ATP

00:00:00 ◼   ► I have a ridiculous product category

00:00:03 ◼   ► that I would like to briefly review.

00:00:05 ◼   ► - This is gonna be interesting, carry on.

00:00:08 ◼   ► - Do you remember a few, maybe six months ago,

00:00:11 ◼   ► it was a while ago, I remember exactly what it was,

00:00:13 ◼   ► it was when the new iPhones came out.

00:00:15 ◼   ► Remember they ran a little warm,

00:00:17 ◼   ► and we were discussing briefly how,

00:00:19 ◼   ► isn't it annoying when you first get a new iPhone

00:00:22 ◼   ► and it has to re-index everything

00:00:23 ◼   ► as it's being all set up and everything,

00:00:25 ◼   ► and it's indexing all your photos,

00:00:26 ◼   ► and it gets really hot, and maybe slows down

00:00:29 ◼   ► and throttles its performance as a result.

00:00:31 ◼   ► And we jokingly discovered that Razer,

00:00:35 ◼   ► the gaming PC company, they make a magnetic clip-on,

00:00:40 ◼   ► a MagSafe mounting cooling fan

00:00:43 ◼   ► with a thermoelectric element, slash Peltier,

00:00:46 ◼   ► however you pronounce those, with one of those in it

00:00:49 ◼   ► to help cool phones while gaming.

00:00:52 ◼   ► - Are you about to pull a KC?

00:00:54 ◼   ► - What does that mean?

00:00:55 ◼   ► - Talk about something that we've talked about

00:00:56 ◼   ► in past episodes already.

00:00:58 ◼   ► - Oh, well fair, that is 100% pulling a KC.

00:01:00 ◼   ► - Let's see, let's sit back and watch.

00:01:02 ◼   ► - He just said what we spoke about in the past,

00:01:04 ◼   ► so he's clear, he's acknowledged it.

00:01:06 ◼   ► - Go ahead, and also.

00:01:07 ◼   ► - Well, I thought, I have to try this.

00:01:12 ◼   ► - Why, we slowed down, why?

00:01:13 ◼   ► Why do you need to cool your phone?

00:01:16 ◼   ► - About once a year, I need that,

00:01:18 ◼   ► because when I'm setting up a new iPhone,

00:01:21 ◼   ► it's really kind of terrible, so I thought,

00:01:22 ◼   ► you know what, what the heck, let me try it,

00:01:24 ◼   ► I can talk about it on this show.

00:01:26 ◼   ► I placed an order on Amazon, like, shortly after

00:01:28 ◼   ► we recorded that episode, it was back ordered immediately

00:01:31 ◼   ► and just has never arrived, like, it never will arrive

00:01:34 ◼   ► at this point, in the intervening times,

00:01:36 ◼   ► random Amazon sellers with those vowel combination names

00:01:40 ◼   ► that, you know, they last about six months

00:01:42 ◼   ► before they disappear and make new vowel combination names,

00:01:44 ◼   ► they have come up with entries in this product category.

00:01:47 ◼   ► And at the same time, I have been facing a problem,

00:01:50 ◼   ► as I am using my iPhone 15 Pro on the dashboard of my car,

00:01:55 ◼   ► like, in the dash mount on road trips,

00:01:57 ◼   ► as it's getting warmer, it's starting to become a problem

00:02:01 ◼   ► that my phone is really getting, like, quite hot in the car

00:02:05 ◼   ► while being used this way.

00:02:06 ◼   ► - Why don't you use CarPlay?

00:02:07 ◼   ► - Mm-hmm. - Mm, you should.

00:02:09 ◼   ► You definitely should use CarPlay.

00:02:10 ◼   ► - I'm gonna say that every time you bring this.

00:02:11 ◼   ► - You should.

00:02:12 ◼   ► So anyway, so I, and it's getting so hot that, you know,

00:02:14 ◼   ► like, the screen starts dimming and the battery

00:02:17 ◼   ► can't keep up with the charge and everything,

00:02:18 ◼   ► so I'm like, this is kind of annoying.

00:02:20 ◼   ► It's also probably not good for the phone,

00:02:22 ◼   ► for its battery life and et cetera,

00:02:23 ◼   ► and for its components to be, like, super hot for hours

00:02:26 ◼   ► when I'm doing, like, long road trips.

00:02:28 ◼   ► - You should buy your car on an iPhone.

00:02:30 ◼   ► Remember when you bought your car a watch?

00:02:31 ◼   ► (laughing)

00:02:33 ◼   ► Just to justify iPhone is just for the car.

00:02:35 ◼   ► (laughing)

00:02:36 ◼   ► - Anyway, so I ordered two product entrants

00:02:41 ◼   ► in this category.

00:02:41 ◼   ► One is just, like, basically a direct copy

00:02:45 ◼   ► of the Razer thing, so it's just a fan

00:02:48 ◼   ► with a thermoelectric plate in it

00:02:50 ◼   ► to chill below ambient temperature,

00:02:53 ◼   ► and it just has a USB-C input,

00:02:55 ◼   ► and it mounts via MagSafe, so it doesn't charge the phone.

00:02:59 ◼   ► I was a little wary, like, you know,

00:03:02 ◼   ► 'cause the way thermoelectric plates work,

00:03:05 ◼   ► it's incredibly power consuming.

00:03:07 ◼   ► It's nothing like a heat pump, like an air conditioner,

00:03:09 ◼   ► where, like, you can get pretty cool temperatures

00:03:12 ◼   ► that are not that much energy.

00:03:13 ◼   ► No, like, those Peltier plates use a ton of power

00:03:17 ◼   ► to make a little bit of cooling,

00:03:18 ◼   ► and so I thought, how much could I possibly get

00:03:21 ◼   ► out of, like, a USB-powered fan for 30 bucks on Amazon?

00:03:25 ◼   ► Like, how cool could it possibly get?

00:03:27 ◼   ► Let me tell you, surprisingly cool.

00:03:30 ◼   ► It's cool to the touch immediately upon plugging it in,

00:03:33 ◼   ► and it actually cools the phone pretty well.

00:03:36 ◼   ► That being said, the one that just mounts to the back

00:03:38 ◼   ► and just like a standalone thing,

00:03:39 ◼   ► it's not really made to be used in a car.

00:03:42 ◼   ► What I wanted for the car was a MagSafe charging puck

00:03:46 ◼   ► on some kind of mount that could, like,

00:03:48 ◼   ► stick to the dash in some way

00:03:50 ◼   ► that also had cooling built in.

00:03:53 ◼   ► - Does this actually exist?

00:03:54 ◼   ► I'm assuming you're about to tell me it does.

00:03:56 ◼   ► - Of course it exists because of Amazon.

00:03:58 ◼   ► So, not only does it exist, there's like 25 of them,

00:04:01 ◼   ► all again from these weird, you know, no-name brands.

00:04:04 ◼   ► So, of course, I got two of them to try,

00:04:08 ◼   ► 'cause they, like, I had to know what would fit my car.

00:04:10 ◼   ► There's not a lot of options that will,

00:04:12 ◼   ► that have, like, a stick-on mount.

00:04:13 ◼   ► Usually they require, like, the vent mounting,

00:04:15 ◼   ► which the placement of the vents in Rivian's,

00:04:18 ◼   ► it's like, it's only down low, so there's not really

00:04:21 ◼   ► anywhere to put something like that.

00:04:22 ◼   ► Anyway, so I tried one of these,

00:04:24 ◼   ► and I actually took my last road trip with it.

00:04:26 ◼   ► I went upstate about a week ago,

00:04:29 ◼   ► and it worked fantastically.

00:04:32 ◼   ► So, let me just set expectations accordingly.

00:04:35 ◼   ► That version of it that has, like,

00:04:37 ◼   ► charging and a cooler in it, the phone does not get cool.

00:04:42 ◼   ► It is not strong enough, it just can't keep up.

00:04:44 ◼   ► That's a huge thermal load.

00:04:45 ◼   ► It does not keep up, it does not keep the phone cool.

00:04:47 ◼   ► However, it does keep the phone a lot less warm,

00:04:51 ◼   ► and so it keeps it from overheating.

00:04:54 ◼   ► It keeps the screen from dimming,

00:04:55 ◼   ► it keeps it very slowly charging,

00:04:58 ◼   ► and so it actually does serve a pretty useful function here.

00:05:02 ◼   ► Like, in the summertime, when my phone was getting so hot

00:05:05 ◼   ► in, like, regular car rides, it was, like, hot to the touch,

00:05:08 ◼   ► it actually works.

00:05:09 ◼   ► So, this ridiculous thing in my car

00:05:12 ◼   ► that was from some no-name brand,

00:05:14 ◼   ► that has a thermoelectric plate and a cooling fan

00:05:17 ◼   ► behind my phone and MagSafe charges it, actually works,

00:05:21 ◼   ► and is totally worth the whatever 35 or 40 dollars it was.

00:05:25 ◼   ► So, I gotta say, if you have this problem,

00:05:29 ◼   ► check out this category of products.

00:05:30 ◼   ► You don't think it would work, it shouldn't work,

00:05:33 ◼   ► and yet it does work.

00:05:35 ◼   ► - Yeah, you already mentioned this,

00:05:36 ◼   ► like the Rivian Vans are down low or whatever,

00:05:37 ◼   ► but the regular car version of this, like in my car,

00:05:40 ◼   ► is where my MagSafe, my charging MagSafe magnetic mount

00:05:44 ◼   ► is clipped to an air vent.

00:05:46 ◼   ► - Just turn the AC on and it blows AC air

00:05:48 ◼   ► on the back of my phone the whole time I'm driving

00:05:50 ◼   ► and my phone gets cold, literally cold, so.

00:05:52 ◼   ► - Yeah, I mean, that is the better approach

00:05:54 ◼   ► if you have that option.

00:05:55 ◼   ► - I mean, it's not, I don't know if that's,

00:05:56 ◼   ► it's not intentional, I don't actually like it

00:05:58 ◼   ► that it's blowing air on it,

00:05:59 ◼   ► because in the winter it blows hot air,

00:06:00 ◼   ► but, you know, the fact is, my phone doesn't overheat.

00:06:04 ◼   ► - Can you put a link to this, certainly in the show notes,

00:06:07 ◼   ► can you share it with us so I can laugh

00:06:08 ◼   ► at how ridiculous this is?

00:06:09 ◼   ► - Oh yeah, it's bad, I mean, 'cause it's hideous,

00:06:12 ◼   ► I mean, here, let me show you, hold on.

00:06:13 ◼   ► - So, you said, like, the Rivian Vans,

00:06:15 ◼   ► like, you can't clip anything to them, they're just too,

00:06:16 ◼   ► they're too weird, like, weird electric car-y type vents

00:06:18 ◼   ► and they're down too low?

00:06:20 ◼   ► - They are weird electric car events,

00:06:21 ◼   ► but yeah, the bigger problem is that they're mounted down,

00:06:24 ◼   ► like, at the bottom of the dashboard,

00:06:25 ◼   ► it's actually, it's not a great place for them

00:06:26 ◼   ► even just to be vents, although Hopps loves them.

00:06:30 ◼   ► All right, so I just put a link in the chat.

00:06:33 ◼   ► (laughing)

00:06:33 ◼   ► - It's like a big tongue.

00:06:35 ◼   ► - It's like a gigantic tongue. - It's like a big tongue

00:06:37 ◼   ► licking an ice hockey puck.

00:06:39 ◼   ► - So I have that, I'm using that mount

00:06:43 ◼   ► with this second one, with this charging pad,

00:06:47 ◼   ► 'cause this one was a little bit,

00:06:49 ◼   ► it had a little bit stronger cooling.

00:06:51 ◼   ► - I love that the artwork makes it look like

00:06:52 ◼   ► it's gonna make frost form on your phone,

00:06:54 ◼   ► which I think would not be good.

00:06:55 ◼   ► - No. (laughing)

00:06:57 ◼   ► - Truthfully, it's not as bad as I thought,

00:06:59 ◼   ► but it's not great.

00:07:01 ◼   ► - No, I mean, look, it's ugly, but you don't see it,

00:07:03 ◼   ► and one thing I was surprised by,

00:07:05 ◼   ► the fan really is not loud.

00:07:07 ◼   ► Like, you really don't hear the fan in a car.

00:07:10 ◼   ► - Cars are loud, like, it's gonna be drowned out

00:07:11 ◼   ► by the sound of the car. - Not his.

00:07:12 ◼   ► Yeah, electric cars aren't.

00:07:13 ◼   ► I was actually concerned about that.

00:07:15 ◼   ► - No, just like the wind noise and tire noise alone.

00:07:17 ◼   ► - Yeah, I mean, the tire noise, yeah, for sure.

00:07:19 ◼   ► Like, you do hear that, but I was very pleasantly surprised

00:07:23 ◼   ► that it actually is really not too loud.

00:07:26 ◼   ► Here, this and this one, if you just want a cooler,

00:07:30 ◼   ► not a charger, this is the other one

00:07:33 ◼   ► that actually cools it substantially better.

00:07:35 ◼   ► I think they're all kind of limited by how much power

00:07:36 ◼   ► they're going to draw from USB.

00:07:38 ◼   ► Like, I have one of those little USB power meter things.

00:07:40 ◼   ► The one, the lysin one, that one draws the most,

00:07:44 ◼   ► that's about 20 watts while charging a phone and cooling.

00:07:47 ◼   ► That seems to be the max.

00:07:48 ◼   ► Everyone else was around like 12 to 15 watts.

00:07:51 ◼   ► So, like, they could theoretically use USB power delivery

00:07:54 ◼   ► to have like even higher rates,

00:07:56 ◼   ► but for some reason they don't do that,

00:07:58 ◼   ► and I haven't looked into that further,

00:07:59 ◼   ► but like, I haven't found one that does for whatever reason.

00:08:01 ◼   ► But this, look, these are fine.

00:08:04 ◼   ► Like, they work fine.

00:08:06 ◼   ► It's a bit of a, and look, this is ridiculous.

00:08:09 ◼   ► You shouldn't need to do this.

00:08:10 ◼   ► Our phones shouldn't be overheating

00:08:12 ◼   ► when just operating in a car, but they do, and here we are.

00:08:16 ◼   ► - Well, I don't know if you can blame it on the phone.

00:08:19 ◼   ► I mean, I would blame it on the car maker

00:08:20 ◼   ► for making you use your phone in that way,

00:08:21 ◼   ► but like, because the car is like,

00:08:24 ◼   ► even when you're driving,

00:08:24 ◼   ► I know it's like air conditioned or whatever,

00:08:26 ◼   ► but it's like in direct sunlight, right?

00:08:28 ◼   ► So that's not a great-- - Pretty much.

00:08:30 ◼   ► - And it's like a little greenhouse.

00:08:31 ◼   ► It's a difficult situation.

00:08:33 ◼   ► - Yeah, and I'm using wireless charging,

00:08:34 ◼   ► which of course, like, generates some heat,

00:08:37 ◼   ► and I have a case around the phone,

00:08:39 ◼   ► which of course also basically insulates the heat

00:08:42 ◼   ► on the inside.

00:08:44 ◼   ► So there's a lot of factors working against it,

00:08:47 ◼   ► but it is kind of a ridiculous problem,

00:08:48 ◼   ► but a ridiculous problem deserves a ridiculous solution,

00:08:50 ◼   ► and I found one.

00:08:51 ◼   ► - All right, let's do some follow up.

00:08:55 ◼   ► Cywazam or something like that wrote,

00:08:58 ◼   ► with regard to the Snapdragon X10, whatever this is,

00:09:03 ◼   ► trillions of operations per second,

00:09:04 ◼   ► and that person wrote, "While 40 tops is the minimum spec

00:09:08 ◼   ► for something to be called a co-pilot plus PC,

00:09:10 ◼   ► all Snapdragon X Elite and plus chips

00:09:13 ◼   ► are actually rated at 45 trillion operations per second."

00:09:17 ◼   ► - I apologize for underselling the Snapdragon X, 45 tops.

00:09:21 ◼   ► - Excellent, I am glad to hear it.

00:09:24 ◼   ► Justin Long has come back around.

00:09:26 ◼   ► I don't know, I actually really liked him

00:09:28 ◼   ► when he was very popular like 15 years ago,

00:09:30 ◼   ► but he is the actor that you would know as the Mac

00:09:33 ◼   ► to the foil of John Hodgman's PC,

00:09:36 ◼   ► and he came back to promote Qualcomm-powered Windows PCs.

00:09:41 ◼   ► I don't even know what to make of this,

00:09:42 ◼   ► but The Verge writes, "Apple swarmer on the Mac

00:09:44 ◼   ► actor Justin Long defected to Intel a few years ago,

00:09:47 ◼   ► and now he's looking to switch

00:09:47 ◼   ► to a Qualcomm-powered Windows PC

00:09:49 ◼   ► during Qualcomm's Computex 2024 keynote today.

00:09:52 ◼   ► Long appeared in a brief 30-second skit

00:09:53 ◼   ► where he was bombarded by macOS notifications

00:09:55 ◼   ► and NAG screens causing him to start searching

00:09:57 ◼   ► for a Snapdragon-powered PC instead."

00:10:00 ◼   ► Cool.

00:10:00 ◼   ► - I'll put a link to Gruber's post about this,

00:10:03 ◼   ► which covers pretty much everything I wanted to say about it,

00:10:05 ◼   ► but yeah, the main thing that's really baffling is like,

00:10:08 ◼   ► okay, so you get the actor that was in an Apple ad campaign,

00:10:11 ◼   ► you try to ride in Apple's coattails to do a thing,

00:10:13 ◼   ► Intel has done it before,

00:10:14 ◼   ► apparently he's done this for Huawei as well, whatever.

00:10:18 ◼   ► But you have to have some motivation

00:10:20 ◼   ► for the person in the ad to be saying,

00:10:22 ◼   ► "I'm dissatisfied with Apple,

00:10:24 ◼   ► so I'm gonna get a CoPilot Plus PC," or whatever, right?

00:10:27 ◼   ► But the reason he's dissatisfied is he gets notifications,

00:10:32 ◼   ► like the email has arrived and stuff.

00:10:35 ◼   ► I mean, you can turn off notifications.

00:10:37 ◼   ► They should have asked an actual Mac user,

00:10:38 ◼   ► "What's annoying about macOS?"

00:10:40 ◼   ► We can tell you.

00:10:40 ◼   ► There are things you could put in the ad that are legit,

00:10:43 ◼   ► but what they had in the ad was like,

00:10:44 ◼   ► "Why does this make him want a new computer?"

00:10:46 ◼   ► It is just not a sensible ad,

00:10:49 ◼   ► and it shows real, just not...

00:10:52 ◼   ► Again, just ask anyone who has a Mac.

00:10:53 ◼   ► They'll tell you what's annoying about macOS.

00:10:55 ◼   ► For example, I think that people have done ads on this,

00:10:57 ◼   ► all the stupid permission things, right?

00:10:59 ◼   ► The pop-up, you know, we've been complaining about for ages

00:11:02 ◼   ► that Jason Stenell had a big post about

00:11:03 ◼   ► when he had to set up a new computer.

00:11:04 ◼   ► That's actually annoying.

00:11:06 ◼   ► Show those flying in his face and saying,

00:11:07 ◼   ► "Boy, I just tried to set up a new computer,

00:11:09 ◼   ► "and I had to give a million different permissions

00:11:10 ◼   ► "to a million different apps.

00:11:12 ◼   ► "I'm gonna get a Snapdragon X Co-Pilot Plus PC

00:11:15 ◼   ► "because then I won't have to deal with that."

00:11:16 ◼   ► I don't know if that's actually true,

00:11:17 ◼   ► but that would be a sensible ad

00:11:19 ◼   ► showing an actual pain point, but this failed to do that.

00:11:21 ◼   ► So anyway, I'm glad he's getting paid to do ads,

00:11:24 ◼   ► and it is really just a...

00:11:27 ◼   ► It's not, I don't know, embarrassing move

00:11:32 ◼   ► to be trying to play off of a very, now very old

00:11:37 ◼   ► Apple ad campaign rather than making your own ad campaign,

00:11:39 ◼   ► but you know, advertising is what it is.

00:11:42 ◼   ► - All right, so Windows Recall,

00:11:45 ◼   ► everyone seemed to be really worked up

00:11:47 ◼   ► about how this was gonna be a security nightmare,

00:11:48 ◼   ► and I'll actually admit that I assumed that, no,

00:11:53 ◼   ► they, Microsoft knows what they're doing.

00:11:55 ◼   ► This surely can't be as bad as everyone fears.

00:11:58 ◼   ► Well, it's as bad as everyone fears, if not worse.

00:12:01 ◼   ► Recall is part of the, Nith is reading from Verge,

00:12:03 ◼   ► "Recall is part of the new Co-Pilot Plus PCs

00:12:05 ◼   ► "that are debuting on June 18th,

00:12:06 ◼   ► "but experts who have tested the feature

00:12:08 ◼   ► "are already warning that Recall could be a quote-unquote

00:12:10 ◼   ► "disaster for cybersecurity."

00:12:12 ◼   ► Kevin Beaumont writes,

00:12:13 ◼   ► "Stealing everything you've ever typed or viewed

00:12:15 ◼   ► "on your own Windows PC is now possible

00:12:16 ◼   ► "with two lines of code."

00:12:18 ◼   ► Cool.

00:12:19 ◼   ► - Kevin's post has a big mock Q&A

00:12:22 ◼   ► about the way it's implemented.

00:12:23 ◼   ► So as we said in the last show,

00:12:25 ◼   ► or maybe it was the show before that,

00:12:27 ◼   ► the people who have not sort of mentally prepared

00:12:29 ◼   ► for the idea of a feature like this

00:12:31 ◼   ► by knowing about rewind or live streams or whatever,

00:12:35 ◼   ► it's just blowing their mind

00:12:36 ◼   ► that something's gonna be recording their screen

00:12:37 ◼   ► all the time and it seems really terrible

00:12:39 ◼   ► and they don't want it.

00:12:40 ◼   ► And now that people see the actual implementation,

00:12:43 ◼   ► yes, the implementation does have problems,

00:12:45 ◼   ► but a lot of the freaking out about it

00:12:47 ◼   ► is about sort of policy decisions.

00:12:50 ◼   ► So one of them is that it's on by default,

00:12:52 ◼   ► and apparently during the setup process for a new PC,

00:12:56 ◼   ► you can't opt out during the setup process.

00:12:59 ◼   ► You have to finish the setup process

00:13:00 ◼   ► and then go in settings and turn it off, right?

00:13:02 ◼   ► So you can turn it off, but it's on by default

00:13:04 ◼   ► and you have to hunt down the thing to do it,

00:13:05 ◼   ► which is mind boggling.

00:13:06 ◼   ► Everyone has, like all the enterprise admins

00:13:08 ◼   ► for Windows PC, enterprise IT people are already paranoid.

00:13:13 ◼   ► I'm sure they're slamming Microsoft saying,

00:13:15 ◼   ► "We need a way to turn this off by default

00:13:17 ◼   ► "because we don't want everything recorded on our computers

00:13:19 ◼   ► "'cause during legal discovery, during lawsuits,

00:13:22 ◼   ► "we don't wanna give a lot of information," all that stuff.

00:13:25 ◼   ► And then as you alluded to, Casey, there's the,

00:13:27 ◼   ► "Okay, but how is that actually implemented?"

00:13:29 ◼   ► Some aspects of it are unavoidable.

00:13:31 ◼   ► If you are recording everything that's on the screen

00:13:34 ◼   ► and those recordings are accessible to the logged in user,

00:13:36 ◼   ► then those recordings are accessible to the logged in user.

00:13:39 ◼   ► You know what I mean?

00:13:40 ◼   ► People want it to be recorded, but they're like,

00:13:42 ◼   ► "Yeah, but I don't want bad people to see it."

00:13:43 ◼   ► Well, guess what?

00:13:44 ◼   ► Bad people can gain access to your account as you,

00:13:47 ◼   ► and then if you can see it, they can see it.

00:13:49 ◼   ► So the mere existence of a treasure trove of recordings

00:13:52 ◼   ► of everything you've done is in itself a security problem

00:13:56 ◼   ► no matter how quote unquote secure it is.

00:13:58 ◼   ► There's no way you can make it so secure

00:14:00 ◼   ► that it can't be hacked because then the user

00:14:02 ◼   ► wouldn't be able to see it either.

00:14:04 ◼   ► Like for it to be useful, you have to be able to go back

00:14:07 ◼   ► in time and look at stuff.

00:14:08 ◼   ► If you take away that ability, you can make it real secure.

00:14:12 ◼   ► It's not a very useful feature, right?

00:14:13 ◼   ► So that I think is just, that's never gonna be fixed.

00:14:18 ◼   ► It's just people are just gonna have to live with it

00:14:20 ◼   ► and deal with it and it'll be fine.

00:14:21 ◼   ► But the second part is, okay, but you should at least

00:14:23 ◼   ► make it so that other people can't see it easily.

00:14:25 ◼   ► Like it should be encrypted on disk and stuff like that

00:14:27 ◼   ► and they didn't even do that.

00:14:28 ◼   ► It's like in a plain unencrypted SQLite database

00:14:31 ◼   ► that's accessible to, of course the logged in user,

00:14:35 ◼   ► apparently it's also accessible to any admin user

00:14:37 ◼   ► on the system.

00:14:37 ◼   ► It's just a plain text file.

00:14:38 ◼   ► Like they didn't even encrypt it at rest.

00:14:40 ◼   ► So implementation of this does not look great

00:14:43 ◼   ► and turning it on by default does not seem great.

00:14:45 ◼   ► And it's basically giving the Co-Pilot Plus PCs

00:14:47 ◼   ► kind of a marketing black eye before they have a chance

00:14:51 ◼   ► to impress people with all the good things about them.

00:14:53 ◼   ► You know what I mean?

00:14:54 ◼   ► Like this is not the fault of the Snapdragon X processor.

00:14:57 ◼   ► This is not the fault of the hardware.

00:14:59 ◼   ► It's the fault of Windows, including this feature

00:15:01 ◼   ► and turning it on by default

00:15:02 ◼   ► and apparently implementing it poorly.

00:15:03 ◼   ► So I don't know, Microsoft, they've got a lot of work.

00:15:06 ◼   ► But by the time you hear this episode,

00:15:08 ◼   ► presumably Microsoft will have issued an apology

00:15:09 ◼   ► and said they're not gonna have it on by default

00:15:11 ◼   ► and yada yada, but right now things don't look good.

00:15:13 ◼   ► - Yeah, I mean, I assume they will have it off by default

00:15:16 ◼   ► before it's released because there's way too many problems.

00:15:19 ◼   ► And we talked about all this when Rewind AI came out.

00:15:23 ◼   ► You know, whenever we talked about that like months ago.

00:15:25 ◼   ► And then again on our overtime a couple weeks ago.

00:15:27 ◼   ► You know, as you said, there's really no good way

00:15:31 ◼   ► to have this data be like, you know,

00:15:35 ◼   ► quote only accessible to good reasons

00:15:38 ◼   ► or good people or whatever.

00:15:39 ◼   ► I think the ultimately, as long as this data

00:15:43 ◼   ► is being collected somewhere on your computer

00:15:45 ◼   ► or on some service, like whatever it is,

00:15:47 ◼   ► like even if you have it all local, all encrypted,

00:15:50 ◼   ► it's still being collected.

00:15:52 ◼   ► And there will be ways to exploit that.

00:15:55 ◼   ► People will find security holes,

00:15:56 ◼   ► malware will try to access it,

00:15:58 ◼   ► some malware will succeed in accessing it.

00:16:01 ◼   ► There are also things like legal discovery risks

00:16:03 ◼   ► and things like that to, you know,

00:16:05 ◼   ► that will make a lot of companies not wanna do it.

00:16:06 ◼   ► - And just plain old social engineering,

00:16:07 ◼   ► because again, the user can access it.

00:16:09 ◼   ► So if you can convince the user to do something,

00:16:11 ◼   ► like that's most of hacking and malware

00:16:13 ◼   ► is based on tricking people into doing something, right?

00:16:16 ◼   ► And there's, again, there's no way to collect this

00:16:19 ◼   ► and have it be totally safe.

00:16:21 ◼   ► If the user has access to it, the user is the weakest link.

00:16:24 ◼   ► - Yeah, and there are so many practical and privacy concerns

00:16:29 ◼   ► and just legal concerns with this.

00:16:32 ◼   ► How are they getting around things like wiretapping laws?

00:16:35 ◼   ► There's so many problems with these kind of approaches.

00:16:39 ◼   ► I think it's interesting that these tools exist.

00:16:42 ◼   ► They make for fantastic demos

00:16:44 ◼   ► and some people will use them and will love them.

00:16:49 ◼   ► - It is so far from being appropriate to be on by default.

00:16:53 ◼   ► Like, we are nowhere, it is nowhere near

00:16:57 ◼   ► that universally good that it should be on by default.

00:17:00 ◼   ► This should be something that it's fine for companies

00:17:02 ◼   ► like Rewind to have products that do this.

00:17:05 ◼   ► It's interesting that Microsoft's doing it

00:17:06 ◼   ► at the OS level now.

00:17:07 ◼   ► It's wonderful to have it as an option

00:17:10 ◼   ► for those like power productivity users who know about it,

00:17:14 ◼   ► who know exactly what it's doing and who will opt into it.

00:17:17 ◼   ► It has to be only for them though.

00:17:19 ◼   ► It cannot be everyone opted in by default.

00:17:22 ◼   ► That is just irresponsible.

00:17:25 ◼   ► - Yeah, not good, Bob, but here we are.

00:17:28 ◼   ► AMD's next generation of AI laptop processors

00:17:31 ◼   ► have been announced.

00:17:32 ◼   ► Reading from The Verge, AMD announced at Computex 2024

00:17:35 ◼   ► that it's next generation of Ryzen laptop processors

00:17:38 ◼   ► for generative AI workloads.

00:17:40 ◼   ► The Ryzen AI 300 series.

00:17:42 ◼   ► It's a rebrand of its top tier Ryzen 9 chips.

00:17:44 ◼   ► The new Ryzen AI chips are built

00:17:45 ◼   ► on AMD's latest architectures for neural,

00:17:47 ◼   ► integrated graphics and general processing.

00:17:50 ◼   ► The first two processors in the series

00:17:51 ◼   ► are the Ryzen AI 9HX370, just rolls right off the tongue,

00:17:55 ◼   ► and the similarly eloquent Ryzen AI 9365.

00:17:59 ◼   ► Both have 50 trillion operations per second on their NPUs,

00:18:03 ◼   ► but the HX variant is the higher end of the two.

00:18:06 ◼   ► You know that because it says HX.

00:18:08 ◼   ► - So we mentioned last episode

00:18:10 ◼   ► or whenever we talked about the Copilot Plus PCs

00:18:11 ◼   ► that AMD and Intel also have processors

00:18:15 ◼   ► that they're gonna be rolling out

00:18:16 ◼   ► that are going to qualify as Copilot Plus PCs.

00:18:18 ◼   ► It's not just an ARM thing, so they're behind,

00:18:21 ◼   ► but they're catching up.

00:18:22 ◼   ► And look, there's AI right in the processor name.

00:18:24 ◼   ► I'm sure they'll never regret that branding

00:18:26 ◼   ► and it won't look dated when we look back on it.

00:18:28 ◼   ► (laughing)

00:18:29 ◼   ► But 50 tops, hey, better than 45, right?

00:18:31 ◼   ► So good job, AMD.

00:18:32 ◼   ► - Hooray.

00:18:33 ◼   ► All right, Intel has also detailed its new lunar lake CPUs

00:18:38 ◼   ► that will go up against the aforementioned AMD,

00:18:40 ◼   ► Qualcomm and Apple, reading this time from Ars Technica,

00:18:42 ◼   ► "Lunar Lake will be Intel's first processor

00:18:44 ◼   ► "with a neural processing unit or NPU

00:18:46 ◼   ► "that meets Microsoft's Copilot Plus PC requirements.

00:18:49 ◼   ► "Intel rates lunar lakes NPU raw performance at 48 tops.

00:18:52 ◼   ► "Lunar Lake has two functional tiles.

00:18:55 ◼   ► "The compute tile combines all of the processors,

00:18:57 ◼   ► "performance and efficiency cores, the GPU, the NPU,

00:19:00 ◼   ► "the display outputs, and the media encoding

00:19:02 ◼   ► "and decoding engine.

00:19:03 ◼   ► "And the platform controller tile handles wired

00:19:06 ◼   ► "and wireless connectivity, including PCIe

00:19:08 ◼   ► "and USB, Thunderbolt 4, and Wi-Fi 7 and Bluetooth 5.4.

00:19:12 ◼   ► "Another big packaging change is that Intel

00:19:13 ◼   ► "is integrating RAM into the CPU package."

00:19:15 ◼   ► Well, just like Apple.

00:19:16 ◼   ► - Yeah, so does this sound familiar?

00:19:18 ◼   ► This is, you know, Apple sort of led the charge here

00:19:21 ◼   ► in the mass market, in the market for mass market PCs, right?

00:19:25 ◼   ► They made their system on a chip,

00:19:26 ◼   ► it's got the RAM in the same package,

00:19:28 ◼   ► it's got the whole thing in a giant SOC.

00:19:30 ◼   ► Here's Intel finally catching up with that philosophy.

00:19:33 ◼   ► I think the AMD one is similar with the chiplet type thing,

00:19:36 ◼   ► but this is, I really feel like this is Apple leading.

00:19:41 ◼   ► And people are gonna say, well, Apple isn't the first one

00:19:42 ◼   ► to do that, lots of people have done it before,

00:19:43 ◼   ► so on and so forth.

00:19:44 ◼   ► But Apple showed that it can perform very well

00:19:47 ◼   ► in personal computers that they sell millions of

00:19:50 ◼   ► to regular people.

00:19:51 ◼   ► Put it all on one chip, put the GPU on there,

00:19:54 ◼   ► put the neuro-processing, the media encoder engine,

00:19:56 ◼   ► everything, and so here's Intel, basically many years later

00:20:00 ◼   ► saying, let's do that.

00:20:01 ◼   ► Now they have things separated,

00:20:03 ◼   ► I think we've talked about this before,

00:20:03 ◼   ► I think they're using like different processes

00:20:05 ◼   ► for different parts of the chip.

00:20:06 ◼   ► I think the connectivity thing is not built

00:20:10 ◼   ► on the same three nanometer processor,

00:20:12 ◼   ► whatever I think they might be.

00:20:13 ◼   ► I forget what the details are,

00:20:14 ◼   ► if they're using TSMC to fab this or whatever,

00:20:16 ◼   ► but so they're taking a slightly different approach to Apple

00:20:18 ◼   ► because Apple can afford to just put it on all one thing,

00:20:20 ◼   ► which has some efficiency benefits,

00:20:23 ◼   ► but of course it is also more costly.

00:20:25 ◼   ► But it's really kind of weird to see the PC world

00:20:29 ◼   ► essentially following Apple's lead and deciding,

00:20:32 ◼   ► you know what, that thing Apple decided to do with the M1,

00:20:35 ◼   ► we should make all our PCs like that too.

00:20:38 ◼   ► Snapdragon, AMD, Intel,

00:20:41 ◼   ► all seem to be moving in that direction,

00:20:43 ◼   ► at least for like the low end and medium end

00:20:46 ◼   ► laptop type things.

00:20:47 ◼   ► When it comes to big desktop PC and gaming PCs,

00:20:50 ◼   ► they still have just giant discrete GPUs

00:20:52 ◼   ► that use a ton of power and giant chips that aren't SOCs.

00:20:56 ◼   ► So in that respect,

00:20:57 ◼   ► they're not following Apple down that path,

00:20:59 ◼   ► but boy, even if I predicted that,

00:21:03 ◼   ► you know, when the M1 came out,

00:21:04 ◼   ► by the time Apple has the M4,

00:21:06 ◼   ► Intel, AMD and the ARM things will be following.

00:21:09 ◼   ► I feel like this is unexpected.

00:21:10 ◼   ► I would have been more likely to think,

00:21:13 ◼   ► oh, maybe the whole PC world will go ARM

00:21:15 ◼   ► as discussed in past episode.

00:21:18 ◼   ► Doesn't look like they're all going ARM,

00:21:20 ◼   ► especially since the Co-Pilot Plus PC thing

00:21:22 ◼   ► isn't firing off like gangbusters,

00:21:24 ◼   ► but I'm still holding out hope.

00:21:25 ◼   ► And anyway, in the meantime,

00:21:26 ◼   ► Intel and AMD are in the game too.

00:21:30 ◼   ► - Speaking of Apple and M chips,

00:21:32 ◼   ► Apple says the M2 iPad Air,

00:21:35 ◼   ► no, it doesn't have 10 GPU cores.

00:21:37 ◼   ► We actually meant nine.

00:21:40 ◼   ► Cool, cool, we're cool, right?

00:21:42 ◼   ► Reading from MacRumors,

00:21:43 ◼   ► Apple has seemingly updated its tech specs webpage

00:21:45 ◼   ► for the latest M2 iPad Air models

00:21:47 ◼   ► to indicate the M2 chip has a nine core GPU

00:21:49 ◼   ► rather than 10 cores as previously stated.

00:21:52 ◼   ► The change was first reported by a nine to five Mac

00:21:53 ◼   ► based on available web snapshots.

00:21:55 ◼   ► The update to Apple's US website

00:21:56 ◼   ► was made within the last 10 days.

00:21:58 ◼   ► This was written two days ago as we record.

00:22:00 ◼   ► However, the equivalent webpage

00:22:02 ◼   ► on many of Apple's regional stores

00:22:04 ◼   ► still lists the M2 iPad Air as having a 10 core GPU.

00:22:07 ◼   ► Eventually a statement was provided to nine to five Mac

00:22:10 ◼   ► and Apple said that the details it shared

00:22:13 ◼   ► with the iPad Air's performance

00:22:15 ◼   ► were always based on a nine core GPU.

00:22:17 ◼   ► So Apple said, and I'm quoting,

00:22:18 ◼   ► "We are updating apple.com to correct the core count

00:22:20 ◼   ► for the M2 iPad Air.

00:22:21 ◼   ► All the performance claims for the M2 iPad Air are accurate

00:22:24 ◼   ► and based on a nine core GPU."

00:22:26 ◼   ► Whoopsie-dipsies.

00:22:27 ◼   ► - What a weird flub.

00:22:28 ◼   ► Like, I mean, it doesn't even seem like

00:22:30 ◼   ► it's a last minute decision.

00:22:32 ◼   ► This is something that had to have been decided long ago,

00:22:34 ◼   ► but like maybe there was a miscommunication.

00:22:36 ◼   ► Like they decided to go with the nine core

00:22:37 ◼   ► instead of the 10 core to save money on the iPad Air

00:22:39 ◼   ► and just the web team didn't hear about it

00:22:42 ◼   ► even though that decision was made four months ago.

00:22:44 ◼   ► Like, it's not like they started manufacturing the iPad Air

00:22:46 ◼   ► on the day of the announcement or something.

00:22:48 ◼   ► Very strange.

00:22:49 ◼   ► Not a big deal for the iPad Air.

00:22:50 ◼   ► That's fine, but I don't know what's going on over there.

00:22:53 ◼   ► - We are brought to you this episode by Squarespace,

00:22:57 ◼   ► the all-in-one website platform for entrepreneurs

00:23:01 ◼   ► to stand out and succeed online.

00:23:03 ◼   ► Whether you're just starting out

00:23:04 ◼   ► or managing a growing brand,

00:23:05 ◼   ► Squarespace makes it super easy for you

00:23:06 ◼   ► to make a beautiful website,

00:23:08 ◼   ► engage with your audience, and sell what you make,

00:23:10 ◼   ► whether it's products or content or even your time,

00:23:13 ◼   ► all in one place and all on your terms.

00:23:16 ◼   ► Squarespace makes everything super easy.

00:23:18 ◼   ► You don't have to be a nerd to use this.

00:23:21 ◼   ► So that's great.

00:23:22 ◼   ► Like not only you as a nerd listening to the show,

00:23:24 ◼   ► you don't have to like waste your time

00:23:25 ◼   ► dealing with like installing packages on servers

00:23:27 ◼   ► and stuff, but also if you're recommending it

00:23:29 ◼   ► to someone else in your life, they can do it themselves.

00:23:32 ◼   ► It's super empowering.

00:23:33 ◼   ► They don't need to go through a gatekeeper nerd like us

00:23:36 ◼   ► to do their own website.

00:23:37 ◼   ► They can do it themselves.

00:23:38 ◼   ► And Squarespace supports everything you might need,

00:23:41 ◼   ► especially for business sites.

00:23:43 ◼   ► You can of course sell products, digital or physical.

00:23:45 ◼   ► They have all sorts of features, tax handling,

00:23:48 ◼   ► shipping handling, inventory, all this stuff.

00:23:50 ◼   ► You can also sell content.

00:23:52 ◼   ► Suppose maybe you're selling exclusive content

00:23:54 ◼   ► by having a paywall to sell memberships or courses,

00:23:57 ◼   ► or you can sell files to download

00:23:58 ◼   ► like PDFs or music or ebooks.

00:24:01 ◼   ► Squarespace also now has an AI feature.

00:24:03 ◼   ► You can kickstart or update written content

00:24:05 ◼   ► on any website, product description or email

00:24:07 ◼   ► where Squarespace AI generating instant personalized results

00:24:10 ◼   ► that know and show your brand identity.

00:24:12 ◼   ► So you can explain what your site is about,

00:24:14 ◼   ► choose your tone, enter what you need

00:24:16 ◼   ► to get short or long form text.

00:24:18 ◼   ► Squarespace AI makes it easier to go live,

00:24:20 ◼   ► stand out and succeed online.

00:24:22 ◼   ► And of course Squarespace has so many other features.

00:24:24 ◼   ► I can't fit them all in one ad read.

00:24:25 ◼   ► Anything you might want from a business site

00:24:27 ◼   ► or a personal site, Squarespace probably supports it.

00:24:29 ◼   ► And they have this new fluid engine,

00:24:30 ◼   ► makes everything super easy.

00:24:32 ◼   ► It's great, go check it out.

00:24:33 ◼   ► See for yourself, start a free trial.

00:24:36 ◼   ► Squarespace.com, build your site in trial mode.

00:24:38 ◼   ► When you're ready to launch, go to Squarespace.com/ATP

00:24:41 ◼   ► to save 10% off your first purchase of a website or domain.

00:24:44 ◼   ► So once again, start that free trial at signup,

00:24:47 ◼   ► Squarespace.com/ATP for 10% off your first purchase

00:24:51 ◼   ► of a website or domain.

00:24:52 ◼   ► Thank you so much to Squarespace for sponsoring our show.

00:24:55 ◼   ► (upbeat music)

00:24:58 ◼   ► Let's talk about LLMs and how they can learn.

00:25:02 ◼   ► Adeiko writes, "There's some quote unquote new tech

00:25:05 ◼   ► "that aims to enable LLMs to learn in context,

00:25:08 ◼   ► "in production, avoiding limitations of relying solely

00:25:10 ◼   ► "on RAG or retrieval augmented generation.

00:25:14 ◼   ► "While promising this approach

00:25:15 ◼   ► "is significantly more expensive,

00:25:16 ◼   ► "here's a paper explaining,"

00:25:17 ◼   ► and we'll put it in the show notes,

00:25:18 ◼   ► "the methodology from earlier this year

00:25:19 ◼   ► "and a video discussing the current limitations

00:25:21 ◼   ► "you mentioned on the episode

00:25:22 ◼   ► "and a potential solution that is clearly explained."

00:25:25 ◼   ► So we'll put all three of these links in the show notes.

00:25:27 ◼   ► John, care to summarize and/or dissect for us?

00:25:30 ◼   ► - Yeah, so there's one video that says,

00:25:31 ◼   ► "What is RAG?

00:25:32 ◼   ► "What is retrieval augmented generation?"

00:25:34 ◼   ► And my summary of it would be,

00:25:37 ◼   ► you take what the person typed into their chat bot

00:25:39 ◼   ► and you stick a bunch of text in front of it.

00:25:42 ◼   ► And then you send that to the chat bot

00:25:44 ◼   ► and picking the text you stick in front of it,

00:25:47 ◼   ► I mean, they're not actually doing this,

00:25:49 ◼   ► but you can imagine if they ask some question about cars,

00:25:52 ◼   ► you would do a Google search based on what they asked

00:25:55 ◼   ► and take the results of that Google search

00:25:56 ◼   ► and stick it to the front of their text

00:25:58 ◼   ► and then take that whole wad of text

00:26:00 ◼   ► and throw it through the LLM, right?

00:26:02 ◼   ► It's taking more than what you said,

00:26:04 ◼   ► but trying to augment it with some context.

00:26:06 ◼   ► Okay, LLM, maybe you weren't trained

00:26:08 ◼   ► on any data about cars,

00:26:10 ◼   ► but I'm gonna pull the latest up-to-date information

00:26:12 ◼   ► about the car they mentioned in their prompt,

00:26:14 ◼   ► and I'm gonna put that in front of their prompt,

00:26:16 ◼   ► and then I'm gonna shove all that text, their question,

00:26:18 ◼   ► plus all the text I grabbed about the latest cars,

00:26:20 ◼   ► and I'm gonna chuck that into the LLM,

00:26:21 ◼   ► because the LLMs are like,

00:26:22 ◼   ► oh, I don't know anything after 2021, right?

00:26:25 ◼   ► Like the old versions of chat DPT would say that

00:26:27 ◼   ► because they were trained on older data.

00:26:29 ◼   ► Well, what if someone's asking about something

00:26:30 ◼   ► that's more recent that the LLM could not have been trained

00:26:32 ◼   ► on or maybe just wasn't trained on

00:26:33 ◼   ► because it's really narrow interest?

00:26:36 ◼   ► That's what RAG is, right?

00:26:37 ◼   ► But of course that has lots of limitations.

00:26:39 ◼   ► I will put the second video says,

00:26:41 ◼   ► what are the limitations of RAG?

00:26:42 ◼   ► And you learn all about that.

00:26:43 ◼   ► And then the final thing is continual learning

00:26:45 ◼   ► for large language models, the paper,

00:26:48 ◼   ► the scientific paper that we'll link,

00:26:50 ◼   ► and this is the thing from the abstract,

00:26:52 ◼   ► it says large language models are not amenable

00:26:53 ◼   ► to frequent retraining due to high training costs

00:26:56 ◼   ► arising from their massive scale.

00:26:57 ◼   ► However, updates are necessary to endow LLMs

00:26:59 ◼   ► with new skills and keep them up to date

00:27:01 ◼   ► with rapidly evolving human knowledge.

00:27:02 ◼   ► This paper surveys recent work

00:27:04 ◼   ► on continual learning for LLMs, right?

00:27:06 ◼   ► So the idea is, we said before,

00:27:07 ◼   ► when you throw things through an LLM,

00:27:09 ◼   ► the weights, the little numbers inside the LLM

00:27:12 ◼   ► don't change, right?

00:27:13 ◼   ► They take input, they produce output,

00:27:14 ◼   ► but the inside of that LLM never changes.

00:27:16 ◼   ► So you can't teach it anything.

00:27:17 ◼   ► All you can do is make your prompt bigger and bigger

00:27:19 ◼   ► and hope that the information in your prompt

00:27:22 ◼   ► influences the answer.

00:27:23 ◼   ► And it does influence the answer,

00:27:24 ◼   ► but not as much as training the LLM

00:27:26 ◼   ► on a whole bunch of that information, you know?

00:27:28 ◼   ► Anyway, so that's RAG.

00:27:30 ◼   ► - Right, so then Steven Tierney writes

00:27:33 ◼   ► that Sundar Pichai's infinite context window claim

00:27:36 ◼   ► from Google I/O seems to stem from this paper,

00:27:39 ◼   ► which is entitled Leave No Context Behind,

00:27:41 ◼   ► Efficient Infinite Context Transformers

00:27:42 ◼   ► with Infinite Attention.

00:27:44 ◼   ► This work introduces an efficient method

00:27:47 ◼   ► to scale transformer-based large language models

00:27:49 ◼   ► to infinitely long inputs with bounded memory and computation

00:27:52 ◼   ► A key component in our proposed approach

00:27:54 ◼   ► is a new attention technique

00:27:56 ◼   ► that incorporates a compressive memory

00:27:58 ◼   ► into the vanilla attention mechanism.

00:28:00 ◼   ► The attention mechanism in Transformers

00:28:01 ◼   ► exhibits quadratic complexity

00:28:03 ◼   ► in both memory footprint and computation time.

00:28:05 ◼   ► Compressive memory systems promise to be more scalable

00:28:07 ◼   ► and efficient for extremely long sequences.

00:28:10 ◼   ► Instead of using an array

00:28:11 ◼   ► that grows with the input sequence length,

00:28:13 ◼   ► a compressive memory primarily maintains

00:28:15 ◼   ► a fixed number of parameters to store and recall information

00:28:17 ◼   ► with a bounded storage and computation cost.

00:28:20 ◼   ► In the compressive memory,

00:28:21 ◼   ► new information is added to the memory

00:28:23 ◼   ► by changing its parameters with an objective

00:28:24 ◼   ► that this information can be recovered back later on.

00:28:26 ◼   ► John, translate that for me, please.

00:28:28 ◼   ► - So the scaling thing for people

00:28:30 ◼   ► who ever took computer science class with big O notation

00:28:32 ◼   ► is like how does this algorithm scale

00:28:34 ◼   ► with the size of the input?

00:28:35 ◼   ► For example, linear scaling would be,

00:28:37 ◼   ► if you asked me to process 20 of these things,

00:28:39 ◼   ► I'd take twice as long as if you asked me to process 10.

00:28:42 ◼   ► Right, that's linear scaling.

00:28:43 ◼   ► You can have other kinds of scaling

00:28:45 ◼   ► where it scales to the square of the input

00:28:47 ◼   ► and that's worse, right?

00:28:49 ◼   ► And quadratic, that's what they're getting at there.

00:28:53 ◼   ► The point is if you make longer and longer inputs

00:28:55 ◼   ► to the LLM, it just, the time goes like a hockey stick graph

00:29:00 ◼   ► it gets really, really bad.

00:29:01 ◼   ► It doesn't even scale linearly.

00:29:02 ◼   ► And even linear would be bad

00:29:03 ◼   ► because Sundar Pichai was saying,

00:29:06 ◼   ► you're gonna have infinite context.

00:29:07 ◼   ► We have a limitation now of like 2 million tokens

00:29:10 ◼   ► and that's pretty good.

00:29:10 ◼   ► You can put a large wad of text through our LLM.

00:29:13 ◼   ► Pretty big, right?

00:29:14 ◼   ► But you can't, like I was mentioning,

00:29:16 ◼   ► the infinite context opened the door to the idea of like,

00:29:18 ◼   ► can you actually teach an LLM?

00:29:20 ◼   ► Well, if your context is infinite,

00:29:22 ◼   ► you can just keep typing stuff until the LLM gets it.

00:29:24 ◼   ► No LLM, let me tell you more about this thing.

00:29:27 ◼   ► And let me tell you more.

00:29:28 ◼   ► Like essentially, you're not training it,

00:29:29 ◼   ► you're not changing the weights,

00:29:30 ◼   ► but it's like anytime I type anything,

00:29:33 ◼   ► the last billion characters I've written to it

00:29:35 ◼   ► over the past five years get sent through the LLM

00:29:37 ◼   ► plus the question I just asked stuck at the end.

00:29:40 ◼   ► And that's, as you can imagine, incredibly inefficient

00:29:42 ◼   ► because of the way it scales.

00:29:43 ◼   ► So this infinite context thing is like,

00:29:45 ◼   ► well, we're not gonna do that.

00:29:46 ◼   ► Instead, we're gonna use this thing where we kind of like,

00:29:48 ◼   ► we try to do with a fixed amount of memory and computation.

00:29:51 ◼   ► I don't understand how they're doing it,

00:29:53 ◼   ► but it's like, I feel like they have some sort of block

00:29:55 ◼   ► that the input goes into

00:29:56 ◼   ► and then it modifies the values in that block

00:30:00 ◼   ► and it just keeps doing that over and over again.

00:30:01 ◼   ► And finally, it gets the final block

00:30:02 ◼   ► and sends that through the LLM

00:30:04 ◼   ► after it's been modified by all the input.

00:30:06 ◼   ► But it's much less compelling.

00:30:08 ◼   ► Like many people who wrote in about this is like,

00:30:09 ◼   ► yeah, it's not really infinite context.

00:30:11 ◼   ► It's more like taking an infinite amount of input

00:30:14 ◼   ► and squishing it down to a small finite size

00:30:16 ◼   ► and sending that through the LLM

00:30:17 ◼   ► and hoping you've preserved enough information

00:30:20 ◼   ► to get a useful answer out of it.

00:30:22 ◼   ► And by the way, the other thing,

00:30:23 ◼   ► the previous link about the continual learning

00:30:27 ◼   ► is that some people are looking into the idea

00:30:29 ◼   ► of what if we did change the weights

00:30:31 ◼   ► when you sent a question or something through?

00:30:33 ◼   ► If you talk to an LLM and you try to tell it something,

00:30:36 ◼   ► can we retrain based on what you put in

00:30:39 ◼   ► and change the weights of the model?

00:30:40 ◼   ► And that's an area of research or whatever.

00:30:42 ◼   ► And as I mentioned, we talked about this earlier,

00:30:45 ◼   ► that opens the door for people to screw up the LLMs

00:30:48 ◼   ► and make them terrible.

00:30:49 ◼   ► You could tell it a whole bunch of bogus stuff

00:30:51 ◼   ► and now your LLM just gets dumber and worse

00:30:53 ◼   ► and tells more lies.

00:30:55 ◼   ► It's like children.

00:30:56 ◼   ► You can teach them whatever you want.

00:30:58 ◼   ► You hope you teach them good things

00:30:59 ◼   ► and teach them how to read and write

00:31:01 ◼   ► and do math correctly and all

00:31:04 ◼   ► and be a nice polite person or whatever.

00:31:06 ◼   ► And what you will get out of all that effort

00:31:08 ◼   ► is hopefully a good adult.

00:31:11 ◼   ► But if you are terrible to them and do terrible things

00:31:13 ◼   ► and tell them lies, the results will be worse.

00:31:17 ◼   ► So the ability of, as you can imagine,

00:31:20 ◼   ► if you allowed anyone on the internet

00:31:22 ◼   ► to type anything they want

00:31:23 ◼   ► that would actually change your LLM,

00:31:26 ◼   ► that would be bad instantly.

00:31:27 ◼   ► We already saw that with the non-trainable LLMs.

00:31:31 ◼   ► Remember when Microsoft put their chatbot up

00:31:32 ◼   ► and it was spewing Nazi stuff?

00:31:34 ◼   ► That's just because it's trained on the internet

00:31:36 ◼   ► and it's filled with all sorts of terrible things.

00:31:37 ◼   ► So they're trying to tamp that down.

00:31:38 ◼   ► But this would be like, no, it will be retrained

00:31:41 ◼   ► based on the things you tell it.

00:31:43 ◼   ► Nightmare fuel.

00:31:44 ◼   ► But of course, if you had your own independent LLM

00:31:46 ◼   ► and you were sensible, then you could mess up yours

00:31:49 ◼   ► and other people would have their own

00:31:50 ◼   ► and they could mess up theirs.

00:31:51 ◼   ► And it's an interesting idea.

00:31:53 ◼   ► But the infinite contact window was like,

00:31:54 ◼   ► no, the LLM doesn't change until we release a new version.

00:31:57 ◼   ► The weights are unchangeable,

00:31:59 ◼   ► but we'll allow you to type an infinite amount

00:32:01 ◼   ► of information in your prompt

00:32:02 ◼   ► and maybe that'll make it do something sensible.

00:32:05 ◼   ► None of these approaches seem super great to me,

00:32:08 ◼   ► but I'm glad people are trying to figure out

00:32:10 ◼   ► how to get around the current limitations

00:32:12 ◼   ► because the current limitations are crippling

00:32:15 ◼   ► for some use cases.

00:32:15 ◼   ► Like they just basically make it impossible

00:32:18 ◼   ► to do certain things reliably.

00:32:20 ◼   ► It doesn't mean they're not useful.

00:32:21 ◼   ► They're useful for other things,

00:32:22 ◼   ► but certain use cases, it becomes completely useless.

00:32:25 ◼   ► So I will look forward to this research advancing,

00:32:28 ◼   ► but when it's appearing in scientific papers and stuff,

00:32:31 ◼   ► it makes me think it's many years

00:32:32 ◼   ► from being practically implemented, but we'll see.

00:32:35 ◼   ► - All right, as we sit here,

00:32:37 ◼   ► it is the evening of Wednesday the 5th.

00:32:41 ◼   ► And in just a few short days,

00:32:43 ◼   ► we will all be flying west to these greater San Jose area

00:32:47 ◼   ► in order to go to WWDC,

00:32:50 ◼   ► Apple's annual worldwide developer conference.

00:32:52 ◼   ► As we mentioned last episode,

00:32:53 ◼   ► all three of us will be there

00:32:54 ◼   ► barring any plane catastrophes.

00:32:56 ◼   ► And I know all three of us are super excited.

00:32:58 ◼   ► - What do you have to say?

00:32:59 ◼   ► (laughing)

00:33:00 ◼   ► - You know what?

00:33:01 ◼   ► You never know what'll happen.

00:33:02 ◼   ► So I just mean like a delay for the record.

00:33:05 ◼   ► I didn't mean--

00:33:06 ◼   ► - That's not what it sounded like.

00:33:07 ◼   ► - Please, please.

00:33:07 ◼   ► - Sorry.

00:33:08 ◼   ► It's not getting any better.

00:33:09 ◼   ► Don't worry, we'll fix it in post.

00:33:11 ◼   ► So yeah, WWDC is coming up.

00:33:13 ◼   ► So we should probably talk about what to expect.

00:33:16 ◼   ► And honestly, I'm very surprised,

00:33:20 ◼   ► well maybe not surprised,

00:33:20 ◼   ► but impressed with the lack of like concrete leaks.

00:33:23 ◼   ► And let me go through a few things here,

00:33:25 ◼   ► unless one of you has an opening statement.

00:33:26 ◼   ► Let me go through a few things

00:33:27 ◼   ► that we haven't really talked about

00:33:28 ◼   ► that we can use as jumping off points.

00:33:30 ◼   ► But by and large, I feel like it's pretty,

00:33:34 ◼   ► pretty well under lock and key

00:33:35 ◼   ► what we're gonna see on Monday.

00:33:37 ◼   ► - Why do you think that?

00:33:38 ◼   ► It's just because the things we've talked about

00:33:39 ◼   ► have been rumored.

00:33:40 ◼   ► I believe most of the things will come to pass,

00:33:42 ◼   ► and I think that's what they're gonna show.

00:33:44 ◼   ► But we'll say, we'll go through the items we have here.

00:33:45 ◼   ► We've talked about a lot of things on past episodes

00:33:48 ◼   ► that have been rumored for WWDC,

00:33:50 ◼   ► and I don't think many of them have been debunked.

00:33:52 ◼   ► They're just sitting out there waiting for us,

00:33:54 ◼   ► waiting to see if they're true or not.

00:33:56 ◼   ► - Yeah, so we'll start.

00:33:58 ◼   ► Mark Ermin has said, "Read my lips, no new hardware."

00:34:01 ◼   ► And so there's apparently not gonna be any hardware

00:34:04 ◼   ► this year, which I don't think is particularly surprising,

00:34:06 ◼   ► but Mark has stated it is definitely not happening.

00:34:09 ◼   ► - Yeah, that sometimes happens.

00:34:10 ◼   ► We have the no hardware WWDCs.

00:34:12 ◼   ► Sometimes Apple itself kind of leaks it

00:34:13 ◼   ► to control expectations.

00:34:15 ◼   ► In this case, I don't think there were any

00:34:17 ◼   ► particular expectations that Apple needed to leak

00:34:19 ◼   ► to tamp down, other than my unrealistic expectations

00:34:23 ◼   ► about the Mac Pro and the Mac Studio,

00:34:24 ◼   ► but there's cold water in that.

00:34:26 ◼   ► - My expectations about Siri not sucking.

00:34:29 ◼   ► - Well, yeah, but no hardware.

00:34:30 ◼   ► No hardware, WWDCs.

00:34:31 ◼   ► And we just had a bunch of hardware.

00:34:34 ◼   ► The only hardware that Apple should be updating right now

00:34:36 ◼   ► are the ones they apparently can't update

00:34:38 ◼   ► because the chips aren't ready,

00:34:39 ◼   ► and that would be the Mac Studio and the Mac Pro.

00:34:42 ◼   ► I mean, he was pretty definitive,

00:34:44 ◼   ► and when he comes out this close to WWDC

00:34:46 ◼   ► and says no hardware, I just have to accept

00:34:47 ◼   ► there's gonna be no hardware.

00:34:48 ◼   ► So I'm sad about it, but perhaps not surprising.

00:34:51 ◼   ► - You know, let me jump in and go a little off script here.

00:34:54 ◼   ► Do you think, Jon, that this is the year,

00:34:57 ◼   ► that the forthcoming macOS release,

00:34:59 ◼   ► whatever it may be called,

00:35:01 ◼   ► do you think that's the year that they start

00:35:03 ◼   ► really hammering down, or taking away stuff for Intel?

00:35:07 ◼   ► 'Cause you weren't missing out on much

00:35:09 ◼   ► on your ancient-ass Mac Pro, right?

00:35:11 ◼   ► - Performance.

00:35:12 ◼   ► (laughing)

00:35:13 ◼   ► - Well done, Marco.

00:35:14 ◼   ► - Yeah, but didn't we talk about this?

00:35:16 ◼   ► Like what year we thought the Mac Pro should stop?

00:35:18 ◼   ► I forget what I predicted,

00:35:19 ◼   ► but whatever I predicted back then, I stick by,

00:35:22 ◼   ► 'cause I think that was probably a reasonable prediction.

00:35:25 ◼   ► Maybe this year, I don't know.

00:35:26 ◼   ► I feel like it's a little bit early, but you know.

00:35:28 ◼   ► We'll see.

00:35:29 ◼   ► Like, half the reason I think it's not gonna come this year

00:35:33 ◼   ► is because that would require extra work by Apple

00:35:36 ◼   ► to deal with that, and they're like, "Ah, macOS.

00:35:38 ◼   ► "We really don't put that much effort in.

00:35:39 ◼   ► "Let's do it next year."

00:35:40 ◼   ► (laughing)

00:35:41 ◼   ► - Fair enough.

00:35:42 ◼   ► All right, so speaking of macOS and also iOS,

00:35:45 ◼   ► allegedly, according to Mark Gurman,

00:35:47 ◼   ► "The Settings app is getting revamped

00:35:49 ◼   ► "with a cleaner interface, better organization,

00:35:51 ◼   ► "and much improved search."

00:35:52 ◼   ► And that's both iOS and macOS.

00:35:53 ◼   ► Like I said, I'll believe it when I see it,

00:35:56 ◼   ► insofar as I don't doubt that they've touched

00:36:00 ◼   ► and maybe rejiggered some stuff in Settings,

00:36:03 ◼   ► but I don't know.

00:36:04 ◼   ► I am not as offended by Settings as most people,

00:36:07 ◼   ► but I can absolutely tell you it is a mess.

00:36:09 ◼   ► And although I haven't looked at it in a little while,

00:36:11 ◼   ► I feel like they've kind of given a hint about this

00:36:14 ◼   ► on the Vision Pro, because the Vision Pro settings

00:36:15 ◼   ► are vastly different than any other platform.

00:36:18 ◼   ► And to my recollection,

00:36:20 ◼   ► they do make a modicum of sense in there.

00:36:23 ◼   ► So maybe we'll see that getting cribbed

00:36:25 ◼   ► and put into iOS and macOS.

00:36:27 ◼   ► But one way or another, we're gonna get something different.

00:36:29 ◼   ► - Yeah, this rumor doesn't have enough detail

00:36:31 ◼   ► to be that interesting, either because it's just a vague

00:36:34 ◼   ► rumor and they don't have details,

00:36:35 ◼   ► or because the person conveying this rumor

00:36:37 ◼   ► doesn't understand what's wrong with Settings.

00:36:40 ◼   ► - Right.

00:36:41 ◼   ► - They just say, "I was revamped to get a cleaner interface."

00:36:43 ◼   ► I don't think the complaints were that the interface

00:36:44 ◼   ► was not clean.

00:36:46 ◼   ► - I mean, I don't feel clean using it.

00:36:48 ◼   ► (laughing)

00:36:49 ◼   ► - Jason had a good article of like,

00:36:50 ◼   ► "Look, you can use hierarchy to arrange information

00:36:53 ◼   ► "visually and conceptually, and that can make things

00:36:56 ◼   ► "easier to find."

00:36:58 ◼   ► And we don't wanna re-litigate everything

00:37:00 ◼   ► that's wrong with Settings, but there's no concrete things

00:37:02 ◼   ► about what are they fixing.

00:37:03 ◼   ► They're like, "Ah, well they're just rearranging stuff.

00:37:04 ◼   ► "Some things, they're putting things in different places,

00:37:06 ◼   ► "and the search will be better."

00:37:08 ◼   ► Those are all good, improvements are welcome,

00:37:10 ◼   ► but I haven't seen anyone say,

00:37:13 ◼   ► "They're totally overhauling it,

00:37:15 ◼   ► "and it's gonna be good now."

00:37:16 ◼   ► It's just, it's like minor improvements.

00:37:19 ◼   ► And I guess if you do minor improvements many years

00:37:21 ◼   ► repeatedly, you'll get better and better, and it'll be fine.

00:37:23 ◼   ► But I'm still not jazzed about the Settings app.

00:37:27 ◼   ► I mean, it's still full of bugs even.

00:37:29 ◼   ► Like, setting aside the design, which is I think still poor,

00:37:34 ◼   ► I don't really necessarily need them to massively revamp

00:37:38 ◼   ► the design if it's still gonna be this buggy.

00:37:40 ◼   ► Like, I'd rather, if they only do one thing to improve it,

00:37:43 ◼   ► I'd rather they make the current design

00:37:44 ◼   ► actually work correctly.

00:37:46 ◼   ► But that being said, a new design would also be welcome.

00:37:49 ◼   ► But I'm hoping whatever this rumor is,

00:37:51 ◼   ► I'm hoping that what they're really saying is,

00:37:54 ◼   ► it's been worked on a lot, and maybe that will include

00:37:56 ◼   ► some bug fixes also.

00:37:57 ◼   ► But it still feels like a broken web app version

00:38:02 ◼   ► of something that's trying to imitate an iOS settings screen.

00:38:06 ◼   ► Like, it does not feel like anything that would ever

00:38:09 ◼   ► have come from Apple if you would have shown people,

00:38:11 ◼   ► like, you know, even five or 10 years ago, like,

00:38:13 ◼   ► "Hey, this is current Apple software in the year 2024."

00:38:17 ◼   ► They would look at it and be like, "Are you serious?"

00:38:19 ◼   ► - Yeah, when we're talking about Mac OS settings,

00:38:21 ◼   ► but this rumor's actually primarily about iOS settings,

00:38:24 ◼   ► which, you know, we all are familiar with,

00:38:26 ◼   ► and doesn't need as much attention as Mac OS settings,

00:38:28 ◼   ► but I think they're rearranging stuff,

00:38:31 ◼   ► splitting out some things that were previously combined.

00:38:33 ◼   ► iOS settings, the search has always,

00:38:35 ◼   ► has had many years of work in it,

00:38:36 ◼   ► so that's how a lot of us navigate settings,

00:38:38 ◼   ► 'cause you can't remember exactly where something is.

00:38:40 ◼   ► But, you know, in all cases, I feel like, you know,

00:38:44 ◼   ► the general philosophy of settings in both Mac OS

00:38:46 ◼   ► or iOS is being maintained to be the same.

00:38:49 ◼   ► They're just like, "Maybe we could arrange things better."

00:38:51 ◼   ► And I think on iOS, the design is reasonable.

00:38:54 ◼   ► On Mac OS, they should have better information,

00:38:57 ◼   ► organization, and hierarchy,

00:38:59 ◼   ► given the larger size of the screens

00:39:01 ◼   ► and the more sophisticated input methods.

00:39:02 ◼   ► And as for the bugs, we've said this before,

00:39:04 ◼   ► when settings was first redesigned,

00:39:05 ◼   ► how much of it is the settings app?

00:39:06 ◼   ► How much of it is Swift UI bugs, you know?

00:39:09 ◼   ► I mean, I know it's all the same company,

00:39:10 ◼   ► and they can complain directly to the Swift UI people,

00:39:12 ◼   ► but when stuff like text fields, like,

00:39:14 ◼   ► don't stay editable or lose focus

00:39:16 ◼   ► when you're in the middle of typing,

00:39:18 ◼   ► I worry, wonder about that.

00:39:19 ◼   ► And sometimes they're using, like,

00:39:20 ◼   ► quote-unquote standard controls

00:39:21 ◼   ► that I think are bad standard controls.

00:39:23 ◼   ► The settings team didn't define those controls.

00:39:25 ◼   ► They're using the standard controls

00:39:26 ◼   ► as part of this form thing in Swift UI,

00:39:29 ◼   ► and I think those controls are bad.

00:39:31 ◼   ► I think they look bad.

00:39:32 ◼   ► I think they don't work well,

00:39:33 ◼   ► but the settings team didn't make those, right?

00:39:36 ◼   ► They're just, you know, using the OS-level controls.

00:39:39 ◼   ► So to the extent that they're buggy,

00:39:40 ◼   ► I think a lot of it is probably on the frameworks teams,

00:39:43 ◼   ► and to the extent that those controls are bad,

00:39:45 ◼   ► that's all on the frameworks team.

00:39:46 ◼   ► Like, one of the things that drives me up a wall,

00:39:49 ◼   ► and I just can't believe, this is not,

00:39:51 ◼   ► I guess it's in settings,

00:39:52 ◼   ► one of the passwords pane in settings,

00:39:53 ◼   ► when you edit, like, a password,

00:39:56 ◼   ► and it makes you edit text in, like, a right-aligned thing.

00:39:59 ◼   ► I feel like I'm writing in a right-to-left language,

00:40:01 ◼   ► or a left-to-whatever,

00:40:02 ◼   ► whatever the opposite of the way English is.

00:40:05 ◼   ► - Right-to-left.

00:40:06 ◼   ► - Have you done this and seen it?

00:40:07 ◼   ► It breaks my brain, how, like, backspacing

00:40:10 ◼   ► and forward-spacing and moving text goes.

00:40:13 ◼   ► You're editing right-aligned text for, like, the password.

00:40:17 ◼   ► I don't want any surprises or unexpected behaviors

00:40:20 ◼   ► when I'm editing the text of a password in plain text, right?

00:40:24 ◼   ► Just, that's not a good way to let me edit information.

00:40:29 ◼   ► It's name-value pairs, username, password, website.

00:40:33 ◼   ► Don't make it right-aligned in a text field

00:40:35 ◼   ► that has no borders,

00:40:36 ◼   ► it's the same color as the background,

00:40:37 ◼   ► that just behaves in bizarre ways.

00:40:40 ◼   ► That's the stuff that really needs to be fixed,

00:40:42 ◼   ► and the settings team,

00:40:43 ◼   ► well, maybe for the right-aligned thing,

00:40:44 ◼   ► they have control over that,

00:40:45 ◼   ► but I'm not sure how much control

00:40:46 ◼   ► the settings team has over that, so.

00:40:48 ◼   ► - Well, but, I mean, this is maybe not their fault,

00:40:52 ◼   ► but it is their problem, you know?

00:40:54 ◼   ► Yes, they were building this app on what seems

00:40:57 ◼   ► to have been definitely a shaky design concept

00:41:01 ◼   ► of let's just make it look like iOS,

00:41:03 ◼   ► and that didn't go so well,

00:41:05 ◼   ► building it on top of SwiftUI for Mac,

00:41:07 ◼   ► which is not the greatest platform for SwiftUI

00:41:11 ◼   ► to date so far, but at the same time,

00:41:14 ◼   ► nobody was forcing them to do this.

00:41:15 ◼   ► Nobody was, or at least, you know, the SwiftUI team

00:41:18 ◼   ► doesn't run the company,

00:41:19 ◼   ► so the SwiftUI team was not forcing the rest of the company

00:41:21 ◼   ► to say, "Hey, you better rewrite all your stuff

00:41:23 ◼   ► "in SwiftUI right now."

00:41:24 ◼   ► And even if they were, that's not our problem

00:41:28 ◼   ► as the customers, like, this is a self-created problem

00:41:31 ◼   ► internally to Apple, so what we see,

00:41:35 ◼   ► we don't care what it's written in,

00:41:37 ◼   ► we don't care what team dynamics there were,

00:41:40 ◼   ► or what they wanted to do on a technical level

00:41:41 ◼   ► to make it more idealistic,

00:41:43 ◼   ► all we see is the Settings app is really weird and sucks now.

00:41:47 ◼   ► And granted, it was not in perfect shape before,

00:41:50 ◼   ► but it's bad now, and it looks like they did

00:41:54 ◼   ► a whole bunch of work and came out with something

00:41:56 ◼   ► that was just still very bad, just in new and different ways.

00:42:00 ◼   ► You know, the example you gave with the typing

00:42:02 ◼   ► in a write-a-line text field,

00:42:04 ◼   ► so much of what I see in the Settings app

00:42:07 ◼   ► feels like no one has ever used it.

00:42:09 ◼   ► Like, I feel like I'm trying out some code

00:42:13 ◼   ► that I just wrote for the first time,

00:42:15 ◼   ► and I'm running it in the simulator,

00:42:16 ◼   ► and I'm trying it out, and I'm like,

00:42:17 ◼   ► "Oh, this doesn't work.

00:42:19 ◼   ► "Okay, back to the drawing board."

00:42:20 ◼   ► But they shipped it that way.

00:42:22 ◼   ► Like, it just seems like it's like a design concept

00:42:26 ◼   ► that they showed in a meeting one day,

00:42:28 ◼   ► and someone hit the wrong button and shipped it.

00:42:30 ◼   ► - Yeah. - Yeah.

00:42:31 ◼   ► And the thing is, the reason I think

00:42:34 ◼   ► they're sticking with design, it did solve

00:42:36 ◼   ► a bunch of the problems they had

00:42:37 ◼   ► with the old fixed layout.

00:42:38 ◼   ► Like, if the old fixed layout just wasn't scalable,

00:42:40 ◼   ► you couldn't add more stuff easily,

00:42:42 ◼   ► everything was sort of hand-laid out,

00:42:43 ◼   ► and it was hand-laid out over a decade and a half

00:42:46 ◼   ► by different people, so you could sort of see

00:42:48 ◼   ► the different, one preference pane was different

00:42:50 ◼   ► than another was different than another,

00:42:51 ◼   ► based on when it was made and who made it

00:42:53 ◼   ► and what they decided to do and what mood they were in.

00:42:55 ◼   ► And this is uniform and scalable

00:42:56 ◼   ► and scrollable and extensible,

00:42:58 ◼   ► but the building blocks that they used to achieve that,

00:43:01 ◼   ► they're bad building blocks, right?

00:43:03 ◼   ► The individual controls, the way it scales,

00:43:06 ◼   ► the way it looks, the way it works, that's all bad.

00:43:08 ◼   ► So they did solve their problem.

00:43:10 ◼   ► And hey, now we have a scalable solution for settings

00:43:12 ◼   ► in Mac OS, but they've brought on themselves,

00:43:15 ◼   ► to your point, Marco, many new problems

00:43:17 ◼   ► by using these new controls that either weren't ready

00:43:20 ◼   ► or buggy or just bad ideas, and that is a shame.

00:43:23 ◼   ► So we'll see.

00:43:24 ◼   ► Again, I welcome any kind of improvement,

00:43:25 ◼   ► but there is no rumor that it's a complete overhaul

00:43:28 ◼   ► that adds information hierarchy and visual hierarchy

00:43:31 ◼   ► and restores some of the hand-laid out things

00:43:34 ◼   ► for the most common controls,

00:43:35 ◼   ► 'cause that's what they should really do.

00:43:36 ◼   ► Like, the old hand-laid out one, like the good ones,

00:43:39 ◼   ► the nice preference panes that were hand-laid out

00:43:41 ◼   ► and beautiful, they were nice and easy to use,

00:43:43 ◼   ► and you could remember what they look like,

00:43:44 ◼   ► and you could remember where controls were spatially,

00:43:46 ◼   ► and it was like being in a place in a room

00:43:48 ◼   ► with furniture arranged and all my wonderful things

00:43:50 ◼   ► that I love, but it was just nice, right?

00:43:52 ◼   ► And you can't do that for every single one.

00:43:53 ◼   ► So you do need a scalable solution for more settings, right?

00:43:56 ◼   ► You gotta have that, and they should have a better system.

00:43:58 ◼   ► But for the most commonly used screen,

00:44:00 ◼   ► like the five most commonly used settings screens,

00:44:03 ◼   ► and I'm sure Apple knows what those are, hand-lay 'em out.

00:44:06 ◼   ► Bring in an actual good UI designer

00:44:08 ◼   ► who remembers what Mac user interfaces are supposed to look like.

00:44:10 ◼   ► Hell, bring someone who's ever used a web form,

00:44:13 ◼   ► because I can't remember any time I was asked

00:44:15 ◼   ► to update my passport on a web page,

00:44:18 ◼   ► and the text field was right aligned in English.

00:44:21 ◼   ► Doesn't make any sense.

00:44:22 ◼   ► Who thought of that?

00:44:23 ◼   ► Or the labels were massively far away from the values.

00:44:26 ◼   ► Just disappointing.

00:44:28 ◼   ► - It's an adventure in there, but we'll see what happens.

00:44:33 ◼   ► VisionOS 2.0.

00:44:36 ◼   ► There haven't been very many rumors about this.

00:44:38 ◼   ► Somebody pointed out, I don't have the toot in front of me,

00:44:41 ◼   ► but somebody pointed out that, hey,

00:44:43 ◼   ► VisionOS shipped in the middle

00:44:45 ◼   ► of the software life cycle for Apple.

00:44:48 ◼   ► And so it would not be surprising if VisionOS 2.0

00:44:51 ◼   ► is really just getting all the low-hanging,

00:44:52 ◼   ► or some of the low-hanging fruit that didn't ship for 1.0.

00:44:55 ◼   ► But one of the things that's been rumored, maybe,

00:44:57 ◼   ► is that some, or perhaps even all, of Apple's iPad apps

00:45:01 ◼   ► that are running in compatibility mode on the Vision Pro,

00:45:03 ◼   ► perhaps those will become real, honest-to-goodness

00:45:06 ◼   ► Vision Pro apps, so things like home,

00:45:08 ◼   ► calendar, podcast pages, et cetera.

00:45:11 ◼   ► Which I think is possible, although it would not surprise me

00:45:14 ◼   ► if that was issued in favor of doing something more

00:45:18 ◼   ► whiz-bang, but we'll see.

00:45:20 ◼   ► Maybe we would be able to rearrange the home screen.

00:45:22 ◼   ► That'd be cool.

00:45:23 ◼   ► - Yeah, so if these apps, if the things that shipped

00:45:26 ◼   ► as iPad apps are native, it makes me think

00:45:28 ◼   ► that they were already being developed,

00:45:30 ◼   ► they just weren't ready for launch, right?

00:45:32 ◼   ► 'Cause I don't feel like they would launch the Vision Pro,

00:45:35 ◼   ► and then say, okay, now all those things

00:45:37 ◼   ► we had to ship as iPad apps, let's get teams

00:45:39 ◼   ► on those to make them.

00:45:41 ◼   ► Seems like they would have to be in progress already

00:45:42 ◼   ► and just didn't make the deadline.

00:45:43 ◼   ► But either way, this is embarrassing

00:45:44 ◼   ► when you launch a new platform,

00:45:46 ◼   ► and a whole bunch of your really important apps

00:45:49 ◼   ► are the iPad versions.

00:45:51 ◼   ► But Apple has done that in the past, like with the iPad,

00:45:55 ◼   ► not getting versions of Apple's own Pro apps

00:45:58 ◼   ► for years and years and years.

00:46:00 ◼   ► Sometimes it takes Apple a really long time

00:46:02 ◼   ► to get around to doing stuff.

00:46:03 ◼   ► And what was the iPad OS rumor

00:46:05 ◼   ► that's gonna get a calculator now?

00:46:06 ◼   ► It's just they just never got around to it, right?

00:46:08 ◼   ► And so, I mean, if they do this, at least it shows,

00:46:10 ◼   ► if they have, if they're all,

00:46:12 ◼   ► if all the iPad apps are now native,

00:46:14 ◼   ► I think that does show some level of commitment

00:46:16 ◼   ► that they realized it was embarrassing not to have that,

00:46:18 ◼   ► and it was just a timing thing,

00:46:19 ◼   ► and now they're fixing that.

00:46:21 ◼   ► But I agree with you, Casey,

00:46:22 ◼   ► that I don't expect Vision OS 2.0 to be massively different.

00:46:26 ◼   ► There's just so many obvious things they need to fix

00:46:30 ◼   ► and enhance, fix bugs, add minor features.

00:46:32 ◼   ► I don't even expect them to have

00:46:34 ◼   ► a big home screen rearranging thing.

00:46:35 ◼   ► If they do, it'll be very rudimentary,

00:46:37 ◼   ► 'cause this is really just kind of like the,

00:46:39 ◼   ► you know, make it what we wish 1.0 could have been

00:46:42 ◼   ► if we had an extra, you know, six months.

00:46:45 ◼   ► - Yeah, 'cause that's really all it's been.

00:46:46 ◼   ► Like, you know, you think about every year,

00:46:49 ◼   ► you think about these platforms

00:46:50 ◼   ► going up by one whole version number,

00:46:51 ◼   ► and there being a certain kind of minimum amount

00:46:53 ◼   ► of improvement, but keep in mind,

00:46:55 ◼   ► like, Vision Pro was a half-year release.

00:46:58 ◼   ► And so it really hasn't been that long since 1.0.

00:47:01 ◼   ► And from all we've heard, it seems like software

00:47:05 ◼   ► was the, like, gating factor of when it could be released,

00:47:08 ◼   ► not hardware, so like, it's not like they have

00:47:10 ◼   ► a bunch of software work saved up for years,

00:47:12 ◼   ► you know, waiting to be released.

00:47:13 ◼   ► It's like, no, you know, what we got so far,

00:47:15 ◼   ► like, that's what they have so far.

00:47:17 ◼   ► And so I'm not expecting a lot of movement

00:47:20 ◼   ► on the Vision Pro software side yet.

00:47:22 ◼   ► It's just, it's too soon.

00:47:23 ◼   ► And that being said, like, keep in mind, too,

00:47:26 ◼   ► this is a very, like, distant forward-looking platform

00:47:30 ◼   ► for them, this is not gonna be a big seller

00:47:32 ◼   ► for a long time, if ever.

00:47:34 ◼   ► And so I think the level of improvement to expect

00:47:39 ◼   ► for Vision OS releases, even going forward,

00:47:41 ◼   ► even when they have full years to do it,

00:47:43 ◼   ► I'm expecting something more like WatchOS releases,

00:47:48 ◼   ► or tvOS releases, like, not, I'm not expecting, like,

00:47:51 ◼   ► major new changes and major new features

00:47:54 ◼   ► to be coming out every single year.

00:47:55 ◼   ► I think it's gonna be a slower update pace.

00:47:57 ◼   ► Like, I think this is a very long game

00:48:00 ◼   ► they're playing with Vision Pro.

00:48:01 ◼   ► I think, honestly, I would love to see them

00:48:03 ◼   ► step on the gas a lot harder,

00:48:04 ◼   ► but what we're seeing so far shows a different strategy,

00:48:07 ◼   ► but regardless, I don't expect a lot of movement

00:48:11 ◼   ► on this platform every single year.

00:48:13 ◼   ► I think it's gonna be a very slow burn.

00:48:16 ◼   ► - You did get that immersive environment

00:48:17 ◼   ► for working in Xcode, with the wavy neon lines and stuff.

00:48:21 ◼   ► - Oh, yeah, from the developer app.

00:48:23 ◼   ► - I mean, that's cool.

00:48:24 ◼   ► I guess I actually haven't tried this out yet,

00:48:25 ◼   ► but that's a thing, which is neat.

00:48:29 ◼   ► But yeah, I don't know, I would love to see,

00:48:32 ◼   ► even before I see a lot of software motion on VisionOS,

00:48:36 ◼   ► I really would prefer to just see content

00:48:39 ◼   ► more than anything else.

00:48:40 ◼   ► Like, that's what I feel like I want,

00:48:42 ◼   ► and we're just not getting it right now,

00:48:44 ◼   ► which is too bad, but, you know, it is what it is.

00:48:47 ◼   ► I will say, though, I was briefly using Vision Pro

00:48:49 ◼   ► using the Mac virtual display mode

00:48:53 ◼   ► with the developer strap on with a high-speed USB cable

00:48:57 ◼   ► plugged into my Mac, and I do really like that.

00:49:00 ◼   ► I mean, I was at home, so I only needed it

00:49:02 ◼   ► for a minute or two because I was working on something,

00:49:04 ◼   ► but it is pretty cool when it works,

00:49:08 ◼   ► particularly when it's done via a cable,

00:49:11 ◼   ► because then latency is much, much lower,

00:49:14 ◼   ► and I feel like the crispness of the display

00:49:17 ◼   ► is also considerably better, which I enjoy.

00:49:20 ◼   ► But yeah, I mean, I just want content.

00:49:22 ◼   ► Give me more content, 'cause I'm not using this thing

00:49:24 ◼   ► for work a whole heck of a lot.

00:49:25 ◼   ► I'm not sure anyone really is,

00:49:26 ◼   ► so I would really just love more content.

00:49:30 ◼   ► - Yeah, and that would also, I think,

00:49:32 ◼   ► significantly drive sales, too.

00:49:33 ◼   ► Like, that's what we need first, 'cause you're right.

00:49:35 ◼   ► Like, it's gonna be a long time

00:49:37 ◼   ► before there's meaningful software on here, if ever.

00:49:42 ◼   ► So, again, like Apple TV, like this,

00:49:44 ◼   ► when we had gone to the thing in November

00:49:50 ◼   ► that we can't talk about, the one thing I was dying

00:49:53 ◼   ► to talk about in the intervening months

00:49:54 ◼   ► between that and release was it was very obvious to me,

00:49:57 ◼   ► even from some past time, it was very obvious to me

00:50:02 ◼   ► that this feels like computing on a virtual Apple TV

00:50:06 ◼   ► with an infinitely sized screen.

00:50:08 ◼   ► And it does kind of physically feel like

00:50:11 ◼   ► you are kind of manipulating these giant apps

00:50:14 ◼   ► in your giant space with these kind of indirect mechanisms

00:50:18 ◼   ► of your eyes and pinching and flinging things around

00:50:21 ◼   ► and stuff like that.

00:50:22 ◼   ► And it kind of, like, it felt a lot like computing

00:50:27 ◼   ► on an Apple TV, but just in a much bigger way.

00:50:29 ◼   ► And obviously, better in some ways, worse in others.

00:50:31 ◼   ► But for the most part, it felt like it was closer to that

00:50:36 ◼   ► than using a laptop or an iPad or something like that.

00:50:38 ◼   ► And so, I think what I would expect here is,

00:50:42 ◼   ► I think it's gonna end up being a lot like the Apple TV

00:50:44 ◼   ► as an app platform, which is most apps will not find

00:50:48 ◼   ► much reason to be there, and most users won't want

00:50:51 ◼   ► to use most apps there, but with the right content

00:50:55 ◼   ► and with the right specialized apps,

00:50:57 ◼   ► this could be something people use frequently.

00:50:59 ◼   ► It just needs that content to be there.

00:51:01 ◼   ► Like, I never even think about making an Apple TV app,

00:51:04 ◼   ► but I use my Apple TV almost every day.

00:51:06 ◼   ► What do I use on my Apple TV?

00:51:08 ◼   ► Like, four apps.

00:51:09 ◼   ► But I use them almost every day.

00:51:11 ◼   ► So, the Vision Pro, I think, could be there.

00:51:13 ◼   ► I think it could reach that point where people are using it

00:51:16 ◼   ► for mostly content consumption and maybe occasional

00:51:21 ◼   ► specialized uses or very narrow productivity uses

00:51:25 ◼   ► for certain people.

00:51:26 ◼   ► I see that being possible, but I think to get there,

00:51:29 ◼   ► first, we need the install base to grow,

00:51:31 ◼   ► and the only thing that's really gonna meaningfully grow,

00:51:32 ◼   ► and I think at this point, is content.

00:51:34 ◼   ► So, that's what I'm really watching for,

00:51:36 ◼   ► is where are the content deals,

00:51:39 ◼   ► where's the content coming from,

00:51:41 ◼   ► like how much can there be, how quickly.

00:51:44 ◼   ► That's what will actually move these things.

00:51:45 ◼   ► And then we can talk about computing possibilities

00:51:47 ◼   ► down the road.

00:51:48 ◼   ► - Yeah.

00:51:49 ◼   ► But I tell you what, if you are traveling

00:51:51 ◼   ► and want a huge ass screen that doesn't take

00:51:55 ◼   ► a whole lot of space, it is without peer.

00:51:58 ◼   ► So, there are some uses, as few and far between

00:52:01 ◼   ► as they may be.

00:52:02 ◼   ► Open AI.

00:52:04 ◼   ► According to Mark Gurman, Apple picked Open AI

00:52:07 ◼   ► as its inaugural AI partner for a few reasons.

00:52:09 ◼   ► It got better business terms than Google was offering,

00:52:11 ◼   ► and Apple believes that Open AI's technology

00:52:12 ◼   ► is the best available on the market.

00:52:13 ◼   ► Apple is expected to offer its new AI features

00:52:16 ◼   ► as an opt-in service, according to people familiar

00:52:18 ◼   ► with the matter.

00:52:19 ◼   ► So, yeah, I think everyone's expecting the AI dust

00:52:23 ◼   ► to be sprinkled all over the place.

00:52:25 ◼   ► It's yet to be seen if somebody in the C-suite

00:52:28 ◼   ► got a burr up their butt to put AI on everything,

00:52:31 ◼   ► or if Apple will take what I hope,

00:52:34 ◼   ► if Apple does what I hope they'll do,

00:52:36 ◼   ► which is take a more mature approach and ask,

00:52:38 ◼   ► where are we actually helping?

00:52:41 ◼   ► Let's not do this just because we can.

00:52:43 ◼   ► Let's do it because we should.

00:52:44 ◼   ► And honestly, I think that there's far fewer places

00:52:48 ◼   ► that AI should be used than where AI can be used,

00:52:52 ◼   ► and we'll see what happens.

00:52:53 ◼   ► But I'm hopeful they're gonna be mature about it.

00:52:56 ◼   ► - I was looking at this Gurman article.

00:52:57 ◼   ► I'm doing the search for, like, the Bloomberg style guide

00:52:59 ◼   ► says when you're attributing anonymous sources,

00:53:02 ◼   ► you have to say, "The people said," right?

00:53:05 ◼   ► So, you just search for the people, that string,

00:53:08 ◼   ► 'cause that'll find you the tidbits

00:53:09 ◼   ► that he's saying are sourced.

00:53:10 ◼   ► My sources say this.

00:53:12 ◼   ► And everything you read there was one of those,

00:53:13 ◼   ► the people said things.

00:53:14 ◼   ► So, I removed that text from some of the snippets

00:53:18 ◼   ► 'cause it's awkward.

00:53:19 ◼   ► Anyway, the reasons for picking,

00:53:22 ◼   ► it was they were talking to Google and other partners

00:53:23 ◼   ► or whatever, that Apple thinks that they're the best,

00:53:25 ◼   ► that they're the market leader,

00:53:27 ◼   ► and that they gave better business terms to Google,

00:53:29 ◼   ► which kinda makes sense 'cause OpenAI's younger and hungrier,

00:53:32 ◼   ► you know what I mean?

00:53:33 ◼   ► And the OpenAI thing is interesting

00:53:36 ◼   ► because everything you said about the AI features, right,

00:53:40 ◼   ► we assume that there'll be a bunch of those,

00:53:41 ◼   ► like, "Oh, you can now do a useful thing,"

00:53:43 ◼   ► and, "Yeah, we'll say it's AI," or whatever.

00:53:45 ◼   ► It's the same stuff we used to say that was ML, right?

00:53:47 ◼   ► But the chatbot style, "Hey, you talk to an LLM

00:53:52 ◼   ► as if it's a little person back and forth,"

00:53:55 ◼   ► all the OpenAI rumors are like,

00:53:58 ◼   ► that's what OpenAI would be providing

00:54:00 ◼   ► because Apple has an internal one that they're working on.

00:54:02 ◼   ► They have their own kind of LLM chatbot named Ajax

00:54:04 ◼   ► or whatever, but it's not ready,

00:54:05 ◼   ► and OpenAI's things are better,

00:54:07 ◼   ► so let's do a deal with them.

00:54:10 ◼   ► But I still, the fact that it would be opt-in,

00:54:13 ◼   ► like, what does that mean?

00:54:14 ◼   ► Does that mean that when you're setting up your computer

00:54:16 ◼   ► and says, "Oh, and by the way,

00:54:17 ◼   ► "we have a partnership in OpenAI.

00:54:19 ◼   ► "Do you want to enable feature that puts an icon

00:54:21 ◼   ► "in your menu bar that you can click on

00:54:22 ◼   ► "that props up a text box,

00:54:24 ◼   ► "and now you're talking to chat GPT or something?"

00:54:26 ◼   ► Like, what is opting into this?

00:54:27 ◼   ► How does a chatbot fit into the OS?

00:54:30 ◼   ► We can see how all the other stuff fits in,

00:54:31 ◼   ► and like I said last week,

00:54:33 ◼   ► the rumors that come down from on high

00:54:35 ◼   ► that's telling every single team,

00:54:36 ◼   ► "Find a place to put AI into your product,"

00:54:38 ◼   ► that's bad, and that's kind of the worst of corporate,

00:54:42 ◼   ► like, "Oh, it's a buzzword.

00:54:43 ◼   ► "Let's just find a place to shoehorn this into our product."

00:54:45 ◼   ► But to some degree, I trust the product teams

00:54:48 ◼   ► to think, "What can we do with AI that is useful

00:54:50 ◼   ► "in the Notes app, in Pages, in the Finder?"

00:54:53 ◼   ► Like, I don't think any of them are gonna be like,

00:54:56 ◼   ► "I know how we can use AI in our app.

00:54:58 ◼   ► "Let's put a text box up,

00:55:00 ◼   ► "and they'll send the message to chat GPT."

00:55:01 ◼   ► Like, they'll think of something better, right?

00:55:03 ◼   ► It's bad that they're asking all the teams to do that,

00:55:05 ◼   ► because maybe it doesn't make sense

00:55:06 ◼   ► to some teams to add AI features,

00:55:07 ◼   ► but the mandate, we keep seeing this rumor

00:55:09 ◼   ► that the mandate came down from on high.

00:55:11 ◼   ► Every team, you better find some way to add AI,

00:55:13 ◼   ► because this is the AI year, right?

00:55:15 ◼   ► But that is entirely separate from this OpenAI deal,

00:55:17 ◼   ► which is like, "Oh, you can opt in somehow in Mac OS

00:55:21 ◼   ► "or in iOS to, like, on iOS,

00:55:24 ◼   ► "to just install the chat GPT app?"

00:55:25 ◼   ► Like, what does this mean?

00:55:26 ◼   ► This is the thing that I'm looking for the most

00:55:29 ◼   ► in this presentation, that they're gonna be on stage,

00:55:31 ◼   ► they're gonna say, "We've partnered with OpenAI,"

00:55:32 ◼   ► or maybe they won't say that, and somehow,

00:55:35 ◼   ► some way, they're going to say,

00:55:37 ◼   ► "In our operating systems that we're shipping,

00:55:39 ◼   ► "there's a way for you to start talking to chat GPT

00:55:41 ◼   ► "and having a little conversation

00:55:43 ◼   ► "in a little window somewhere."

00:55:44 ◼   ► I don't even think that's that useful a feature.

00:55:46 ◼   ► There's already a chat GPT app.

00:55:48 ◼   ► There's a Mac one with Swift UI, there's a native iOS one,

00:55:51 ◼   ► you can use it on the web, like, what is this?

00:55:53 ◼   ► I don't get it, and this is the thing that's baffling me,

00:55:56 ◼   ► because, as we said, I think this is the highest potential

00:55:58 ◼   ► for badness, because it is such a non-Apple,

00:56:05 ◼   ► non-controlled, non-careful, just like,

00:56:07 ◼   ► "Wild West, talk to chat GPT, good luck, it's not our thing."

00:56:10 ◼   ► How is that a selling point?

00:56:12 ◼   ► How are you going to promote that as a new feature

00:56:16 ◼   ► of your operating systems?

00:56:17 ◼   ► If a non-hardware WWDC, I don't think there's probably

00:56:20 ◼   ► gonna be a developer story, unless Apple's gonna pay

00:56:23 ◼   ► for your API tokens, and now you can write an iOS

00:56:26 ◼   ► or a Mac app and use their framework

00:56:27 ◼   ► and get free access to chat GPT,

00:56:29 ◼   ► because Apple did some multi-million dollar deal

00:56:31 ◼   ► with them or something.

00:56:32 ◼   ► This baffles me, that's why I added this thing

00:56:34 ◼   ► of the whole opt-in, like, what does that mean?

00:56:36 ◼   ► What are we opting into?

00:56:37 ◼   ► What are you providing?

00:56:38 ◼   ► I have red flags all over this.

00:56:41 ◼   ► (laughing)

00:56:44 ◼   ► - Yeah, I don't know, I'm so turned off

00:56:49 ◼   ► by Sam Altman and OpenAI.

00:56:52 ◼   ► I don't know, it's funny, the dude who we all snickered

00:56:54 ◼   ► about wearing the two popped collar polo shirts,

00:56:58 ◼   ► wasn't that WWDC, like, forever and a day ago?

00:57:00 ◼   ► - A long time ago, yeah.

00:57:01 ◼   ► - Yeah, and we all laughed and had a good,

00:57:04 ◼   ► you know, laugh at his expense, and, you know,

00:57:06 ◼   ► oh, look at this Silicon Valley bro

00:57:08 ◼   ► just being the world's bro-iest bro.

00:57:11 ◼   ► And fast forward five, 10 years, and it seems like,

00:57:14 ◼   ► same as it ever was, nothing's ever really changed,

00:57:16 ◼   ► like, it's, you know, move fast, break things,

00:57:18 ◼   ► and have no consequences, and, I don't know,

00:57:21 ◼   ► like, again, we'll see what happens,

00:57:23 ◼   ► like, I am cautiously optimistic, and again,

00:57:25 ◼   ► I like to think of Apple as generally one of the more mature

00:57:30 ◼   ► and adult organizations in the Valley.

00:57:33 ◼   ► I'm sure there's exceptions, I'm sure that's not always true,

00:57:35 ◼   ► but broadly I think that to be the case,

00:57:38 ◼   ► and I don't know, I feel, I hope,

00:57:40 ◼   ► and I so desperately want Apple to be deliberate,

00:57:45 ◼   ► be considered, be mature, and not just put the AI at dust

00:57:50 ◼   ► on everything they see, but, hey, we'll see.

00:57:54 ◼   ► - I mean, like, I don't, I find Chat TV useful.

00:57:57 ◼   ► I use, it's in my rotation of things that I do.

00:58:00 ◼   ► I do all the chat bots, I do Google searches,

00:58:02 ◼   ► like, I get useful things out of it.

00:58:05 ◼   ► There is use to be had there, but I'm doing that already.

00:58:08 ◼   ► Like, I don't need Apple to add any features to macOS

00:58:10 ◼   ► for me to do this.

00:58:11 ◼   ► I can have a free account on OpenAI, I can have a paid one,

00:58:14 ◼   ► I can use Google Gemini, I can do it on iOS,

00:58:16 ◼   ► I can do it on a Mac, like, all of that is already there.

00:58:19 ◼   ► I don't think I need any kind of OS integration

00:58:22 ◼   ► to make that experience better.

00:58:23 ◼   ► I can open a webpage, I can have, you know,

00:58:25 ◼   ► let's me, it's already open in a tab somewhere.

00:58:27 ◼   ► (laughing)

00:58:29 ◼   ► Like, you know, maybe it's not on my phone or whatever,

00:58:32 ◼   ► like, I guess they could use voice thing,

00:58:34 ◼   ► like, it integrates with Siri and you talk to Siri

00:58:36 ◼   ► and it sends it out.

00:58:37 ◼   ► Like, I'm not saying there's no room

00:58:38 ◼   ► for OS integration, there is, but I just,

00:58:40 ◼   ► this just seems such an uneasy partnership,

00:58:42 ◼   ► but also because, you know, like, the sources in this

00:58:44 ◼   ► essentially seem to indicate, some of the sources

00:58:47 ◼   ► are not inside the Apple,

00:58:48 ◼   ► sometimes they're ex-Apple people or whatever,

00:58:49 ◼   ► like, this is a stopgap.

00:58:51 ◼   ► Apple doesn't have anything Chat HTTP caliber.

00:58:53 ◼   ► They don't have anything Google Gemini caliber.

00:58:55 ◼   ► They want to have something like that,

00:58:57 ◼   ► they wanna have a story for that for W3C,

00:58:58 ◼   ► therefore they have to partner,

00:58:59 ◼   ► we talked about this before, but they are working on stuff.

00:59:02 ◼   ► Apple does have its own internal LLM.

00:59:04 ◼   ► It does have its own internal chat bot

00:59:05 ◼   ► that its own employees have been using.

00:59:08 ◼   ► It's just not up to snuff yet, which makes sense,

00:59:10 ◼   ► 'cause they got a late start, right?

00:59:11 ◼   ► But they're doing research in that area,

00:59:13 ◼   ► so presumably, if this turns out to be useful,

00:59:15 ◼   ► Apple would wanna pull a Google Maps here and say,

00:59:19 ◼   ► yeah, we'll partner with this other company

00:59:21 ◼   ► because they're the market leader,

00:59:22 ◼   ► but eventually we'll roll out our own thing

00:59:24 ◼   ► and eventually our own thing will be,

00:59:26 ◼   ► if not better than good enough, right?

00:59:28 ◼   ► And hopefully they'll do it better

00:59:29 ◼   ► than they did with Maps rollout,

00:59:30 ◼   ► but we'll see how it goes.

00:59:32 ◼   ► But still, this is what I'm looking at the most,

00:59:37 ◼   ► what's the pitch here?

00:59:37 ◼   ► How is it integrated?

00:59:38 ◼   ► How is this presented?

00:59:40 ◼   ► If it's opt-in, what does that look like?

00:59:44 ◼   ► It's so weird to sort of,

00:59:45 ◼   ► if this ends up being a tentpole feature

00:59:46 ◼   ► and not just an off to the side,

00:59:47 ◼   ► 'cause I can imagine just being like,

00:59:48 ◼   ► oh, and by the way, we partner with them and blah, blah,

00:59:50 ◼   ► and that's all they say about it, it's quick, right?

00:59:52 ◼   ► But if this is a tentpole feature,

00:59:54 ◼   ► usually tentpole features aren't like,

00:59:56 ◼   ► and we're so confident in it that it's not on by default,

00:59:58 ◼   ► it's kind of like the Windows recall thing, right?

01:00:01 ◼   ► That's a sign of something like,

01:00:03 ◼   ► we think it might be okay, but there is some danger here,

01:00:06 ◼   ► so if you just hit okay, okay, okay during setup,

01:00:10 ◼   ► whatever we're doing, you won't be exposed to, right?

01:00:14 ◼   ► But if you opt into it somehow,

01:00:16 ◼   ► then again, when you talk to Siri,

01:00:19 ◼   ► and it can't figure out the answer,

01:00:20 ◼   ► instead of telling you to check the web,

01:00:21 ◼   ► it'll ask chat GPT or something.

01:00:23 ◼   ► - Yeah, yeah, I don't know.

01:00:26 ◼   ► We'll see what happens.

01:00:28 ◼   ► I mean, that's all the items that we had in the show notes.

01:00:31 ◼   ► I just, I really don't know what to expect,

01:00:33 ◼   ► and I am so very, very interested

01:00:35 ◼   ► to see what the AI story is.

01:00:37 ◼   ► It seems that Apple's previous dedication to ML,

01:00:41 ◼   ► or machine learning, as we've talked about,

01:00:42 ◼   ► many other shows and blogs have talked about,

01:00:45 ◼   ► you know, they finally embraced AI as a term.

01:00:48 ◼   ► And so we'll see what happens.

01:00:50 ◼   ► I think it's tough because, you know,

01:00:52 ◼   ► Apple is a publicly held company,

01:00:54 ◼   ► and publicly traded company, and you know,

01:00:56 ◼   ► I think investors and, you know,

01:00:58 ◼   ► rank and file shareholders are gonna expect them

01:01:00 ◼   ► to have that AI dust sprinkled everywhere,

01:01:04 ◼   ► you know, like a graffiti cannon of,

01:01:06 ◼   ► or not graffiti, a glitter cannon of AI dust

01:01:09 ◼   ► just shooting all over, you know, everything WWDC touches.

01:01:12 ◼   ► But I don't know that that's appropriate or reasonable,

01:01:15 ◼   ► so we'll see what happens.

01:01:17 ◼   ► And I'm just curious to see, you know,

01:01:19 ◼   ► now that we've got a whole new platform,

01:01:20 ◼   ► I'm so excited to see what VisionOS 2 is.

01:01:24 ◼   ► If history tells us anything, it will be disappointing.

01:01:26 ◼   ► - Do you think VisionOS will have AI dust sprinkled on it?

01:01:29 ◼   ► - Oh no. - At all?

01:01:31 ◼   ► Anywhere?

01:01:31 ◼   ► - I'd be surprised if iPadOS even has it.

01:01:34 ◼   ► Like, keep in mind, like, you know, again,

01:01:35 ◼   ► we're still dealing with Apple's kind of hierarchy

01:01:38 ◼   ► of platform importance.

01:01:39 ◼   ► iOS is gonna get all the cool stuff first.

01:01:41 ◼   ► I would be surprised if we see it anywhere else.

01:01:45 ◼   ► I think it's gonna be iOS certainly, you know, first,

01:01:49 ◼   ► and then maybe they'll put some of the basics

01:01:52 ◼   ► into iPad and Mac.

01:01:54 ◼   ► - Well, the cross-platform stuff will get it.

01:01:56 ◼   ► Like, the whole, like, removing people

01:01:57 ◼   ► from the background of photos,

01:01:58 ◼   ► that's just gonna be everywhere,

01:01:59 ◼   ► because it's a common framework, you know what I mean?

01:02:01 ◼   ► - Yes, of course, yeah.

01:02:02 ◼   ► But like, you know, in terms of like,

01:02:04 ◼   ► features like specific to, say, like,

01:02:07 ◼   ► the Vision Pro or the watch, or like,

01:02:09 ◼   ► I would be surprised to see anything like that this soon.

01:02:13 ◼   ► - And speaking of like, sprinkling AI sparkles everywhere

01:02:16 ◼   ► and everyone expects it to do it,

01:02:17 ◼   ► at this point though, and this is what everyone expects,

01:02:20 ◼   ► but at this point, it's true that the only way

01:02:22 ◼   ► for Apple to really make news is to do something different

01:02:27 ◼   ► than everyone else does.

01:02:28 ◼   ► I mean, they can make news by saying

01:02:29 ◼   ► we're not doing any AI stuff, and that would be news,

01:02:31 ◼   ► but that's obviously not what they're doing, right?

01:02:32 ◼   ► But they're not gonna do what every other company did,

01:02:34 ◼   ► which is like, AI everywhere,

01:02:35 ◼   ► sprinkle, sprinkle, sprinkle, sprinkle, like, you know,

01:02:37 ◼   ► because that'll just be like, everyone else did that.

01:02:39 ◼   ► Like, throwing spaghetti against the wall, right?

01:02:41 ◼   ► Like Windows did.

01:02:42 ◼   ► Anything we could think of, we're just gonna do.

01:02:44 ◼   ► The only way Apple could actually make news

01:02:46 ◼   ► is by saying, we're doing AI, but we're doing it

01:02:49 ◼   ► the Apple way, where everything is useful,

01:02:51 ◼   ► and blah, blah, blah, like, they have to present it that way.

01:02:52 ◼   ► That is a marketing decision.

01:02:54 ◼   ► We already know they're already doing that.

01:02:55 ◼   ► We talked about that last week.

01:02:56 ◼   ► There's tons of features that I used to call ML

01:02:58 ◼   ► that are just features of our phone, right?

01:03:00 ◼   ► If they do that this year and they present it as,

01:03:03 ◼   ► we're not doing AI like everyone else,

01:03:04 ◼   ► we're not throwing spaghetti against the wall,

01:03:06 ◼   ► we don't have a thing where you draw a picture

01:03:07 ◼   ► and AI draws along with you,

01:03:08 ◼   ► because we didn't think that was that useful yet.

01:03:10 ◼   ► So we didn't put it in, right?

01:03:12 ◼   ► That's the story, that's the Apple story.

01:03:13 ◼   ► And the news would be, Apple, WWDC,

01:03:16 ◼   ► rolls out a bunch of AI features,

01:03:17 ◼   ► but unlike other people's AI features,

01:03:18 ◼   ► they only introduce the ones that are useful,

01:03:21 ◼   ► and also a changing thing.

01:03:22 ◼   ► Right, because that's the wild card,

01:03:24 ◼   ► but I don't understand that, right?

01:03:25 ◼   ► That is the way to make news.

01:03:27 ◼   ► Otherwise, the story will be,

01:03:28 ◼   ► Apple catches up with everyone else

01:03:29 ◼   ► and does what everyone else has already done,

01:03:31 ◼   ► which is, I guess, better than the bad story,

01:03:33 ◼   ► which is Apple doesn't do any AI stuff

01:03:35 ◼   ► and they're doomed, right?

01:03:36 ◼   ► - Well, I mean, I think that's the most likely.

01:03:38 ◼   ► I think by far the most likely reaction

01:03:40 ◼   ► by the overall press and business and Wall Street communities

01:03:44 ◼   ► is gonna be they didn't do enough.

01:03:46 ◼   ► Now, as actual Apple users, for me,

01:03:51 ◼   ► like what you were saying earlier,

01:03:52 ◼   ► I would rather they do things that are useful to me.

01:03:55 ◼   ► I don't need them to do the most stupid things that they are,

01:03:59 ◼   ► 'cause other companies will gladly do

01:04:01 ◼   ► the most stupid things that they are.

01:04:02 ◼   ► Apple doesn't need to do that.

01:04:04 ◼   ► They are, as you were saying,

01:04:05 ◼   ► they are the more grown-up company.

01:04:07 ◼   ► I don't think that is their style

01:04:10 ◼   ► to be the spaghetti against the wall, cutting edge,

01:04:13 ◼   ► let's do a bunch of stupid stuff.

01:04:15 ◼   ► That's not Apple's style.

01:04:15 ◼   ► Apple is very conservative in that kind of approach.

01:04:18 ◼   ► And to some degree, that's kind of just in their DNA.

01:04:22 ◼   ► To another degree, they need to be,

01:04:23 ◼   ► because they have much more scrutiny on them

01:04:25 ◼   ► and their moves have, they make bigger waves.

01:04:29 ◼   ► - And their brand, not that they're Disney,

01:04:31 ◼   ► but their brand is very kind of like,

01:04:33 ◼   ► let's not be hasty and do the radical stuff

01:04:37 ◼   ► just because everyone else is doing it, right?

01:04:39 ◼   ► And I honestly think if they market this well,

01:04:41 ◼   ► the story won't be Apple didn't do enough.

01:04:42 ◼   ► The story will be what Apple wanted to be,

01:04:44 ◼   ► is that Apple took a more considered approach.

01:04:46 ◼   ► Everyone else has just been like,

01:04:47 ◼   ► "Yeah, yeah, yeah, yeah."

01:04:49 ◼   ► And Apple made all their features useful.

01:04:53 ◼   ► Here are the things that it does.

01:04:55 ◼   ► It only does things that are provably useful, right?

01:04:58 ◼   ► Like they did last year and the year before

01:04:59 ◼   ► and the year before, like, oh, OS wide, you can OCR text.

01:05:03 ◼   ► Like that's an AI feature, but it's useful.

01:05:06 ◼   ► Like, and it's not, if that is the story Apple presents,

01:05:09 ◼   ► I think most of the stories from outlets

01:05:11 ◼   ► that don't have it in for Apple will be,

01:05:13 ◼   ► Apple decides to take a different approach to AI.

01:05:15 ◼   ► Unlike its competitors, Apple says that it's only rolling out

01:05:18 ◼   ► features that are provably useful.

01:05:19 ◼   ► And I'm not gonna use that term

01:05:20 ◼   ► because that's not a good marketing term,

01:05:21 ◼   ► but you know, some marketing person will come up with

01:05:23 ◼   ► a two or three word phrase that distinguishes

01:05:25 ◼   ► Apple's approach to AI, which again,

01:05:27 ◼   ► is gonna be no different than what they did last year.

01:05:30 ◼   ► The only difference is the magic of marketing, right?

01:05:32 ◼   ► And I, you know, you're right that there will be

01:05:33 ◼   ► some stories like Apple's not doing enough.

01:05:35 ◼   ► They're falling behind.

01:05:36 ◼   ► Microsoft has 40 LLMs running in the background

01:05:39 ◼   ► and Windows 11, right?

01:05:40 ◼   ► But I feel like the whole Windows co-pilot PC thing

01:05:43 ◼   ► is just so wrapped up in the recall disaster

01:05:46 ◼   ► that no one is even talking about the draw along

01:05:49 ◼   ► or whatever, or the OS wide like translation,

01:05:51 ◼   ► which I think actually is useful.

01:05:53 ◼   ► So the door is open for Apple to do something smart here.

01:05:56 ◼   ► And again, the wildcard is what the hell is OpenAI?

01:05:59 ◼   ► Because OpenAI is not that.

01:06:00 ◼   ► OpenAI and ChatGPT, that is not,

01:06:03 ◼   ► we only do the things that we are sure are useful and safe.

01:06:06 ◼   ► Because, you know, for all the uses that ChatGPT has,

01:06:10 ◼   ► it's not conservative.

01:06:12 ◼   ► (laughing)

01:06:13 ◼   ► Basically the floodgates.

01:06:14 ◼   ► It's like type anything you want, hope it works,

01:06:16 ◼   ► fingers crossed.

01:06:17 ◼   ► (laughing)

01:06:19 ◼   ► - Yeah, like that, I would be very happy with,

01:06:22 ◼   ► take most of the basics that we already know

01:06:25 ◼   ► that quote AI stuff can do,

01:06:27 ◼   ► even take a lot of the stuff the phones already do

01:06:29 ◼   ► and just make it better, make it more reliable,

01:06:31 ◼   ► but give it better results.

01:06:32 ◼   ► As we were saying last episode,

01:06:33 ◼   ► like basic dictation, speech-to-text kind of APIs,

01:06:38 ◼   ► I believe, is it all Android phones

01:06:41 ◼   ► or just the Pixel phones that have supported

01:06:44 ◼   ► system-wide closed captioning?

01:06:45 ◼   ► You can just have constant transcription showing on screen

01:06:48 ◼   ► of whatever audio is playing on the phone.

01:06:51 ◼   ► - Likely added to Windows.

01:06:52 ◼   ► - Right, like why haven't iPhones been able to do that yet?

01:06:55 ◼   ► Like that's such low-hanging fruit for Apple.

01:06:58 ◼   ► That should be a feature, of course it should be a feature.

01:07:01 ◼   ► And that's, again, things like removing people

01:07:03 ◼   ► from photos that shouldn't be there.

01:07:05 ◼   ► Like that kind of stuff,

01:07:06 ◼   ► these are not cutting-edge new ideas.

01:07:10 ◼   ► These are features that the rest of the industry has done.

01:07:12 ◼   ► And yes, Apple will be playing quote catch-up,

01:07:14 ◼   ► but that's fine as long as they do a good job of it.

01:07:17 ◼   ► The reason why some of these features stand out to us

01:07:20 ◼   ► as things Apple should do is because they are good ideas.

01:07:24 ◼   ► We are not saying Apple should do everything

01:07:25 ◼   ► everyone else has done.

01:07:26 ◼   ► Most of the other things people have done with quote AI

01:07:29 ◼   ► are weird and useless and creepy or something.

01:07:33 ◼   ► But there are a few really cool killer features

01:07:36 ◼   ► that it would be nice if Apple had good versions of.

01:07:38 ◼   ► And then beyond that, what we've already talked about.

01:07:40 ◼   ► Things like, obviously the big one is make Siri better.

01:07:43 ◼   ► Like that is the big one.

01:07:45 ◼   ► Please, for the love of God, make Siri better.

01:07:48 ◼   ► - Make Siri, this is interesting, make Siri better,

01:07:51 ◼   ► but don't turn Siri into chat GPT.

01:07:53 ◼   ► There's a fine line there because you want Siri to be better.

01:07:56 ◼   ► It's like all the things that Siri's supposed to do now,

01:07:58 ◼   ► imagine if it did them reliably

01:08:00 ◼   ► and with lots of flexibility and sophistication,

01:08:02 ◼   ► and maybe you could do one or two new things.

01:08:04 ◼   ► But that is very different than, hey, guess what?

01:08:06 ◼   ► Now you can just ask Siri anything and it will do something.

01:08:08 ◼   ► That's not what we want, right?

01:08:10 ◼   ► Because that is the throwing spaghetti against the wall

01:08:12 ◼   ► because you can get yourself into trouble real fast

01:08:14 ◼   ► with just, especially since Siri has essentially control

01:08:17 ◼   ► over your phone and when it gets it wrong,

01:08:19 ◼   ► there are consequences, bad consequences.

01:08:22 ◼   ► So that is something, I mean, there's no rumor

01:08:24 ◼   ► that like, you know, the Siri suddenly

01:08:26 ◼   ► could become chat GPT, right?

01:08:27 ◼   ► Apple needs to, it's kind of like what you were just saying,

01:08:30 ◼   ► things that the phone already does

01:08:31 ◼   ► just make it do them better,

01:08:32 ◼   ► but that is different than Siri is now

01:08:35 ◼   ► an all-knowing wizard that you can ask anything

01:08:37 ◼   ► and cross your fingers.

01:08:39 ◼   ► - Yeah, like I don't need Siri to become a chat bot.

01:08:44 ◼   ► I don't need Siri to generate images

01:08:47 ◼   ► that don't exist already for me.

01:08:48 ◼   ► Many of the things that we think of as modern AI apps,

01:08:53 ◼   ► I don't need Siri to do those things.

01:08:55 ◼   ► What we need Siri to do is what the promise of Siri

01:08:59 ◼   ► has always been promising and it has just never

01:09:02 ◼   ► reliably delivered on that promise.

01:09:04 ◼   ► That's what we want to do.

01:09:05 ◼   ► And I think, you know, looking at modern AI techniques

01:09:08 ◼   ► and models and things like that,

01:09:10 ◼   ► I think this gives them the ability to tackle

01:09:13 ◼   ► what really is a pretty old set of problems

01:09:16 ◼   ► in a new way that should theoretically

01:09:18 ◼   ► be substantially better.

01:09:19 ◼   ► So that's what I'm mostly looking for.

01:09:21 ◼   ► And everything else that we've talked about,

01:09:23 ◼   ► like being able to control apps without configuration

01:09:26 ◼   ► and stuff like that, that's all gonna be great

01:09:28 ◼   ► down the road whenever that comes together.

01:09:30 ◼   ► I don't even need that right now.

01:09:32 ◼   ► I just want Siri to work.

01:09:34 ◼   ► That'd be great.

01:09:35 ◼   ► And maybe give me a good captioning model

01:09:37 ◼   ► on the phone that I can use.

01:09:38 ◼   ► And otherwise, I think that could be enough for us.

01:09:42 ◼   ► It's not gonna be enough for the press,

01:09:43 ◼   ► it's not gonna be enough for the analysts,

01:09:45 ◼   ► it's not gonna be enough for Wall Street,

01:09:47 ◼   ► but I think that would be enough

01:09:48 ◼   ► for Apple's actual customers.

01:09:50 ◼   ► - I have an addendum to an earlier statement

01:09:52 ◼   ► that I meant to correct myself immediately

01:09:53 ◼   ► and then the time passed and I couldn't.

01:09:55 ◼   ► But talk about when they're dropping Intel support,

01:09:57 ◼   ► I forgot about all the AI stuff.

01:09:59 ◼   ► The whole requirement to have an NPU,

01:10:01 ◼   ► like that's why rewind doesn't work on my computer,

01:10:04 ◼   ► that may push the schedule forward.

01:10:07 ◼   ► I mean, that wasn't really a glimmer in anyone's eye

01:10:09 ◼   ► when back when we were talking about this years ago.

01:10:12 ◼   ► But now the whole like, you know,

01:10:14 ◼   ► the Co-Pilot Plus PCs requiring a 40 tops NPU,

01:10:16 ◼   ► blah, blah, blah.

01:10:18 ◼   ► You know, my CPU doesn't have an NPU in there, right?

01:10:21 ◼   ► And so every single one of these quote unquote AI,

01:10:23 ◼   ► if they bother to add quote unquote AI features,

01:10:27 ◼   ► yeah, they could just make them not work on Intel,

01:10:28 ◼   ► but they could all just not make Mac OS,

01:10:30 ◼   ► also just not make Mac OS work on Intel.

01:10:32 ◼   ► So I feel like that nudges the possibility

01:10:34 ◼   ► that this is the year my Mac isn't supported,

01:10:35 ◼   ► nudges it a little bit.

01:10:37 ◼   ► I don't think it makes it a slam dunk,

01:10:38 ◼   ► I'm thinking the sure thing,

01:10:39 ◼   ► I still think I'll probably be supported

01:10:41 ◼   ► and just won't have features visible,

01:10:42 ◼   ► but the LLM stuff is a problem for me.

01:10:46 ◼   ► - Can you reprogram the Afterburner card to be an NPU?

01:10:51 ◼   ► - It would still be so bad, right?

01:10:53 ◼   ► That is not, they're not spending any time

01:10:55 ◼   ► optimizing any of this stuff.

01:10:57 ◼   ► It's like, well, it runs in the neural engine

01:10:58 ◼   ► and also someone did an implementation

01:11:00 ◼   ► for like the SIMD instructions and the Xeon.

01:11:02 ◼   ► Nope, not happening.

01:11:04 ◼   ► - Yeah, it's not, sorry.

01:11:05 ◼   ► You're out of luck on that one.

01:11:06 ◼   ► - Yeah, so it's really just, I mean,

01:11:08 ◼   ► like I said, what Mac has going for it is like,

01:11:10 ◼   ► yeah, Apple probably doesn't care that much.

01:11:12 ◼   ► Oh, sorry, Intel Mac users,

01:11:13 ◼   ► you don't get these new features, right?

01:11:15 ◼   ► Just be glad Mac OS still,

01:11:16 ◼   ► we still ship it for your CPU.

01:11:21 ◼   ► And that, I feel like that has a likelihood

01:11:23 ◼   ► because Apple's like, yeah, it's just Mac,

01:11:25 ◼   ► it'll be fine, right?

01:11:26 ◼   ► But if they really, it was the old Apple

01:11:29 ◼   ► where it's like everyone on the same page,

01:11:30 ◼   ► they'd be like, no, we're cutting them off this year

01:11:32 ◼   ► 'cause we got so many of the AI features

01:11:33 ◼   ► and they all require the neural processor

01:11:35 ◼   ► and Intel doesn't have that, so tough luck.

01:11:37 ◼   ► We'll see, like I, the worst case scenario for me,

01:11:41 ◼   ► it's like, I was gonna ask you too

01:11:43 ◼   ► what you're most looking forward to WWDC,

01:11:45 ◼   ► but the thing I'm fearing the most

01:11:47 ◼   ► is that I won't be able to run this version of Mac OS

01:11:51 ◼   ► or the feature that I want in Mac OS

01:11:54 ◼   ► doesn't run on my thing.

01:11:55 ◼   ► So for example, the photos features

01:11:57 ◼   ► that use quote unquote AI to do cool photos stuff.

01:12:00 ◼   ► Granted, there's ways I can use other AI tools

01:12:02 ◼   ► to edit my photos, I don't need this to be built in,

01:12:05 ◼   ► but if I can't even access those features,

01:12:08 ◼   ► like if the person removal thing

01:12:10 ◼   ► is built into the photos app,

01:12:11 ◼   ► like this is great, saves me a lot of time,

01:12:12 ◼   ► it's right built in, I don't have to edit it

01:12:14 ◼   ► in an external editor, don't have to go to a webpage.

01:12:16 ◼   ► Oh, but not on Intel, that's a big fear for me

01:12:18 ◼   ► because even though I would still,

01:12:20 ◼   ► it's like, look, Mac OS still supports your system,

01:12:22 ◼   ► no problem, I'd be like, oh, but I really wanna use

01:12:25 ◼   ► the new AI powered photo editing features on my Mac

01:12:28 ◼   ► and now I'm sitting at my wife's computer

01:12:29 ◼   ► to do photo editing and she's complaining

01:12:31 ◼   ► that I'm hogging her computer and that hastens the timeline

01:12:35 ◼   ► for me getting a new Mac, but. (laughs)

01:12:38 ◼   ► Yeah, which Apple hasn't shipped yet,

01:12:39 ◼   ► like an M4 based Mac Studio or something,

01:12:41 ◼   ► but anyway, that's, I wanna throw it,

01:12:43 ◼   ► anyway, so what things are you most looking forward to?

01:12:45 ◼   ► Mark, you already said Siri, but is there anything else,

01:12:47 ◼   ► like speaking of developer stuff,

01:12:48 ◼   ► like developer tools, Xcode, Swift,

01:12:50 ◼   ► like what big thing are you looking forward to at WWDC,

01:12:54 ◼   ► assuming there's no hardware?

01:12:56 ◼   ► - So, I mean, for me, I really wanna see,

01:12:59 ◼   ► kind of what I was talking about last episode,

01:13:01 ◼   ► I want there to be good models that developers can use

01:13:06 ◼   ► on the phone for free with no limits,

01:13:08 ◼   ► like that's what I really want,

01:13:09 ◼   ► give us built in AI SDKs that we can just use

01:13:14 ◼   ► the same way we can use almost every other API on the phone.

01:13:18 ◼   ► Like that would be game changing in so many ways,

01:13:21 ◼   ► and so that's the big thing I want in terms of capability.

01:13:24 ◼   ► Now, in smaller ways, like, you know,