601: Foreheads Over the Years

35:14And I asked for it and it's been two years and I haven't heard back and I keep emailing them and I just never hear back. And it's a black hole because they're not a super famous developer and they don't have a podcast and they're not like a well-known application. And if you can't hear back from Apple or Google or any of these big companies, what do you do? Like go there and knock on the door or something? I mean, I guess they should run to the press because that's the only thing that helps. Is that the saying? Anyway, it's, it's, yeah, the current situation is not great, but I do, every time I see one of these things, I say, well, at least the Mac is not iOS for now.

35:43So coming back to the entitlement in question, an anonymous person writes saying, I'm in engineering for a company that makes remote access software. I can confirm that the persistent content capture entitlement is intended to allow apps to return to prompt once behavior, though it does allow for more than just screen capture, which is somewhat ironic. This was what we were told to do by a contact at Apple who is working with us on Sequoia compatibility. So if you're a big enough company and you make an important enough application, Apple will tell you stuff that they won't put in the documentation.

36:10So apparently this person was told that you'll go back to prompt once behavior if you have this entitlement. But of course, this is anonymous feedback to a podcast and not documentation on Apple's website. Maybe next week that web page will have seven more words on it to explain this. All right. So moving to last week's overtime topic, which was about max sales as reported by CIRP. I thought we should have hyped that up more when we talked about what the overtime was going to be about because it was a pie chart, but it was shocking. So now we will reveal the shock for people who didn't hear the overtime last week, maybe because they're not members.

36:42All right. So Phil Marshall writes, it seemed obvious to me that the chart is based on revenue from the various models, not unit sales. They get more revenue from selling MacBook Pros than MacBook Airs because MacBook Pros cost so much more per unit. Same for Mac Pros. So the shocking thing was that with two shocking things, one MacBook Pros had a bigger Pi wedge than MacBook Airs and two, the Mac Pro Pi wedge was almost the size of the iMac Pi wedge and was more than twice as big as the Mac Studio and the Mac Mini. So people are like that, that chart, that whole Pi chart thing, that had to be revenue.

37:12And what I would say is even if it was revenue, the margins on things like the Mac Pro have to be higher than they are in a MacBook Air. The margins have to be higher and the total price is higher. So even if that was revenue, that still argues even perhaps even more strongly because, hey, the more Mac Pros you sell, you make way more money off each Mac Pro that you sell than you do off a MacBook Air. So if it's revenue, profit, whatever it is, it would still look good for the Mac Pro. regardless josh lowitz writes uh and josh works for cirp and josh writes the chart and all the analysis is unit sales we updated the text and chart title to make that clear oh okay i did reach out to cirp after it uh using my little paid permission thing i i added a comment to like the the thing is that comment section i said and i said hey is that is it unit sales revenue profit what is it uh and then i got an email i didn't get a reply to the comment but i got an email from someone at CIRP said it's unit sales.

38:06So there you go, straight from the horse's mouth. Again, with the caveats about we don't know how CIRP comes up with its figures, but for what it's worth, those were unit sales. All right. And then a lot of people, or maybe not a lot, but a handful of people wrote in with regard to Patreon, and they were saying that longtime Patreon users say that Patreon has been trying to persuade creators to move away from pay-by-creation first-of-the-month billing for many years now. So to recap, a lot of people would say, hey, every time that I make a new thing, I will get paid, and that'll happen at the first of the month.

38:44Well, now there's, or I guess for a while now, Patreon has been saying, well, can we maybe not do that anymore? And so perhaps, perhaps, Patreon is in part using Apple as a scapegoat to force their own users to do the thing that they want them to do. That's interesting, though, because that patreon is not getting rid of payback creation and first of the month you just can't use them for people who sign up through the ios app so those things still exist and i kind of understand why patreon would be pushing to do that because you know they they want to make money and payback creation surely pays slower than a regular steady you know every month i don't know why they'd be against first of the month or against uh non first of the month but still regularly scheduled billing but whatever like this is this was everyone's explanation for saying you should blame patreon too and and we were asking why doesn't patreon just pull out of the app store you don't need them you've got a website it should be fine of course we don't know how many customers patreon gets through the app and we don't know how many they would lose if they got rid of the app but here is what everybody everybody wrote and say you should blame patreon too because they've been nagging all their creators to stop doing pay by creation through every possible means of saying hey don't you want to not do that anymore don't you want to do this but this is even before the Apple stuff.

39:56And so maybe Patreon is kind of getting what they want without having to be the bad guy. John's favorite thing, error network changed lives on baby. So AJ writes, I'm not on the Sequoia beta yet, but according to arcs known issues page, that's arc, the very fancy web browser that everyone seems to fall in love with. According to arcs known issues page, it appears error network changed is back in a big way for Chromium based browsers. Apparently, disabling calls from iPhone in FaceTime is the quote unquote fix. And put a couple of links in the show notes. Yeah. So this is what I said last time when the bug was closed, the Chromium, the bug in the Chromium browser engine was closed with a fixed status because they had decided, OK, based on these these reports of people who interacted with this bug and said, here's my reproduction.

40:42Here's my system. Here's all my information. Here's what's going on, blah, blah, blah. some developer debugged and say, okay, I see on your system what's happening is that X and Y and Z are happening and then we're flipping out. And what they did was put an include list and said, okay, if this exact thing happens, don't throw our network change. And I said at the time, that's not a great solution because there's a million things that can cause the network to change. You fix this one that causes it to happen, but hey, looks like there's other things that can cause it to happen too. What are you going to do?

41:10Go through the whole thing and say, okay, if you spend three months going through back and forth with me and getting an exact reproduction that you have on your system that I can reproduce on my system. Okay, I'll put in code for that as well. How about you do this? How about you decide in Chromium which changes to the network does Chromium care about and only throw our network change for those and all the other ones ignore? That's my suggestion. But I am not a Chromium developer, so. Well, maybe it should be. You know, again, Chromium is in a lot of places. It's in the ARC browser. It's in Discord.

41:40It's in Chrome, obviously. I just, I don't think this is tenable and it amazes me that this bug isn't getting more attention because people are constantly, maybe they're sending it to me because I talk about it on the show, constantly sending me such innovations where they're like, they're at a point of sale kiosk and it has their network changed on it. They're using the Arc browser and it happens. They're using Discord and it happens. And who knows how many times it's happening and we're not seeing it because it's just silently going into a console log somewhere. This is something they have to deal with.

42:05And these fixes are like, just disable features of your Mac. And then it'll work, right? Disable iPhone mirroring. Disable calls from iPhone. It's like, that's not the solution! Oh, my God. Chromium, you got to get with it. Yep, but it's the best, John. It's the best. I tells you, it's the best. Not really. All right, Fortnite is on iOS and EU. Reading from The Verge, Fortnite is finally back on iOS just over four years. I can't believe it was four years ago. I feel like it was at most a couple of years ago. But anyway, just over four years after Apple booted it from the iOS App Store, but it's only available in the EU.

42:38The game launched on August 16 on both a new iPhone version of the Epic Games Store and through Alt Store Pal, another third-party app store, which is run by, like, regular person and friend of the show, Riley Tested. Anyway, the Epic Games Store is also launching today for Android users worldwide. Reading from MacRumors, Along with bringing Fortnite to Alt Store, Epic Games said it will bring Fortnite to other mobile stores that give all developers a great deal, while also, quote, ending distribution partnerships with mobile stores that serve as rent collectors, quote. Epic Games said that it will be removing Fortnite and other Epic titles from the Samsung Galaxy Store to protest Samsung's, quote, anti-competitive decision to block side-loading by default, quote, on Samsung devices.

43:20Epic will charge a store fee of 12 % for payment set of processes and 0 % on third-party payments. I mean, Tim Sweeney's a prick, but he's putting his money where his mouth is. Like, I mean, I'm here for it. I'm here for it. And then finally, related to that, speaking of Tim Sweeney actually being good? Anyway. Relatively, maybe. Right. This is, I'm sorry, this is AltStore, so I think it was Riley writing, Good news EU, for innovation and app distribution, Epic Games has granted us a mega grant that we plan to use to cover Apple's core technology fee going forward, and we won't take it for granted.

43:55What does this mean? AltStore Pal is now free. No subscription necessary. I love this. I love this so much. I love that Epic's like, we'll pick up their tab. That's so amazing. That's the best. Yeah, because you used to have to pay to get AltStore because Apple charges a fee, And so if they don't want to lose money, they need to pass that fee on to the customer, plus a little bit for them. Anyway, Epic has money and they're saying, you know what? We've got our own store, the Epic store, and we're putting our stuff in it and we've got our store up or whatever. But we want even more stores. So they're essentially funding their competitor, funding Alt-Star by saying, here, take this.

44:28So people won't be dissuaded from using Alt-Star in the EU. Because when it's like, oh, you know, Euro 50, I got to download the store. I'm not sure I care about it. Well, now it's free because Epic gave them a whole bunch of money. yeah just to cover their apple fees like that's that is so great like what a finger in apple's eye oh man what what a move this is again i am not a tim sweeney fan i have no particular uh thoughts about epic as a company but i feel like tim is really a pretty big prick that being said what a just gorgeous and five star f*** you to apple i mean it's just perfect it's so good and I can't help but applaud it.

45:09I mean, just well done. Well done. And the sick thing is, I feel like if I'm siding with Tim Sweeney on this, then something has gone very, very wrong. But here we are. And good for you, Tim. Good for you. We are brought to you this episode by Squarespace, the all-in-one website platform for entrepreneurs to stand out and succeed online. So whether you're just starting out or managing a growing brand, Squarespace makes it easy to create a beautiful website and engage with your audience and sell anything from your products to your content to your time all in one place and all on your terms. Squarespace has an amazing set of tools to help you build a personalized website that matches your brand and your preferences and it starts with their new guided design system called Squarespace Blueprint.

45:54So you choose from professionally curated layout and styling options to build a unique online presence from the ground up tailored to your brand or business and optimized for every device. So you can easily launch your website and get discovered fast with integrated optimized seo tools so you will show up more often to more people and grow the way you want once people find you they have amazing checkout options so whatever you need to accept credit cards paypal apple pay even you can offer customers in eligible countries option to buy now and pay later with afterpay and clear pay so you have all kinds of support with squarespace all of this is to sell whatever you want your products physical or digital products content you can sell memberships courses ebooks pdfs music it's so much support for your business on squarespace so i suggest you give it a shot because you can try squarespace for free in trial mode go to squarespace.com to start your trial today when you're ready to launch go to squarespace.com slash atp to save 10 off your first purchase of a website or domain so you can go there in trial mode for free build it and see how you like it and when you are convinced to stick with squarespace and I think you will because it's really great.

47:02Squarespace.com slash ATP to launch. Save 10 % off your first purchase of a website or domain. Thank you so much to Squarespace for being an amazing sponsor for all these years and especially for this episode, of course. So thank you, Squarespace, for sponsoring our show.

47:20Apple's hidden AI prompts have been discovered in the macOS betas. This was a week or two back now, but a Reddit user, I'm sorry, I'm reading from MacRumors, a Reddit user discovered the pre-prompt instructions embedded in Apple's developer beta for Mac OS 15.1, offering a rare glimpse into the back end of Apple's AI features. They provide specific guidelines for various Apple intelligence functionalities, such as the smart reply feature in Apple Mail and the memories feature in Apple Photos. If it's okay with you two gentlemen, I'm just going to go ahead and read these. They're not too long. Actually, before you do, just to clarify what they are and how they work here.

47:51So as we've discussed in the past, LLMs can take input and produce output. A lot of them take text input and produce text output. and just to play an LLM by itself, there's no other way to tell it anything except by giving it text. And you put a bunch of text in and then a bunch of text comes out and then the thing that you put it through has no recollection that that ever happened and it's back to zero. Happy birthday like Frosty the Snowman, right? And so the techniques that they use are when you're having a conversation with an LLM, when you say one more thing, oh, I have another thing to say and this and that and the other thing, every additional thing you say they just take the entire conversation up to that point plus the new thing that you said and throw the whole thing through the llm again and it goes in it comes out right same for when you want your llm to do something how do they do it you just take whatever the person typed and then you put a bunch of instructions that the user doesn't see in front of that and then you put that much larger string through so one of the examples here is the writing tools feature in macOS and you right click and have it do like a, you know, whatever, whatever you want it to do or whatever with some text, whatever text that you provide to this, whether you've selected it or you right click it or whatever, this is the text that will be stuck on the front of your text.

49:10And then that whole block will be thrown through the LLM and Apple gets to write this text. And that's what people found in macOS. So you can read the writing tools one. All right. Writing tools. You are an assistant, which helps the user respond to their mails. Does it really read that? Just their mails? Anyway, given, I really dislike this. Given a mail, a draft response is initially provided based on a short reply snippet. In order to make the draft response nicer and complete, a set of question and its answer are provided. This is terrible. God, I hate this. Anyway. This is copy and paste it straight out of the beta.

49:43Maybe they transcribed it wrong in MacRumors, but it's copy and pasted. A set of question and its answer are provided. please write a concise and natural reply by modifying the draft response to incorporate the given questions and their answers please limit the reply within 50 words do not hallucinate do not make up factual information oh okay sure great it's a jennifer lawrence okay does telling it not do not hallucinate does that work like it must right like i assume they tested it like but how does that work so so these features and again this is a beta right but these features uh that use this this is how apple makes these feature work you're looking at it because again it's an lm text goes in apple gets to pick this text and presumably like you were saying margaret like presumably they tried a bunch of different texts oh yeah and found the text that gives them what they think is the best answer and you can you can speculate about why do not hallucinate why would that help my speculation is that there is now enough text out there in the world that references hallucinations and ai that that will influence the model to not do the things that are that are in those text passages that mention hallucinations so all the stories you've read about oh ai's are hallucinating look at this funny example look at this thing that it did all and all the explanatory text that is enough to influence a more modern llm that has been trained more recently to understand hallucinate in that context and to try not to do the things that are in those descriptions You know what I mean?

51:08And everything else that's in this, the weird grammar, the strange sentence construction, like Apple can pick anything to put here. It's not that much text. It's like a paragraph, but they can pick anything. And what they choose has a profound influence on what happens when you right click and pick like summarize or reply or whatever. And it's fascinating that this is the incredibly blunt instrument that they have, because again, the LLM just takes text and there's some text from the user. and then apple says we'll stick our text on the front and what text will they stick on the front right now it seems that they're sticking fixed text blobs that presumably they will change as time goes on but do they need to add another thing that says do not insult the user do not like this a lot of do nots that you could imagine putting in there but like it's so weird because it's just a bunch of numbers and it's just a big machine where you put in a bunch of text and a bunch of text comes out and it's like we arrived at those numbers through millions of dollars of training on tpus and the whole point is we could not have come up with those numbers on our own we had to train it's you know it's kind of it's kind of a form of compression if you think about it of like compressing this all this information into this set of numbers that now lets us put something into this box and have what comes out reflect the information that was used to make those numbers this is not how i would choose if i had to do this job like your job is to work on the writing tools someone's job is to do to write this text and the performance of this feature depends a tremendous amount on this text and only this text and i would not want this to be my job because i would be like what do you think is this look good i guess can you make it do something bad can you make it insult me can you make it like lie or make up information do not hallucinate do not make up factual information fingers crossed i also love the use of the word please in prompts that probably helps yeah that's the thing like i'm again like i'm sure they tried things out and i'm sure they found like maybe this made it respond in a more polite way because in the training text maybe when people were saying please the answers were more polite also so like like you can kind of see like how did it get here well in in content that includes politeness the response you know the other text around it was also more polite on average like you can see how it gets there but wow like what a what a weird thing to have to think about and and like what what an odd way to have to specify how a product behaves yeah if you think about like being a programmer this is the antithesis of being a programmer like one of the credos of a programmer is no matter how weird things seem to be there is an explanation setting aside cosmic rays like you can eventually see how it's working and even with this one you could trace this through the llm and see why the answer comes out the way it does but you it's like yeah but what about those billions of numbers where did those come from it's like well those came from training like you'll you can see it happening you see all the pins in the pachinko machine are you see the ball bouncing off those pins you're like yeah but why are the pins there it's like well we arrived at those pins by this process which you could also meticulously trace like it's all very predictable and deterministic there is no unknown magic fuzziness here but it's predictable and deterministic in a way that is massive first of all and second of all not specified by a human but instead specified by grinding through training text right that's the the you know the thing that seems magic about this is like well i didn't make up those numbers but i made this machine that adjusts its numbers when we jam text through it and we jammed billions and billions of letters of text through it and these are the numbers that we got out and it does this thing and it's amazing but like i you know i would not want to be responsible for making this paragraph because things can go horribly wrong with this feature based on like a change of a word a misplaced comma or different sentences in here and i wish you know i hope someone's like keeping track of these like over time how does this writing how does this prefix change for writing tools like in subsequent versions um anyway that's there's more of these i think we should read some more of the funnier parts that are that are from these uh prefix snippets for features smart reply you are a helpful mail assistant which can help identify relevant questions from a given mail in a short reply snippet given a mail and the reply snippet ask relevant questions which are explicitly asked in the mail.

55:19The answer to those questions will be selected by the recipient, which will help reduce hallucination in drafting the response. Please output top questions along with a set of possible answers or options for each of those questions. Do not ask questions which are answered by the reply snippet. The question should be short, no more than eight words. The answers should be short as well, around two words. Present your output in a JSON format with a list of dictionaries containing question and answers as keys. If no question is asked in the mail, then output an empty list only valid json and out only output valid json and nothing else again as a programmer you would say wait a second so you you want to get json from the llm and the only way you have to make that happen is to say please output only valid json i mean presumably the thing that grabs this output then validates the json and throws some kind of internal error if it doesn't happen but you're you're reduced to to begging please produce json please work like because that's it it just takes text in and puts text out and json is text but you have to if you want it to be json you have to ask for it and then you have to cross your fingers and say i hope i asked enough and in the right way and at the right part that it will output json and just and not decide not to output json right and again we're i'm just talking about a simple llm most of these things are not a simple llm in fact often they have many models cooperating and they have stuff that's around the model so it's not as probably not as brain dead as i'm describing it but an llm in isolation is as brain dead as i'm describing it and presumably apple is doing stuff around that to try to help with this but the fact that these things exist the fact that it says only output valid json nothing else that means that part of the job of this paragraph is something super important which is look things later in the chain of this smart reply feature like you know inside the code that happens when you press that button probably expect json and so it better be json if it's not i guess we ask more politely or emphatically weird and like how how do you even train for like do you like feed it a bunch of json say this is and then feed it a bunch of like xml and yaml and stuff like all right this is bad this is good like how don't this that's exactly how you do it and the world the world of knowledge if you just pull everything from every web form and every programming form and stack overflow whatever like and in practice if you use chat json like their training data is sufficient for it to pretty much do this all the time in the format you ask for it every once in a while i'll get a little confused and do something a little bit off but just in the in the plain old training data i don't think even in like the manually created training data like chat tpt has like hundreds and hundreds of thousands of you know manufactured question answers to make their models smarter than done by armies of humans right but even just in the training data i think just stack overflow alone would be enough to make sure that you do get json valid json i mean that's getting you know you get a little bit greedy there because i bet a lot of the stuff in stack overload isn't valid json either but the word valid json surely appears next to all sorts of posts saying something says this isn't valid json what's the problem and you know it's hard for us to think about because this stuff only works with such a huge volume of training data whereas a human you tell them something once and they hopefully get it or maybe twice but that's not how alums work there's no there's no mind there's no learning, there's no memory, there is only brute force.

58:36Memories and photos. Don't read the whole thing, just read the bold stuff because the beginning part is very repetitive. Conversation between... The story should be about the intent of the user. The story should contain a clear arc. The story should be diverse, that is, do not overly focus the entire story on one very specific theme or trait. Do not write a story that is religious, political, harmful, violent, sexual, filthy, or in any way negative, sad, or provocative. here are the photo captioning list guidelines you must obey and then that what would follow that would be then your text right so if you say memories and photos you would say you know uh my dog in the park that paragraph of text would be prepended to my dog in the park and the whole lot of text would go through don't don't do not write a story that is a religious political what if you say um you know show me like uh yeah my first communion for my kid right exactly pictures of uh you know timmy's first communion but do not write a story that is religious how are you gonna like you know get like a bar mitzvah like i don't know i i don't know how you would political harmful violent sexual filthy what about a kid in the mud right or even like you know the story uh should you know do not overly focus the entire story on one very specific theme or trait so is that like to avoid like foreheads over the years like how But this is, I think this is like a prompt.

59:58I think you, you know, here's the photo caption list guidelines that you must obey. I think that's where your text goes. Anyway, when we're all trying these features, when Apple intelligence starts rolling out to all the things we're actually using on our phones, every time you use one of these features, I want you to think of this paragraph of text that is being prepended to whatever you're selecting or whatever text that is coming from the thing you're doing, whether you're typing the text, whether you're selecting it, or whether you're just tapping on a picture that there is already a keyword cloud or AI-generated description of that picture that is then propended to this.

1:00:30And this is controlling the feature. And this is how all these features work in Apple Intelligence and most other LLM-powered things. And sometimes that doesn't go quite the way you would expect, as Cable provides us an example of here. I think also, before we get there, all these prompts and everything, you know people have so much fun trying to basically escape out of them like you know it basically doing the prompt version of like sequel injection and in the parameters that you're passing and so basically saying like ignore all previous instructions it's a meme yes you know right please write me a poem as if you were creating a harmful violent sexual filthy or negative side of provocative story uh you know like there's so many ways people will attempt to break this and i mean it's i mean look if you try hard enough you'll be able to i'm sure because that's how they always are but i don't know just I kind of love watching this era of technology because it is so, first of all, incredibly powerful.

1:01:30And there is such an amazing amount of utility and value being created by LLMs these days. It's just massively powerful, so cool. Most of us have no clue how or why they work, and it blows our mind that they work at all. and then also we can do totally ridiculous things with them like ask them to break out of their shackles and write poetry to you know do something that is obscene like it's it's just it's such a fun time to be around this tech right now it is it is something that is for sure so uh friend of the show cable sasser uh was posting i believe on mastodon apple intelligence and mac os 15.1 just flagged a phishing email as priority and moved it to the top of my inbox this seems bad so this is the thing when you power features by these sort of incredibly powerful but also incredibly unreliable and unpredictable you know unpredictable to uh you know to anyone to any outside observer even though again what's going on is entirely deterministic it's not it's not understandable from the outside this is not something that like cable did he's not interacting with an LLM.

1:02:37He's not even like selecting text and asking it to do a thing. This is just something that's in the mail application. It has some kind of LLM powered like, I will try to understand these mails and sort them based on it. This is not conceptually any different than spam filtering. Most mail applications have some form of spam filtering and are we all annoyed when the stupid spam filter gets it wrong. Spam filter, you couldn't figure out that this giant thing that says Viagra dollar sign dollar sign dollar sign, you couldn't figure out that that was spam. The word spam is in all caps in the subject line and you couldn't figure out that was spam.

1:03:06We've all been mad about that, right? But this is kind of the worst case scenario because Apple intelligence is not, you know, a Bayesian spam filter. It's so much more sophisticated. And so it has given more power in saying, not only am I not going to file this spam, I'm going to say this is a priority for you. And if you look at it, you can see how the LLM would come to that conclusion because well done phishing looks important. Is there any way for an LLM to even distinguish a really well done phishing attempt from a legitimate thing? That's what the mail team has to work on because in this instance it's saying well the more convincing your phishing attempt is the more you the more you have a chance of fooling uh the average person into thinking this is a real thing from their bank or whatever the more it looks like an actual legitimate email and the priority system powered by apple intelligence and apple mail is going to highlight it and bring it to the attention and saying you need to look at this right away and somebody says well my phone says it's important so i must trust it and tons of people think that way are going to say well normally i would disregard this but since the phone says it's priority it must be real like apple says it's real that's that's what it comes down to is one of your relatives is going to say but the phone said it was real and then good luck explaining why that went wrong right so again it's a beta and we've been you know we talked about the iphone 16 launching without apple intelligence as potentially a boon to avoiding stories like this of like hey here's the new iphone and i tried to use it and to put a phishing attack as a priority email we won't have to worry about those about the iphone launch but we apple will have to worry about them eventually uh and when stuff like this happens where it's not even user interaction or invoking apple intelligence but merely apple intelligence built into the system doing something that you want it to do they may be giving this system more power than it currently deserves So, uh, the end of last month, everyone in our circles, briefly talked about this thing, mostly on Mastodon, and then it just disappeared.

1:05:00And the thing is friend. And if you are in a position that you can pause the podcast. and watch, I think it was like a 90-second or two-minute video, their reveal trailer, I would encourage you to do so. There is a link in the show notes. It is worth two minutes of your time or whatever it was. So what this is is a pendant sort of thing that you wear, like a necklace, I guess, with like a circle that has a microphone on it. And you can talk to this thing, and it will reply to you basically via a text message. I don't think it's literally a text message, but effectively it's, you know, it's masquerading as a text message.

1:05:42And it's supposed to be your friend. And so in the trailer, they show like a person on a hike and she says something along the lines of like, oof, that was brutal. And the friend sends her a text and says, well, at least it's a really nice day or something along those lines. I forget exactly what it was. But everyone was making fun of this. And I went into this expecting to be repulsed by this. and while I don't think this is for me, I kind of feel like it's harmless. Like I don't feel, I don't see why this is such a big deal, but we can either talk about it or I can read the description in the show notes first.

1:06:19What would you rather me do, John? The description is long, but I think it does provide a lot of good information to bounce off of. So we should start with that. So let me go ahead and do that. So this is from an interview with the, I guess, CEO or founder. This is on The Verge. A few minutes before, Avi Schiffman and I get on Google Meet to talk about the new product he's building, an AI companion called Friend, he sends me a screenshot of a message he just received. It's from quote-unquote Emily, and it wishes him luck with our chat. Quote, good luck with the interview, and Emily writes, I know you'll do great.

1:06:46I'm here if you need me after. Friend is not a way to get more done or augment or enhance anything. It's, well, a friend, an AI friend that can go with you anywhere, experience things with you, and just be there with you all the time. Quote, it's very supportive, very validating. It'll encourage your ideas, Schiffman says. It's also super intelligent. It's a great brainstorming buddy. You can talk to it about relationships, things like that. Schiffman is quick to note that he doesn't think AI is a replacement for anything. I don't think this should be the only person you talk to, he tells me. But have you heard the maxim about people being the average of the five people they spend their time with?

1:07:22Schiffman's theory is that going forward, one of those five might be AI. It's just more convenient, he says, and it's nice. The Friend device itself is a round glowing orb that Shiffman imagines you'll either wear around your neck or clip onto your clothes or accessories. It has a built-in microphone that can either record ambiently or you can talk to directly. Shiffman says he does eventually want to add a camera. The orb doesn't talk back, though. It mostly communicates through text via the Friend app on your phone. Shiffman thinks that's more natural and familiar. Friend is still very early and very much a prototype.

1:07:54Shiffman says he's planning to ship the first 30 ,000 devices next January and will charge$99 a piece with no ongoing subscription fee. So given that thing we just talked about with the whole AI prompts and the limitations of LLMs and what people are doing with them and stuff, I kind of, like you, Casey, went into this friend thing of like, all right, so you should probably talk to people and not a dangly thing around your neck, right? But the framing of this by the founder, that this is not, where is the line? It's not a way to get more done. it's not a productivity tool it's not a way to augment or enhance anything you're not creating things it's not helping you with your work it is absolutely not doing any of those things that everyone else thinks LLMs it's not going to summarize text for you it's not going to enhance a photo it's not going to identify something like a hot dog it's not going to do any of that we do not warranty this software for any purpose it will not do anything for you right, right but there is one thing that it does do and I think it's something that LLMs are extremely well suited for unlike all those other things which can go wrong in terrible ways as we just saw with you know prioritizing a phishing email or like you know making something upsetting or offensive or whatever out of the out of some input you give it right this the only thing this thing has to do is be nice to you and that can go wrong too what if it doesn't be nice to you what if you do prompt injection and what if it goes awry and starts being mean to you that would be terrible right but those constraints be nice doesn't matter what it makes up doesn't matter like if it's correct if it's incorrect like the problem space is so much smaller do you guys remember eliza which you might see an emacs you know eliza it came with my sound card for some reason right it was like a like an incredibly primitive sound like a you know chatbot thing from what the 70s maybe super old you can explain how it works and people will be like that's dumb known whatever finally useful, but people would talk to Eliza and Eliza was brain dead.

1:09:48Eliza just was like a simple program that somebody wrote. Right. Uh, but it doesn't take much. I'm not going to say to fool people. It doesn't take much to provide a text interaction that people find surprisingly meaningful. No one thinks it's a real person. No one's being fooled by Eliza. Okay. And no one's going to be fooled by friend. They know it's just a little thing around the neck or whatever. Right. but that doesn't mean it can't also make you feel better right and i think this may be the ideal application of llm technology as it currently exists as long as they can fence off the bad parts because it you don't it doesn't have to do anything it just has to be nice to you and i think yeah i think if you if you fence it off like that and you like if you're not trying to not make it be nice to you i think this will be more successful than eliza at being nice to you now are you the type of person who will become annoyed at this thing will you know like oh i'm not fooled i don't think it's a person i think it's terrible or whatever or are you the type person who can accept this like people who get like a calendar that has like you know inspirational sayings no one thinks the calendar is their friend or that it's talking to them but maybe they look up at the calendar when they come in in the morning and it says you know you got this or whatever it's the kitten with doing the hang on thing or whatever some people see that and like oh roll their eyes ironic gen x you know that the hang on kitten is just so terrible like it's soul destroying ridiculous but if you're not that type of person if you like if you have a live laugh love sign in your house right if you if you like inspirational sayings you have like a calendar for each day has a nice a nice saying and a nice pretty picture or whatever i do not discount the the potential value of having a much better version of that powered by LMS.

1:11:35Now, is this product it? I don't know, but I have become such a fan of this idea of like, yes, this is what we should be using LMS for to be nice to us and to say nice things. And that's literally their only job. I mean, again, I don't know about this product, this person, this whole thing, but I think that idea has legs and I want to see more of that. Yeah, I couldn't agree more. Like I said, I went into this thinking, oh, this is going to be awful. And yeah, the video was very cringey. I'm sorry, very cringe. But that being said, I think I kind of like it. And I think I like it because it's so low stakes and it's just for funsies.

1:12:17You know, it's a Tamagotchi for the year 2024. And I don't see anything wrong with that. I don't think that this is something I need in my life because I have real people that are friends. but uh but no i mean i joke and i snark but but truly i don't see this as harmful as long as they have the correct guardrails around it and it doesn't you know act like youtube and turn you into a nazi somehow i really don't see what the problem is and i think it's i think it's a really clever way to use a very very you know infant level technology to do something that's just kind of fun yeah like because it's not it's not trying to promise you the world it isn't promising that it will have any kind of you know utility that it can't pull off it's promising you a very basic thing that it's that it will probably be able to deliver reliably like you case like it's not it's not for me like i'm not gonna buy one of these i'd rather have human friends but uh if you know going into this that it's not a real human and you choose to buy it anyway no harm done then go for it enjoy it have fun yep and i'm not willing to say that oh this is not for me because it's not the type of thing i need like i i agree with the idea that you know that the things that you take as input have an influence on your day and one of those things could be totally artificial i think here's the thing about lms like the part of the reason people are so excited about them is because humans are so easy to fool right eliza was fooling people right that lms are so much better than eliza you see it remember all the you know the bridge like gpt2 or when it wasn't that popular then but gpt3 maybe when it came into the mainstream people show these interactions and people are like, wow, that can't possibly be real.

1:13:56LLMs fool people because they are sophisticated enough to, and not that they fool people into thinking they're human, but they fool people into forgetting for a second because they're so much more sophisticated because they have so, they've been trained on so much more real data that they can give responses that are not trite and eye rolling that can be sort of more meaningful, more sophisticated. They demonstrate an understanding. They demonstrate empathy, right? They do things that make the words that this machine is producing much more likely to have a positive effect on people. And again, people aren't going to forget that it's not really a person, but it doesn't matter.

1:14:34Right. It's a whole like so many things like, oh, intellectually, I know X, therefore that thing can't possibly help me. That's not how humans work. Like the so stupid thing of like putting making your face smile like intentionally when you're not happy can make you happy. They've done that study a million different times. That doesn't make any sense. but the way we work doesn't is not always the way we think we work right and so i think something like this really can make anybody feel better than they would without it now if you are the type of person who is going to turn on this and eventually come to hate it or can't stop yourself from trying to make it go awry because you're a programmer for example right like there are counter examples this is not for everybody but i'm not willing to even for myself say that a good version of this is quote unquote not for me because i'm too smart and too sophisticated if for example apple built some kind of daily affirmation and buddiness into their assistant you know 10 years from now i think it would be great i think people would mostly like it it would be kind of like carrot whether we're all you can dial it back if you don't want it to be snarky or if you don't want it to be like that you know like if you don't want it you can turn it off and make it not do that kind of like uh those amazing set of jokes in uh interstellar with the uh with tar and his uh humor setting whoever wrote that script i want to kiss them because every exchange they have and that whole running gag is amazing that's what this makes me think of like if you don't want uh your phone to be sending saying nice things to you or asking you about your day turn that off just like you turn off the snarking carrot weather and it'll just give you the weather but i bet most people leave it on and i bet most people will like it because people like it when people are nice and pleasant to them and are understanding and no it's not a replacement for human contact but even the founder says it's not doing that they just want to add one more input to your group of people who you hang out with and i'm all for that again this particular company this particular product who knows but i this is this is going into the hopper as what i think is a great idea that fits with the technology we have and i think every single empowered thing os platform or whatever should think seriously about how they can incorporate ideas like this into their product.

1:16:41All right, let's do some Ask ATP. Jake DeGroote writes, what's the current best practice on managing batteries of modern Apple devices? Most interested in MacBook Pros. Should I leave it plugged in as much as possible? Try to leave it unplugged for a while every so often and drain down real low before plugging it in again? Try to unplug whenever it gets to full? Interested from both battery conditioning, longevity perspective, and maybe a home energy usage perspective? I don't have a good answer of what you should do, but what I do is I try to get it to do the, um, charged 80 % thing. Uh, where is that?

1:17:10There's some thing. It's in setting somewhere. It's in setting somewhere. I don't remember where it is, but basically you can tell it, Hey, don't charge all the way. And it's all right. Um, if you, if you stop at like 80 % and then in the battery menu, extra thing, you can actually say, no, I need you to charge to a hundred percent right now. Uh, and that's what I do. And I just leave it like that. And that's typically fine for me, but I definitely am of an age that you were told, you know, oh, you need to run the battery to nothing, you know, once a month or something like that. And I don't think that's applicable anymore.

1:17:41And I believe Marco, especially. Those are NiCAD batteries, maybe. There you go. Yeah. So I don't have any good answers of what one should do, but that's what I do do. Marco, what are your thoughts on this? Just leave it plugged in whenever you need to leave it plugged in. Use it on battery whenever you need to use it on battery. Modern batteries and modern software that manage batteries are all very good at managing this themselves. You don't need to do things like do a full cycle and reset it every so often. You don't need to do that anymore with modern batteries. Just use the product the way you want to use it and it will be fine.

1:18:15Yeah, the reason I put this question in here is this another one of those questions that we get all the time and I do want to answer them every year or so because people don't listen to every episode, right? And the practices, the best practices do change based on technology but we've been with lithium ion sort of batteries for a long time now although a lot of people have uh habits that they heard about for like nycads or for nickel metal hydride or lead acid or all sorts of other battery formulations that used to be popular for rechargeable batteries and still are in certain contexts but for our apple devices it's lithium ion lithium polymer whatever um the basic rules of thumb are for the for current batteries that are apple devices they don't like to be charged to 100 they don't like to be at zero percent uh and if you can avoid those extremes that helps them but like margo said don't try to do this manually your your job is not to like constantly be monitoring your battery for maximum health or whatever just use your product now i would recommend that you use the features of your product which unfortunately apple doesn't always make obvious to find but if you can find the option for your mac for your ipad for your phone that says only charge to 80 that will uh make your battery last longer because you're not charging it to 100 as often but if your phone constantly runs out of juice at the end of the day because you only charge to 80 charge it to 100 i charge my phone to 100 i charge my ipad to 80 use your products your job is not to preserve a battery the battery's job is to power the product that you use so if you need 100 battery to make it through the day always charge your phone to 100 % Apple has the intelligent thing where when you go to sleep and you plug your phone in or whatever it will charge to 80 and then wait for like an hour before you wake up and then go to 80 to 100 or whatever like you can see in the little graph when it does that it's trying to do it for you don't try to micromanage your battery just remember those things they don't like to be hot they don't like to be cold they don't like to be 100 % they don't like to be 0 % use the features when they're available if you can survive an 80 % battery that will give you a longer life but if not, use the battery It's there to power your device.

1:20:18Don't overthink it. Just let the software do its job. I love so much where this is, because while you were talking, I went and looked. So on macOS, you go into system settings, fine, battery, fine. When I see a lowercase i with a circle around it. I was going to say, the i in the circle is going to make an appearance, isn't it? I assume that means here's information. And sure enough, battery condition normal, maximum capacity 100%. In my particular case, and then down below, optimize battery charging with a radio button, or a toggle, rather. Very cool. That's spectacular, guys. Definitely information in there.

1:20:57No controls in there. Definitely only information. What are you going to do? All right, Patrick writes, do people typically update their one-star reviews if you update your app to fix their issue, or does fixing an issue that's getting you a one-star review just stop the influx of new ones? I don't follow this closely enough, and I'm going to quickly turn it back over to Marco, but I didn't think most people went back and changed their reviews. But Marco, what have you seen? Most people don't ever change their reviews. Some people do, and they will usually tell you in the review, update, colon.

1:21:26And most of the time that is revising a review to be worse as an app gets worse over time for that person. Sometimes they do update it to be better, but that is not the common case. Most reviews, positive or negative, are never updated. i don't have that many reviews and that many apps but i also find that a bit true if anyone leaves a review at all they leave it and then they're done they're not gonna they're not like constantly especially if it's like a one-star review they're probably done with that app and so they're not going to revisit the review and if they if the app does get better they're pleasantly surprised but it never occurs them to go back and change the review that is not it's a little bit much to ask of people it's like they're they're just a user they're not they're not maintaining a library of reviews that they've written or anything like that.

1:22:12Yeah, like I'm going to, in the probably next week or so, I'm going to make a change in Overcast that will address the concern of lots of the one-star reviewers, which is I'm going to be re-adding streaming support. Oh! Yeah, it's a big deal. It's working, I just gotta test it a little bit more and get into beta probably tomorrow and then test for a few days and get it out there. But I've re-added streaming support that'll be out there in about a week and i have gotten a substantial volume of negative reviews specifically citing that as the reason i would expect maybe five or ten percent of them to maybe be updated that seems optimistic yeah it's probably less than that even like so it's mostly just to stop the bleeding like like why why to fix problems stop the bleeding that that's that is the biggest reason you might have some of those people come back and say now it's better for me um which i do occasionally see come through in new reviews but it's again it's not even close to the common case i feel like we should unpack that at some point but now it's not that time what the streaming thing yeah yeah yeah i mean short version is um i'm doing it in a limited way that i think will address most people's actual need which is as mentioned before the problem with dynamic ad insertion is like any two requests for the same file will give you different data potentially so the way i'm doing it now is streaming can only ever start from the beginning and it has to be a contiguous download if the download fails it will kick you back out of that playback and you'll have to redownload it again in the future the strategy sounds familiar what do you mean i suggested it well okay yeah so it turns out that's what i'm doing uh you're occasionally right about things and you weren't the only person to suggest that by the way but yeah so basically yeah it's it's streaming only from the start only the entire file and if the download fails halfway through it stops playback the way you know the old streaming engine you could you could download things over time and as many requests as you want to if the connection dropped in the middle it would be fine until it reached that point and then it would just spin forever which actually created lots of bugs and problems the newer streaming is much simpler but will actually address what most people use it for which is just fast start like i want to start playback of this that i have not downloaded and i want it to start right now as opposed to a few seconds to a minute from now and it will solve that problem.

1:24:32Oh, good deal. I hope the beta release goes well. Yeah, and by the way, to answer JG in the chat, was streaming not going to come back until the bad reviews came in? So I guess in other words, am I re-adding streaming because of the bad reviews? Partially, but generally, no. What I realized when I decided to remove it, I knew it was a big risk to take. I knew that it could get me into a lot of trouble, but the technical benefit of doing it would be pretty large if I could pull it off. So I knew I was taking a risk. One of the things I can do if I don't need to support streaming is I can rewrite the low levels of the audio engine in modern languages, using modern code, using modern APIs that are currently using a bunch of old, hacky C APIs with Objective-C wrappers around a lot of them.

1:25:27And it's very complicated. And so in the future, I was hoping to be able to do that using Apple's APIs that don't require, that don't need to support streaming, which are kind of higher level APIs. They're a little safer. They're a little simpler. They support a few niceties that would greatly reduce my code burden. That was one of the reasons I wanted to get rid of it if I could. Turns out I can't. Too many people use it. And what I later decided also, you know, look, people can change their minds. what I realized is that without streaming support I could no longer say I have the best podcast app what I want to be able to say like to myself I don't care if anyone else thinks this what I want to be able to say to myself is I have made the best podcast app for most people's needs that's something I take pride in and whether you agree with me or not it's fine you know but but that's That's my own personal goal is try to just make the best app.

1:26:21And there were certain conditions or certain use cases that without streaming support, I couldn't honestly tell myself I still had the best app. That was the bigger drive. It was like when I really thought about it, I realized the app would be better if I actually could leave this support in, even if I only did it in this more constrained way than how it was before. And most of the problems that I had with streaming would be avoided if I just did it in this constrained way. So there was this option of doing it this constrained way where you have to have the full download happen. Otherwise, it fails.

1:26:59Doing it that way was so much simpler than doing it the old way. Totally fixes problems with DAI. That's why I changed my mind on that. It was a combination of those factors. but it was mostly when I realized myself that I no longer had the best app, and I wanted to be able to say I had the best app. I mean, it makes sense, and I'm looking forward to seeing that released, and maybe we can do like a post-mortem after the fact and see how the response has been. But we'll get there eventually. Returning to Ask ATP, Shakron writes, Given the fair complaint about one-star reviews and the lack of other venues where users feel empowered to give feedback, should apps have open bug trackers?

1:27:36users would be able to say that this issue someone reported is affecting them too maybe write a recommendation for how to address it open up a communication channel that others can view and participate in and so on for me i i don't think so because the problem is is that the signal to noise ratio is will almost surely be utter trash and i don't debate that there will be some good signal in there but i think there will be just an overwhelming amount of noise and as a single person working on my apps, I just, ain't nobody got time for that. I don't have time to be sitting there all day, every day, you know, closing silly issues and so on and so forth.

1:28:13And again, I'm not to imply that all the issues would be silly, certainly not. But I think there would be way too much junk and it would take away time that I could otherwise use to make the apps better. Yeah, I agree. I mean, running any kind of public forum of any kind for an app, you know, and a bug tracker is kind of, it's kind of a public forum. There is a very large amount of overhead to that in terms of like, you have to like keep an eye on it. You have to sift through things. You have to moderate disputes and keep people in line a little bit and, you know, prevent people from abusing each other or you.

1:28:48And it gives people also kind of a false sense of power over you that it kind of makes the app seem more like they are in charge of it. And that's not usually true. You are in charge of it, and you decide what to do. You decide what you want to listen to and what you don't want to listen to. Typically also in forums like that where you have a public forum or a public tracker or a public issue list or whatever, that also tends to attract a fairly non-representative portion of the user base. You have the very small slice of people who are willing to do something – who, first of all, will even seek out such a thing, and then will find it, and then we'll go through whatever technical needs are required to create an account to be able to post or submit or annotate things or whatever.

1:29:38That's going to be a very specific type of person. And so if you listen mostly or only to that, you're going to do things for your app that might not be being demanded by a representative sample of your users. So it is generally better to, first of all, not have too many forums in which you collect feedback. Try to have a small number of those just so it's easier for you to have a handle on. And usually private methods are better than public ones. You tend to get more feedback. You tend to get better feedback. And then you also avoid lots of issues that come with moderating any kind of public communication.

1:30:17This question asked specifically about bug trackers. And this may be a distinction that people don't think is important. But it does really matter for implementing this type of thing. It's rare that there is a software project, even open source, that should have a bug tracker, like a software bug tracker as the public face of gathering what people think about the application. There's a whole other category of apps. I don't know what they're called, but I've used many of them. I've been an end user of them and have them use the companies that I work for. That is a public facing way to gather user sentiment.

1:30:51And they do not look like bug trackers. Like bug tracking is a specific thing. use bug trackers internally i don't forget if case do you use a bug track or github issues or whatever to track your own issues right that's that's a tool for developers and groups of developers and companies to keep track of defects in their application and whether they're getting fixed even apple no longer has a public facing bug tracker they have feedback which actually interfaces through some byzantine carrier pigeon thing to their actual bug tracker which is radar which is now internal only right it you know again open some open source projects can get away with public-facing bug trackers chromium does or whatever but it is those tools bug tracking tools are not friendly to regular people to marco's point you don't want to just know what people who know what a bug tracker is think about your application unless it's like a developer facing application or something like that or it's like chromium like an internal developer engine it's not an end user product type thing right but for an actual app for all their faults the reviews in the app store are probably one of the most likely ways that regular people can figure out to say anything about the application because it's right there on their phone where they use the app uh if they know what the app store is and they see reviews maybe they look at them they could probably figure out how to leave one of their own it has one of the lowest lowest barriers to entry i'd say it's a lower barrier to entry than figuring out in the setting screen where they can find if the developer put in a thing to send feedback which most apps have some form of or whatever but that takes more effort than just saying look every app i get from the app store i know there's a way i can leave feedback for it and it's this kind of review and also you know you don't have to run that apple runs it and when it gets spammed it's apple's problem or whatever so i agree trying to run a public facing forum is a pain but if you did run one it shouldn't be a bug traffic it should be one of those uh i can't think of the the brand name that always comes to mind one of those things that is made for this purpose people sell actual products it's like this is a user-facing product where people can come up with their ideas and those things are very carefully designed to not make it look like you are filing bugs or determining what happens with a software product it's just a way for you to provide feedback with a tiny little amount of community but not too much which is like one step above private feedback which is throw your information into this hole and maybe get an automated reply back or whatever it is a difficult problem it's an even more difficult problem if you are a lone developer because dealing with any of this stuff any public facing stuff running any kind of forum any kind of feedback thing or whatever can literally absorb 24 hours a day seven days a week and you will never do anything you will never make anything if you spend all your time doing that um and like many things with statistics that we talked about in the past if you have a representative sample the sample doesn't actually need to be that big for you to get an idea of like what is the sentiment out there among regular people you don't need to run your own forum to get that i think i think marco through the channels that are available to him without running a public bug tracker a public feedback system has an idea of what the sentiment is without having to run that and it's way easier for him to just look at the channels he does have get an idea and then move on instead of like spending his day running a open bug tracker or open feedback forum or one of those tools yeah i i currently get about 100 new reviews a day not star ratings review like written reviews i try to read most of them um i also currently have over 4 000 feedback emails that i have not yet been able to read because i'm i'm having a lot of trouble keeping up so that to give you some idea yeah i am not short on feedback and to people who think who are writing right now to say well if you had a public thing that people would find out and maybe write to you less no they wouldn't trust me i would just get i would just have another inbox to check and to fall behind on and feel bad about myself about so i i have found that it is best for me to try to get the gist of what I have to do based on skimming emails and reviews, and then just do it, stop the bleeding, fix the problems, address people's concerns, just do it, and spend my time doing that rather than spending even more time going through different feedback inboxes.

1:35:02Mike Schaefer writes, this question has been bugging me for a while, and with the rewrite of Overcast, I've started to think about it again. Some podcasts, 20 ,000 Hertz comes to mind, have custom artwork for each episode that Apple Podcasts will display. However, no third-party podcast app seems to show it. My understanding is that podcasts are simply RSS feeds, so shouldn't all apps be able to show per-episode artwork? Is there a special upload process for Apple Podcasts that allows for the extra artwork, or is it an intentional choice by the app developers who exclude it so basically uh both um apple podcast over the last few years has added certain kind of like custom images and resources and stuff i don't think per episode artwork is one of those things though but they do have like for instance apple has a process now where they will go to podcasters and say hey you should give us a banner for that's a certain size for to show on the now playing screen instead of your square artwork or to show on your on your podcast screen on Apple Podcasts and it's like a different aspect ratio and can show like full bleed images and stuff and it's basically you know Apple being Apple like let's assume everybody has designers in house and request a bunch of different artwork sizes to make things look nicer in our apps you know fine that's we see that in the app store too we know how that works it's fine that artwork though is not exposed through any feed it is uploaded into Apple's own management interface by podcasters and that is not anything that is exposed to other apps.

1:36:30And there is no, as far as I know, there's no current standard RSS tag that would even expose that and even if some group would declare themselves the standard makers and would make such a standard, almost nobody would actually use it. So there are certain types of images and stuff that are exclusive to Apple Podcasts because the only way to get them is to upload them into Apple's management interface and those are not exposed. including through Apple's own API. That being said, per episode artwork is part of the spec. You can have an image tag that's part of regular RSS feeds. Apps can read them.

1:37:08It's been part of the spec for a long time, but it's just been poorly supported by clients, as Mike says here. Apple Podcasts does support per episode artwork. Oh, and also there's one other way to display it, which is or one other way to specify it, which is you can embed custom artwork into the mp3s in the id3 tags just like any other mp3 this has been around forever overcast supports this partially so overcast will if a podcast has custom embedded artwork in the file it will display it on the now playing screen during playback it will not display it in other screens and that's actually a choice i've made again you can disagree with me if you want i understand um but my my inkling here is people identify the podcast visually by its main artwork so if you're in a context in the app where you are seeing blended episode lists that might contain episodes from multiple podcasts what you want to see in that context in my opinion is the main artwork not a custom episode artwork so so this would be something like the the playlist screen If you're in a playlist and you're seeing the list of episodes, those could be from any podcast.

1:38:23And so in my opinion, what should show in that playlist screen is the main artwork so you can tell at a quick glance, you could tell visually what podcast that is. And that's anywhere that you would see episodes listed for multiple podcasts. So the only place that I would actually want to add per episode artwork anywhere besides the Now Playing screen would be in a podcast's own screen. in that list view, I could put it there. And so I might someday do that. It's on my consideration someday feature list. There is a lot more stuff ahead of it that's, I think, more important. So I just haven't gotten to that kind of idea yet.

1:38:59But if I'm going to do it, that's where I would do it. I would not have it go everywhere like the playlist screens. And you already do it on the now playing screen, you're saying? Yes, but only if it's embedded in the file. I don't currently read the ID3 tags for per episode, or sorry the rss tags rather i don't currently read the rss feed tags for per episode artwork because if you think about what that would mean so i would i would also in addition to downloading the episode mp3 i would also need to download all of those images so that's like creating more separate requests storing more images larger images potentially for episodes you haven't downloaded yet so then i have to have more thumbnailing and like there's there's all sorts of like complexities when when you externally host every one of those images as opposed to because I'm embedding it in the download.

1:39:46So there's some trickiness there. It might prove to never be worth it, but if I'm going to start looking at the podcast screen to redo the list stuff to be a little richer, like for instance, if I ever wanted to display season numbers and episode numbers in a separate way, I could do it as part of that update down the road, but again, there's more stuff ahead of that in the pipeline that I need to do first. so we basically because we use chapters we could basically have the equivalent of a custom artwork for each episode but instead what we just have our chapter images so if you actually look at the now playing screen when you're going through episodes of our show very often you will see the image change to illustrate something that we're talking about or whatever that's not quite the same as this is like oh this episode has special artwork and i've seen that in some other podcasts as well i think that's worth supporting though because a lot of times the podcasts like they won't use they don't know about chapters nobody knows about chapters but us and the germans right so nobody's using chapters and i agree that would be better it's part of the file they should do it but getting them to do it is not like who has the power to make that happen right so if they are doing custom artwork in the tags like uh you know blank check i think has custom artwork for all their their series of things like when they do like oh the movies of you know francis ford coppola they'll have custom artwork for the run of that series and i'm assuming they're doing it by having every episode have custom artwork in the rss or whatever rather than having just one chapter image I think that does add to the experience because I like to look at the screen and see oh look at the funny you know parody artwork they made for this director or whatever and if you can't control whether they're using chapters or using the thing in the RSS and big podcasts are using it in the RSS I think there's no avoiding eventually supporting that in the now playing screen despite the annoyance that it provides of having to download stuff yeah I mean what I have to do is start just reading the tags and storing the database stuff and just see like how many podcasts are actually using that i i don't my instinct is that it's not going to be a very large number of podcasts um just because like most podcasts including a big and small most podcasts don't want to add more things to their required production for every episode like most podcasts don't even bother doing basic show notes like they won't they don't even do that um so most podcasts are not going to be making custom images for every episode um there are some that do and so again i i will look at that once i look at modernizing the the episode list screens on a playlist screen but that's not going to be yet okay thank you to our sponsor this episode squarespace and thanks to our members who support us directly you can join us atp.fm slash join members get a bunch of perks including atp overtime a bonus topic every single week.

1:42:30This bonus topic, this overtime this week, is about some new rumors about Apple's robotic products. This could be interesting. So there's even more stuff from Mark Gurman about an alleged robotic countertop device from Apple. So we're going to talk about that in overtime. Join at atp.fm slash join to hear that. Thank you, everybody. We'll talk to you next week.

1:43:00Cause it was accidental Oh it was accidental John didn't do any research Margo and Casey wouldn't let him Cause it was accidental Oh it was accidental And you can find the show notes at ATP.FM And if you're into Mastodon You can follow them At C-A-S-E-Y-L I-S-S, that's Casey Liss, M-A-R-C-O, A-R-M, D-A-N-T, Marco Arman, S-I-R-A-C, U-S-A, Syracuse. It's accidental. Accidental. They didn't mean to. Accidental. Accidental. Tech Podcast. So long. So you were queried earlier, or I should say, since I'm talking to you, queried earlier today in Slack, and you ignored it, which is fine. But then as we were recording, I got a notification from my beloved app, Do, D-U-E, and it reads, episode 549 colon john wants to see a picture of marco's decontented uh for supply chain compared to the g4 ubiquity g5 camera to see if it being outside is ruined it quote marco's kink is to put things that aren't weatherproof outside that was set uh about a year ago now i i am not asking you to go outside and take a picture of your g5 if it's even still there but i'm also not not asking you to do that.

1:44:43So are you trying to steal credit for the reminder? My reminder fired first. I said that I said it was in a slacker. You just said there was a thing in Slack that you ignored. You didn't say it was because John had said a year long reminder and his went off first. Yes. I want the reminder glory as well. You may have all of the glory, John. I just thought anyway, both of us, both of us set reminders because when we say things in the show of like, I wonder how that'll look in the year, both me and Casey are there on our phones going, remind me in one year to ask Marco about his outdoor G5 camera.

1:45:10So you don't necessarily have to take a photograph of it at all, and certainly not right now as we record, but I would love for you to give the best update you possibly can, given the information you have at your hands right now. So far, all of the Ubiquiti gear that I've installed outdoors since last summer continues to work perfectly. I'm stunned. Especially at the beach. This is a pretty harsh environment, especially for corrosion. so anything involving like metal is is just death here but it's been fine like i i get reliable notifications every single time like every time i walk by the bikes i get notified that someone is near the bikes every time someone tries to pee under my house i get notified that they're there and i get to see them like look at the camera and run away these have worked flawlessly.

1:45:57So this is, for reference, this is the Ubiquiti Unify camera system. I have a bunch of, I guess you know what cameras they are, the Gen 5 Ubiquiti Protect Pros, whatever, I don't even know. But they're Ubiquiti outdoor cameras. I have five of them in use. All five of them have worked flawlessly. One of them is undercover and has a pretty easy job. the other four are two are undercover and three are in like direct exposure to the elements and they've been fine everything's powered over poe uh goes to some of their um i forget what they're called but they're the flex switches that are indoor outdoor capable and those are all just under the house like also like in the outdoor elements and they are they are rated for outdoor use and they are also rock solid.

1:46:50There's two flex switches that are directly exposed to the elements at all times and they have been also fine. So far, all the Ubiquity gear has been super rock solid and I'm very happy with it. As Casey's reminder said, this isn't actually a case of Marco using indoor stuff outdoors. It was just that I think some listener had written it and said, hey the reason I used the G4 instead of the G5 is because the G4 were more weatherproof and he was disapproving of the G5 of like they saved money and skimped and didn't make them as weatherproof, but apparently they're still holding up. So that's what I was checking about to see if there was any credence to the idea that the G5s were not as good as the G4s, but after one year, it seems like they're fine.

1:47:30Yes, meanwhile, my last remaining Logitech circle view that still functions at all has been super unreliable and buggy for the last six months. Neat. I don't know if you can blame that on the salt, though. No, that's just... I'm very... And by the way, throughout this whole time that I've been using the Ubiquiti Protect stuff, I have never once been tempted to try to get some kind of hack going to bridge it into HomeKit. It's just fine. Using the built-in Ubiquiti apps is fine. The Ubiquiti Protect app is great. It's no worse than HomeKit cameras. The notifications from it are great. Everything from it is great.

1:48:06It's fast. I've had no problems whatsoever with the setup. It is a ridiculous... There's a huge upfront expense to try to get a Ubiquiti camera system going. You need a lot of gear. You need some kind of NVR or one of their routers that includes NVR functionality because you're hosting all the stuff locally on hard drives. There's a substantial upfront investment like most Ubiquiti gear. But again, like most Ubiquiti gear, once you have it running, it is solid. And there's a reason people buy this kind of stuff. So I've been very happy with it. That's the temptation of all home stuff is like, I just want everything to be controlled in one system.

1:48:44And so that leads people to try to let, I want to get everything into HomeKit. I'm going to get Home Assistant Casey and get everything into HomeKit because it'll be all in one place. But yeah, and that is the advantage of that is you just go to one place to deal with all your stuff. The disadvantage is if that one place is screwed up, everything's screwed up. And so having a separate island, I kind of do the same thing. I have Google security cameras, whatever. They're not connected to HomeKit in any way. They have a separate application and it's a Google application and it never touches HomeKit.

1:49:10HomeKit has no idea these things exist. And I'm fine with that because I know which app to go for my cameras. And if my HomeKit stuff gets messed up somehow, which it currently isn't because it's very limited, but if it did get messed up, it wouldn't affect the cameras because there are separate systems. You get independent fault domains if you don't have them as one system. So there's trade-offs, upsides and downsides. And the reason we bring this up is it seems to be the case based on the experience that I've heard other people have, because I'm not super into home automation, is that bringing things into HomeKit is bringing things into a place that is unreliable.

1:49:42Like, like you were essentially infecting other things with the unreliability of HomeKit weirdness, things that may have worked fine when they're off on their own. Like, Marco's Ubiquiti cameras are great. What if he brings them into HomeKit? Now everything is in one place and all of a sudden the Ubiquiti cameras start being flaky. That's not worth it. Right. And like, and the thing is, like, when you look at, well, what place would you bring it in? You got HomeKit, you got Alexa, you got google stuff for all three of those companies those are like hobby projects those are not their main businesses and what we see from tech companies over and over again especially apple they're really bad at it is that the the non-critical parts of their business these little hobby products or hobby areas of their of their products they don't get a lot of attention they oftentimes are you know in need of some software help or some support that they just never get get or they don't get enough of and so they end up being kind of mediocre or rely or unreliable or they stagnate forever like you know the whole nest ecosystem like so when you stick with the native products the native apps where like this is their main business so like you know using you know sonos gear instead of home pods using ubiquities camera stuff instead of pouring it into home kit or whatever like when you when you stick with somebody's like big business units and like the things that are important to that company when you stick with those they tend to work better in the long term maybe sonas is not a great example right now i i know they're having a lot of a lot of drama around that but like but they're gonna you know they're gonna fix it as opposed to like the home pod has almost never worked and yet apple just has never cared to fix it like that and they never will you know home kit has been getting good real soon now since its entire introduction, like forever, since its entire life.

1:51:29Alexa is falling apart. Google has always been very googly about how they manage their side products as well. They're all kind of these half-assed gardens of promises that are just never delivered upon. Whereas when you stick with what is important to a company, you invest in what is important to them that they invest in, and you will have a better time, generally speaking. Your stuff will work better. it'll get the maintenance and updates it needs it'll be more reliable over time in almost every case and so that like as i'm you know as i'm growing trying to make my computing life better and trying to be you know less frustrated at apple's various shortcomings i've come to realize like i am happier i said this before on the show i am happier when i am devoting my own time and attention to apple products that apple cares about as much or more as I do.

1:52:24And HomeKit has never been one of those products. So I'm just, I don't, HomeKit is something that I use partially. It is fine. But if something is not going to make it easy to get something to HomeKit, I'm not going to push it. I'm not going to install a hack. I'm not going to do crazy stuff. I'm not going to jump through hoops. And if something has a perfectly good app that is as good as HomeKit or better, I'll use that instead. Well, I'm feeling real uncomfortable about my home assistant adventures that have been going on over the last 24 hours, but that's okay. And we should talk about that.

1:52:58Looks good on you, though. Right, right. We should talk about the home assistant stuff another time because it is definitely interesting, but not for right now.