pretty much any reason,

01:01:30 ◼   ► like he needs to do anything wrong,

01:01:31 ◼   ► like it just happens all the time,

01:01:32 ◼   ► the board's like, we want a different CEO,

01:01:34 ◼   ► we wanna go in a different direction,

01:01:35 ◼   ► that happens all the time,

01:01:36 ◼   ► they don't need to explain,

01:01:38 ◼   ► which is why you can do the,

01:01:39 ◼   ► oh, they're going to spend more time with the family

01:01:41 ◼   ► or whatever, whatever the real reason is,

01:01:43 ◼   ► it's not like they need to make something up,

01:01:44 ◼   ► like we can't fire him unless we say that he lied to us,

01:01:47 ◼   ► they could have just said later,

01:01:49 ◼   ► but they threw that in there to say,

01:01:51 ◼   ► oh, and we have reasons,

01:01:52 ◼   ► we think he has not been quote unquote,

01:01:54 ◼   ► consistently candid,

01:01:55 ◼   ► like they didn't have to put that in there for any reason,

01:01:58 ◼   ► other than to just, I mean, I don't know,

01:02:01 ◼   ► maybe this will give us cover because it's vague enough,

01:02:03 ◼   ► people can imagine some terrible thing happened,

01:02:05 ◼   ► but bottom line is, apparently a majority,

01:02:09 ◼   ► we don't even know how much,

01:02:10 ◼   ► because it's not a public company,

01:02:12 ◼   ► so it's not like they have public records to say,

01:02:13 ◼   ► was this a unanimous vote,

01:02:14 ◼   ► did everybody vote to kick him out or whatever,

01:02:16 ◼   ► but a majority of the people on the board

01:02:19 ◼   ► decided they don't want him to be CEO anymore,

01:02:22 ◼   ► and they made this decision without consulting Sam Altman

01:02:25 ◼   ► at all, I mean, maybe they've had past conversations

01:02:27 ◼   ► about it, maybe he should have known it's coming

01:02:29 ◼   ► because they said in their last meeting,

01:02:30 ◼   ► if you do really well in that dev conference, you're fired,

01:02:33 ◼   ► I don't know what they might have said to him.

01:02:36 ◼   ► - So the impression I get, and I don't have a lot of facts,

01:02:41 ◼   ► nobody has facts here at the moment,

01:02:43 ◼   ► but the impression I got based on the coverage that I've read

01:02:47 ◼   ► is that there were two different factions, or tribes,

01:02:49 ◼   ► I think Sam had called them at one point or another,

01:02:51 ◼   ► two different tribes within OpenAI.

01:02:54 ◼   ► OpenAI was originally founded in 2014, 2015 thereabouts

01:02:58 ◼   ► as a nonprofit, and their theory was,

01:03:01 ◼   ► we wanna make basically artificial general intelligence,

01:03:04 ◼   ► we wanna make AI, but we understand that with this great

01:03:09 ◼   ► power comes great responsibility, and we wanna do this

01:03:12 ◼   ► to improve the lives of people everywhere,

01:03:17 ◼   ► but we're gonna try to do it very methodically,

01:03:19 ◼   ► very deliberately, very safely.

01:03:21 ◼   ► And that was how the company was founded,

01:03:24 ◼   ► and it's not even really a company, I guess, it's--

01:03:28 ◼   ► - It's a nonprofit. - It's a nonprofit, right?

01:03:30 ◼   ► So that's how it was founded.

01:03:32 ◼   ► Well, at some point, you know, Altman swoops in,

01:03:34 ◼   ► I don't recall exactly when it was in the timeline,

01:03:36 ◼   ► but it's clear that Sam Altman is very Silicon Valley VC,

01:03:41 ◼   ► which I personally find to be a very ugly stereotype,

01:03:46 ◼   ► like, I just don't care for that whole like,

01:03:48 ◼   ► grow, grow, grow, take over the world mindset.

01:03:50 ◼   ► - And he is the embodiment of that stereotype,

01:03:52 ◼   ► he used to be with Y Combinator, which is also kind of

01:03:54 ◼   ► a poster child for that type of VC.

01:03:56 ◼   ► - Yeah, and he's not the guy who had the day-of-phone,

01:03:58 ◼   ► night-of-phone, he's the guy who wore the multiple

01:04:00 ◼   ► polo shirts at the same time.

01:04:01 ◼   ► - Yeah, sorry to keep track of all that, Sam, yeah.

01:04:04 ◼   ► - Exactly, now, and I'm trying to make plain,

01:04:07 ◼   ► to quote Merlin, my priors on this,

01:04:09 ◼   ► like, I find that whole Silicon Valley mindset

01:04:11 ◼   ► to be very off-putting, you know, grow,

01:04:14 ◼   ► grow no matter what, we don't care who we run over

01:04:17 ◼   ► in the process, grow, grow, grow, you know,

01:04:19 ◼   ► let's make money, make money, and that's all that matters

01:04:22 ◼   ► is making money, nothing else matters,

01:04:23 ◼   ► grow and make money, grow and make money,

01:04:25 ◼   ► or sometimes you don't even have to make money,

01:04:26 ◼   ► you just grow.

01:04:27 ◼   ► Well, so anyway, so consider that when I give,

01:04:29 ◼   ► one of my, you know, lip turns up,

01:04:31 ◼   ► as I tell you this story, but nevertheless,

01:04:34 ◼   ► so Sam comes in and he's grow, grow, grow,

01:04:36 ◼   ► money, money, money, and everything,

01:04:38 ◼   ► I guess everyone was kind of coexisting okay,

01:04:41 ◼   ► and even though they disagreed internally,

01:04:44 ◼   ► everything was mostly all right.

01:04:45 ◼   ► Well, then they released ChatGPT, what, about a year ago,

01:04:48 ◼   ► and it makes a tremendous splash.

01:04:53 ◼   ► I think a lot of people are saying it's the quickest

01:04:55 ◼   ► adopted consumer product ever, because, you know,

01:04:58 ◼   ► they had something like 100 million signups

01:05:00 ◼   ► in the span of like four minutes,

01:05:01 ◼   ► not literally, but you know what I mean,

01:05:02 ◼   ► and so now money is becoming a thing, like for real,

01:05:07 ◼   ► and suddenly we have to pay the piper on this division

01:05:11 ◼   ► between money, money, money, and let's do this

01:05:14 ◼   ► for the good of all people, and their chief scientist,

01:05:17 ◼   ► Ilya, something or other, I don't have the name

01:05:19 ◼   ► in front of me, I guess was more on the let's play it safe,

01:05:23 ◼   ► let's be deliberate, let's do this for the good

01:05:25 ◼   ► of humanity, and Sam was very much money, money, money,

01:05:28 ◼   ► and at some point, it appears something gave,

01:05:33 ◼   ► and the rumblings that I heard, not from like sources

01:05:37 ◼   ► or anything, just based on the reporting I read,

01:05:41 ◼   ► was that Ilya had convinced the board,

01:05:43 ◼   ► of which I think he is a part of the board,

01:05:45 ◼   ► let's get rid of Sam, and then all hell broke loose,

01:05:49 ◼   ► and now, as we record on Monday night,

01:05:51 ◼   ► we don't really know what the latest and greatest is,

01:05:54 ◼   ► but I understand, you know, both sides of this.

01:05:58 ◼   ► Like on the one side, if the board really did establish,

01:06:00 ◼   ► or if OpenAI really was established as a non-profit,

01:06:02 ◼   ► which I think is factual, I don't think that's up for grabs,

01:06:05 ◼   ► it is within reason for them to say, well, hold on,

01:06:08 ◼   ► suddenly we've taken a turn and pivoted

01:06:10 ◼   ► to money, money, money, we don't like that,

01:06:13 ◼   ► and Sam seems to, by most metrics,

01:06:16 ◼   ► have pivoted them in that direction,

01:06:18 ◼   ► so if we don't like this pivot to money, money, money,

01:06:20 ◼   ► then we probably don't like Sam anymore,

01:06:22 ◼   ► so on the surface, I don't have a problem with that,

01:06:24 ◼   ► and again, my priors tell me, yeah, screw that Silicon Valley

01:06:27 ◼   ► nut job, you know, let's do this for the good of people

01:06:30 ◼   ► rather than the good of the almighty dollar.

01:06:32 ◼   ► - Well, wait till you hear about the other side of that,

01:06:34 ◼   ► because I think that they're both nut jobs.

01:06:36 ◼   ► - Well, that's fair too, that's fair too.

01:06:38 ◼   ► But I'm just trying to make plain, you know,

01:06:40 ◼   ► this is my biases coming to light.

01:06:43 ◼   ► That being said, unquestionably,

01:06:46 ◼   ► ChatGPT and the work of OpenAI and Dolly,

01:06:49 ◼   ► and you know, all this AI stuff,

01:06:51 ◼   ► whether or not you think it's cool,

01:06:52 ◼   ► whether or not you think it's good,

01:06:55 ◼   ► I think unquestionably, it's important,

01:06:58 ◼   ► and it's been making a big damn splash,

01:07:01 ◼   ► and I think that there's a lot of interesting things here,

01:07:04 ◼   ► and unlike blockchain, I think there's a lot

01:07:07 ◼   ► of fascinating threads that we can pull,

01:07:09 ◼   ► and I think that there's a lot there,

01:07:12 ◼   ► and because of that, I also have sympathy

01:07:16 ◼   ► for the let's grow this product

01:07:18 ◼   ► and see what the world can do with it,

01:07:20 ◼   ► and let's just, let's not slow down,

01:07:22 ◼   ► let's not, you know, be deliberate about it, screw it,

01:07:25 ◼   ► let's just figure it out as we go,

01:07:26 ◼   ► which is when Silicon Valley is at its best.

01:07:29 ◼   ► So I have very mixed feelings about this,

01:07:32 ◼   ► and I mean, as someone who likes drama more than I should,

01:07:36 ◼   ► because I'm a grown-ass man,

01:07:38 ◼   ► I really shouldn't enjoy drama this much,

01:07:41 ◼   ► but oh, this is delicious drama,

01:07:42 ◼   ► and I am here for the drama of it,

01:07:44 ◼   ► but I honestly don't know who's right, who's wrong,

01:07:47 ◼   ► I'm not sure any of us do.

01:07:49 ◼   ► I don't know what to make of this,

01:07:50 ◼   ► but it is a mess, and it's been a mess

01:07:53 ◼   ► every other hour since late Friday evening.

01:07:56 ◼   ► - Yeah, by the time you hear this episode,

01:07:58 ◼   ► who knows what will have happened.

01:07:59 ◼   ► I think this is actually, I mean,

01:08:01 ◼   ► a lot of, this is the most,

01:08:03 ◼   ► the most Silicon Valley thing about this

01:08:05 ◼   ► is that a lot of the companies that, you know,

01:08:08 ◼   ► from my youth that sort of came out of Silicon Valley,

01:08:11 ◼   ► they were at the forefront of some technology that,

01:08:14 ◼   ► they were at the right place at the right time, right?

01:08:16 ◼   ► They decided to make personal computers,

01:08:18 ◼   ► and personal computers were just becoming possible, right?

01:08:20 ◼   ► You know, you had your Apples, you had your Intel,

01:08:22 ◼   ► all right, you know, microchips, memory, you know,

01:08:24 ◼   ► CPUs, x86, Microsoft with the software, like,

01:08:27 ◼   ► they're riding a wave of something.

01:08:31 ◼   ► And those particular companies, especially, you know,

01:08:33 ◼   ► we know a lot of the names of the founders,

01:08:35 ◼   ► Bill Gates, Steve Jobs, a lot of them were younger people,

01:08:38 ◼   ► and pretty much none of them went into it thinking,

01:08:43 ◼   ► I'm going to found what's going to become

01:08:46 ◼   ► what we know today as tech giants,

01:08:48 ◼   ► 'cause that's not what, you know,

01:08:49 ◼   ► who's the, if you tried to tell a young Steve Jobs

01:08:53 ◼   ► and Steve Wozniak what Apple will look like today,

01:08:55 ◼   ► they'd be like, yeah, right.

01:08:56 ◼   ► No one thinks that's going to happen.

01:08:57 ◼   ► So what happens with these companies

01:08:58 ◼   ► that are incredibly successful is,

01:09:00 ◼   ► they're made up of people, the founders,

01:09:02 ◼   ► but also usually they bring in other people

01:09:04 ◼   ► to help them run the company,

01:09:06 ◼   ► and especially for companies

01:09:07 ◼   ► that experience explosive growth,

01:09:10 ◼   ► the people who are responsible for running that company,

01:09:12 ◼   ► they're just people.

01:09:13 ◼   ► Sometimes they're people who've never done

01:09:15 ◼   ► anything like this before.

01:09:17 ◼   ► And so if you're wondering how can this company

01:09:19 ◼   ► that's been so financially successful

01:09:21 ◼   ► and apparently has such amazing technology

01:09:24 ◼   ► not have their governance straightened out

01:09:26 ◼   ► and basically, you know, conduct their business

01:09:30 ◼   ► in this ridiculous way with this whipsawing of the CEO

01:09:34 ◼   ► being fired and asking him to come back

01:09:36 ◼   ► and regretting telling him to leave and all that stuff,

01:09:38 ◼   ► it's because they're a young tech company

01:09:41 ◼   ► riding the wave of new technology

01:09:43 ◼   ► run by a pinch of people who are bad at doing this.

01:09:46 ◼   ► And that's, I mean, look, Steve Jobs was fired from Apple

01:09:51 ◼   ► by a bunch of people that were brought on to run the company.

01:09:53 ◼   ► That was probably the wrong decision

01:09:55 ◼   ► for them to do at that time.

01:09:56 ◼   ► If they had, you know, nurtured his talent,

01:09:59 ◼   ► perhaps they would have had a more successful 90s,

01:10:01 ◼   ► but you can also understand why they fired him

01:10:03 ◼   ► because, you know, read the history.

01:10:05 ◼   ► Like, a lot of these companies in their early days

01:10:08 ◼   ► at least run up to the edge of being, you know,

01:10:12 ◼   ► doing something terrible but being run in a way

01:10:15 ◼   ► that is not, I'm not gonna say not sane,

01:10:18 ◼   ► but that is not the way you expect

01:10:21 ◼   ► a very wealthy company to be run.

01:10:23 ◼   ► Even with Google, they have the two founders

01:10:25 ◼   ► and we have to bring in the adults to help run things.

01:10:28 ◼   ► You know, it can go badly in so many different ways.

01:10:32 ◼   ► It's very rare that you get a company

01:10:35 ◼   ► that manages to get through this sort of awkward adolescence

01:10:38 ◼   ► and sustain its success into something that continues

01:10:42 ◼   ► and becomes like a company that, you know,

01:10:45 ◼   ► that is more reliable and steady

01:10:47 ◼   ► and does not have weird boardroom drama like this.

01:10:49 ◼   ► So I'm not saying that this is unique to OpenAI at all.

01:10:52 ◼   ► In fact, like I said, I think this is the most

01:10:54 ◼   ► Silicon Valley thing that they've done.

01:10:56 ◼   ► That said, the people in these two camps here,

01:10:59 ◼   ► the Sam Altman and the OpenAI people

01:11:03 ◼   ► are a little bit weird.

01:11:04 ◼   ► They're, both sides are a little bit weird.

01:11:07 ◼   ► So we didn't talk about the Microsoft stuff,

01:11:08 ◼   ► but like, OpenAI is this nonprofit.

01:11:11 ◼   ► Microsoft wanted to do a thing with them.

01:11:13 ◼   ► So here's the thing with OpenAI.

01:11:14 ◼   ► Their goal is to make artificial general intelligence,

01:11:16 ◼   ► like 9,000 or whatever, that doesn't kill you.

01:11:19 ◼   ► - Big asterisk there.

01:11:21 ◼   ► - Yeah, yeah.

01:11:21 ◼   ► They haven't done that, to be clear.

01:11:23 ◼   ► That's like their aspirational goal.

01:11:24 ◼   ► It's like their mission statement.

01:11:26 ◼   ► But what they did do was make ChatGPT.

01:11:28 ◼   ► And it turns out that ChatGPT is a useful thing

01:11:30 ◼   ► that people can do stuff with.

01:11:32 ◼   ► And everyone's interested in it.

01:11:35 ◼   ► And Microsoft was interested in it and they said,

01:11:37 ◼   ► "Hey, we would like to do stuff with your technology too.

01:11:40 ◼   ► The problem OpenAI has is they want to make artificial

01:11:43 ◼   ► general intelligence, but turns out doing anything,

01:11:46 ◼   ► even approaching that, costs tons and tons of money."

01:11:49 ◼   ► And so OpenAI's idea was like,

01:11:51 ◼   ► "We'll just, we'll raise money somehow.

01:11:54 ◼   ► And people will give us money because they know

01:11:55 ◼   ► we're doing the good work to make how 9,000 not kill us."

01:11:58 ◼   ► And they, the amount of money they were able to raise

01:12:00 ◼   ► was a fraction of the amount that they would need.

01:12:02 ◼   ► Microsoft came and said,

01:12:04 ◼   ► "We kind of like that things you're doing over there.

01:12:06 ◼   ► How about we let you use our massive computing resources

01:12:09 ◼   ► and our data centers to the tune of billions of dollars

01:12:12 ◼   ► worth of free credits, like little Chuck E. Cheese tokens?

01:12:15 ◼   ► Now you can run your stuff in Azure because OpenAI,

01:12:17 ◼   ► OpenAI can't do anything without large amounts of computers,

01:12:22 ◼   ► which takes large amounts of money.

01:12:23 ◼   ► And they're a nonprofit.

01:12:24 ◼   ► And their idea was like, "We'll raise that money.

01:12:26 ◼   ► People will give us money to pursue this."

01:12:28 ◼   ► They just did not get as much money as they would need.

01:12:30 ◼   ► But ChatGPT, Microsoft's like, "Hmm, kind of like that."

01:12:33 ◼   ► So they did this deal where Microsoft is like

01:12:37 ◼   ► giving them $10 billion,

01:12:39 ◼   ► most of which is in the form of credits

01:12:40 ◼   ► to run stuff in their Azure, cloud computing stuff.

01:12:44 ◼   ► But how can they do that?

01:12:46 ◼   ► Well, if you look at the org chart,

01:12:47 ◼   ► it's like there's a nonprofit

01:12:48 ◼   ► and the nonprofit controls this other for-profit thing

01:12:52 ◼   ► that gets money from Microsoft.

01:12:54 ◼   ► But the for-profit part, it's not,

01:12:58 ◼   ► I think it's like what, profit capped or something like that.

01:13:00 ◼   ► Anyway, it's entirely controlled by the nonprofit,

01:13:03 ◼   ► but still Microsoft has this financial interest.

01:13:05 ◼   ► And by the way, as part of this deal,

01:13:06 ◼   ► Microsoft gets all the rights to the open AI IP,

01:13:10 ◼   ► like their intellectual property,

01:13:12 ◼   ► I don't know what their intellectual property is,

01:13:13 ◼   ► but presumably whatever it is they use to make ChatGPT,

01:13:15 ◼   ► Microsoft now has the rights to that.

01:13:17 ◼   ► I think forever, as part of this $10 billion deal.

01:13:21 ◼   ► The only thing Microsoft doesn't have the rights to,

01:13:22 ◼   ► and this is another one of those great deals,

01:13:23 ◼   ► kind of like Microsoft doing the Internet Explorer thing

01:13:27 ◼   ► and saying, "Oh, don't worry.

01:13:28 ◼   ► We'll give you X percent

01:13:30 ◼   ► of all of our Internet Explorer sales,"

01:13:31 ◼   ► and then giving Internet Explorer away for free

01:13:34 ◼   ► to everybody, and so that person got nothing.

01:13:35 ◼   ► Anyway, the company that got nothing from that.

01:13:38 ◼   ► Open AI is like, "Okay, we'll do this $10 billion deal.

01:13:41 ◼   ► We'll get access to your computing resources,

01:13:42 ◼   ► which we kind of need to literally do anything

01:13:44 ◼   ► 'cause we don't have enough money to do this AI stuff.

01:13:47 ◼   ► We don't have enough money to pay for the computers to do it

01:13:49 ◼   ► but what we won't give you,

01:13:50 ◼   ► and we'll license you our current technology,

01:13:53 ◼   ► but what we won't give you is anything having to do

01:13:56 ◼   ► with artificial general intelligence.

01:13:57 ◼   ► So if we invent HAL 9000, Microsoft, you don't get it."

01:14:00 ◼   ► And Microsoft's like, "Okay, I guess."

01:14:03 ◼   ► And Microsoft's actually thinking,

01:14:04 ◼   ► "They're never gonna do that.

01:14:05 ◼   ► It doesn't matter."

01:14:06 ◼   ► (laughing)

01:14:07 ◼   ► So they retained the right to the fantasy thing

01:14:10 ◼   ► that they wanna make,

01:14:11 ◼   ► but gave Microsoft the rights to all the other stuff.

01:14:13 ◼   ► And so Microsoft's doing all this stuff