Google AI Studio Deep Dive: From Vibe Coding to AGI with Logan Kilpatrick

29:18Well, some other potentially more niche use cases that are kind of interesting. So legal professionals are either extremely cautious about AI or they're not even fact checking the cases referenced in an AI output before they go to make an argument in court. So how would someone like that, where they're like saying like, hey, like I'm maybe not coding, but I need it for a very specific use case. How could they use AI Studio and Gemini and how would you advise someone? Not in a legal setting, but just like for a niche use case like that, you know? Yeah. One, you should definitely make sure that your law firm allows usage of these tools.

30:01I have maybe a slightly different perspective, which I talked to the CEO of Case Text, which was one of the early sort of successful GPT-based products. and they had early access and they did a bunch of stuff. They ended up selling their company to Thomas, uh, Thompson Reuters, uh, which is like a massive, uh, legal firm. If folks, folks have done legal stuff in the past. Um, and in that conversation with, uh, with Jake talking about like the, the actually how ubiquitous the penetration of AI is in the law in legal today, which was really interesting and not like my, what would have been my conventional wisdom default assumption.

30:46Uh, but they, thompson reuters ran a bunch of i think it was them or one of their partners ran some studies and it was like 99 of lawyers had tried gemini claude or chad gbt um which is crazy and i'm sure they there's like probably some bias sampling i'd assume the like absolute number is probably slightly lower than that um but even in a small subset of people like it's incredible to see how much usage they're getting i think there's a bunch of reasons for that in the legal domain like you know access to you know they have money to spend their high potential roi of like making operations go faster because they're you know uh digging through case law stuff like that that you can only remember so many you can only process so many in a search you can exactly so i think it's one of those examples where like the use case is just like so mission critical to what those folks are doing that like it just becomes easy to buy in um yeah and yeah i think like what one of the biggest points of feedback from from them in the conversation was just around like how much long context matters for this use case and gemini has obviously been at the at the bleeding edge of this with our 1 million token context window and 2 million um and it's been interesting to see how much that still comes up as like a limitation for them um is they they just want long context to bring more context and more documents and more information into the memory of the model.

32:16So I think, and obviously we're still early in that domain, but I think it'll be cool to see how much people are accelerated when you like 10x the context window or 100x the context window and things like that in the future. And it's very distinct from RAG in a lot of ways. Like I think if folks have gone into the weeds of rag versus long context, it really is a fundamental trade-off that you're making. So I'll be interested to see people not have to make that trade-off in cases where their use case would support it. Do you think that that's something that's realistic near-term timeline? I mean, without giving away anything.

32:56What do you think? There's a bunch of like architectural challenges like llms in the current form are not designed to scale up to the the 10 to 100 million token context window like it's it's really tough like you could do some hacks to sort of get slightly farther and like we did show a bunch of like research of what it would look like to bring 10 million to people and even with the original gemini launch showed some of that in in like practice and production environments it becomes extremely like very very very costly um and like not easy to maintain and continue to scale up so i do think we'll need some like architectural uh innovation at the model level in order to enable things like 100 million tokens um which i'm excited about and i think the world needs so i'm hopeful we'll keep pushing the rock up the hill 100 million tokens man what's a 100 million token use case yeah i I mean, some of these code bases is actually a good example of like, if you look at like a large company and if you're using a mono repo, really interesting to see like, if you imagine like your assistant, like you probably at maybe 100 million tokens is like too much or is like slightly on the extreme of this, but like accumulated through your lifetime, you actually do have a lot of this data.

34:16I think the challenge then becomes like, how do you, and like the attention mechanism in language models, um, and transformer specifically doesn't have this intrinsically in it, but like, how do you up sample the right data and down sample the wrong data, all that stuff. So I think you'll need these sit, like human memory has this like interesting, uh, has all these interesting mechanisms to make sure like the stuff that's not useful is removed from your memory. And the stuff that is useful is cemented and then you build these this sort of abstraction layer on top of like core ideas that you've had I assume we'll get something closer to like what looks like memory because then it it decreases like you don't need a hundred million you probably need like 10 million with the system on top of it and that's much more reasonable than 100 million totally that makes sense food analysis is a thing that really interests me this the food analysis via image it seems super practical and how does this work is this just simply take a photo of your lunch and get detailed nutritional information from that photo yeah this depends probably a little bit on like the use case you could definitely do something like that um i would i would suggest folks to like try to do some validation as far as like how accurate that is and i imagine there's cases where it works really well and there's probably like foods like different food genres or categories or different examples where it does not work well.

35:45Yeah. So this is where having, making it easy for people to build evals and benchmarks is really important. I'd love to see like your personal eval of different food understanding use cases. But yeah, I do think I've thought about this for a while. If you've ever like tried to use one of those apps to like track what you're eating um it's a huge pain in the ass i'm like i i envy people who are going on that journey and that mission because it's such a pain to have to like keep track of everything and being able to just snap a picture and have the model deeply understand um and use things like bounding boxes and it can do it the size proportional understanding of size and all the like nuanced details um has just never been possible without language models that have like a deep vision capability and i think that's that's something that works like actually works right now and is a good example of like the model capability has actually in some cases like uh outpaced the ability for us to build products around the model and the ability of people to like wrap their head around the understanding of the model can actually do that like if you were to go and ask, you know, a hundred million random Americans can, does a product like that exists?

37:03Can AI models do this, et cetera? Like I would assume the vast majority would say it's not possible. Um, and the reality is it works right now today. Like you can go download an app that does that or build an app yourself even better that does that. I'm guessing it would probably where you'd hit limitations would be with things like it, not being able to recognize the difference in maybe almond milk and dairy milk. No, I think it could do that. I think it could do that. I mean, in a cup without the packaging, it wouldn't be able to do that. Yeah. In containers, I mean, the models can read the text extremely well, like superhuman ability to read that text.

37:38I think you'd be able to do that right now. So I'm curious. I have almond milk and regular milk in my fridge. I'll go test after this, but I'm almost positive it'll be able to pick up on it. That's awesome. Well, I am actually doing that journey right now. I'm using an app to track my calories and it takes a lot of mental effort every day to put it in. And I was thinking like, man, if I could just go make a micro app version of this. And then, you know, I also read that it's really good to instead of just walking 30 minutes every day to actually switch every three minutes between regular walking and fast walking.

38:13And I was like, there's no reason with today's AI tools that I couldn't make myself a micro app to do both of those things for me like right now. Do it in AI Studio, Grant. Do it in AI Studio and let me know if it doesn't work and send the feedback of whatever doesn't. But I have to imagine this is like super achievable, single prompt, and it works out of the box. Yeah, I think so too. I will spin it up and I'll let you know how it goes. yeah i think the thing we're missing right now in ai studio for the like keeping track of what you eat example is we don't have um we don't have like persistent database support right now which is what you would want you'd want to store it over time and then build a bunch of infrastructure to retrieve that context where it's useful so we're working on that i think that will make that use case it will like the demo will work really well you could do it for like a bunch of stuff in a single like browser session but as soon as you refresh that's gonna uh you're gonna lose it yeah you're going to potentially lose that stuff.

39:07So we will, we'll solve that problem. But I think the technology should like exist. The models can do it for you, which is awesome. Well, Google, you, Google also has a lot of different other tools. Like you have Firebase Studio, you have some other ones, Jules, Colab. I mean, Firebase Studio in theory could, could do the database portion of that. No? Yeah. Yeah. It should be able to, I think that's one of the cool things about Firebase's and Firebase Studios. You get the sort of batteries included with, compute and storage, et cetera. So that should just work out of the box. And they also have a bunch of Gemini integration.

39:39So you should be able to get it to do sort of image understanding with Gemini through Firebase, get a database, all that stuff. Well, you know, Google's been releasing a lot in even the Gemma family. I think it's over 100 million downloads now. How does the open source strategy fit with AI Studio? Can people access those Gemma models within the platform? And what's the thinking behind the difference? Yeah, yeah. That's a great question. So what ends up happening with Gemma is sort of it's an acknowledgement in a couple of ways. Like one, obviously open source and open models are good for the ecosystem.

40:20So it's great for Google to participate. and Google's done a ton of stuff in open source for many, many years. I think the Gemma story is also an acknowledgement of it's us meeting developers and customers where they are. And there's a ton of ecosystems where you either need to have the level of model customization that open source provides or the stability of long-term deployments. I was emailing with a customer over the weekend and they were saying how they basically need more than six months to get a model approved to actually put it into production because they're in a regulated industry. And at the pace of what ends up happening with the Gemini models, it ends up being difficult for them to like use Gemini in production because like we have so many models and then the old models get deprecated and new ones come out and all this stuff.

41:11And like this, the pace of innovation that's happening actually makes adopting AI difficult. And that's where Gemma is like a great example of like, you can take the models, you can put them on a server somewhere you can deploy them and you can own that yourself long-term and sort of take take ownership in the journey of like how you migrate between different ai uh models that are coming so i think there's i think there's something interesting there there's two other angles one of them is um i think i think people also just want to have this like customization story and there's like a few a few good examples of this like we did dolphin gemma um and dolphin gemma is like trying down a bunch of dolphin data and sort of helps, helps understand some of the stuff that's happening, um, that the like actions that dolphins are taking, which is like kind of a funny example, but also like interesting of the level of customization that you can do.

42:03Um, med Gemma is another one where there's like a huge amount of interest in the medical space. And obviously open AI is making a bunch of, um, a bunch of noise about that with GPT five and the stuff that they're doing. So like, it's clearly an example that I think they said it was like some, one of the highest traction query use cases in chat GPT was this medical use case. And I think, you know, I talk to customers all the time who are blown away with MedGemma and MedGemini, actually the sort of Gemini variant of it. So it's been super cool to see those use cases also get love by being able to like build these custom open source models.

42:43And it usually ends up being that the models the open source models are like six months behind um the proprietary models for for google at least um yeah but it's all the same innovation that ends up being in there it's like we you know make hill climb on multimodal or long context or reasoning or something like that and then it all feeds back into the open model ecosystem what are you most excited about right now in the AI space? Like, where do you, where do you think all this is headed? Not necessarily, you know, giving away Google secret sauce or anything, but just as someone who's been in this space through your open AI days and now at Google DeepMind and watching it expand and grow, what, what, what gets you excited?

43:28Yeah, it's a good question. I think probably vibe coding. And the reason I, the reason I say that is I think as someone who's like gone through the pain of learning how to build software. Um, the, the reality is it's like the, it's, everyone isn't going to learn how to build software. Um, but the cool thing, and like, actually, if I coding is just one of these examples where like AI is lifting is like the rising tide that lifts all boats. Um, and, and we see this and, and, and that's why I think some of the like broad conversations about, um, you know, the, the negative, I think there are very real potential negative impacts of AI for people.

44:09But I think it's just so incredible and exciting to see this like continued trend of the tide lifting all the ships. And I think vibe coding is one of those examples. I think there's something interesting about this medical use case, which is really great. I think there's like the tutoring use case, which is really great. There's all these examples where consistently the model is making it so that something that was reserved for a small group of people is now able to be done by a much larger group of people. And the direct outcome of that is like hopefully a benefit to humanity. And I think there's...

44:45I agree. Yeah, it's fun to see the Vibe Coding example because I love building software. And I think it's like a great force for economic prosperity and all that stuff. And I'm hopeful we'll see other examples like this where something that is super high economic value to the world can be done by more people. And at the same time, this is the like the very interesting dovetails, like at the same time that more people are able to do it, the value of the expertise actually isn't diminished and it's further increased. And I think that's the missing part of this like vibe coding developer story, which is like just because you can vibe code doesn't make it so that developers aren't important.

45:23Actually, the amount of software in the world is going to be a million X what it is right now in 10 years. and who do you think is going to be fixing and maintaining all that software? There'll be AI in there and there'll be agents in there, but there'll also be a lot of human developers who are sort of orchestrating these tools and making sure that there's actual value creation happening. So I'm just incredibly bullish on this market expansionary force that AI has and also just creating all these new markets that didn't exist before. Yeah, it's incredible to see. If someone listening has never touched AI Studio, what's the first thing that they should use it to build?

46:02Clone your favorite website. I feel like as a use case, I do all the time. So I take a screenshot of AI Studio. And when I'm ideating, I'm like, what is the next generation of our product look like? I literally just take a screenshot of AI Studio, put it back in AI Studio, and I say, clone this. And it remakes it. And then I have this sort of canvas to reimagine what the product could be. So I think it's a magical use case. I like that. That's cool. I don't know what this says about me and that you can read into this all you want, but my go-to is always remake MySpace. So I guess MySpace is my favorite website, which I didn't realize.

46:39Yeah, that was a special tell. Yeah. But what would you say is the most underrated feature of AI Studio that like knowledge workers, for example, who aren't developers, but what are they missing from not using it? Yeah, I think maybe the knowledge worker persona probably should be in the Gemini app. So we're definitely building ASGDO for more of like the AI builder audience. But I still think like deep research is a great example of like people don't understand. My dad texted me over the weekend and was like, hey, he was sending me some article and was like, hey, does Gemini have this like deep research functionality?

47:12I'm like, yes, we actually invented deep research. and uh and it's it's just a mind-blowing experience if folks haven't tried deep research before in the gemini app you can ask your like a really simple query and the model will go visit you know between tens and hundreds or even thousands of different web pages in order to help answer the question and then like show you this proof of work and citations and all this stuff so it's a it's a mind-blowing experience and is a great example of like what ai products should feel like um so that's definitely the like my knowledge worker best uh best use case yeah that's great and yeah it's gotten a lot better too since the first version came out that yes google invented we remember because we covered it it's so funny it's like yeah people only use or they don't only use open ai but they they think of it with open ai now because open ai sort of like took that and ran with it yeah 100 the ai space has a lot of that that was never one of that that just really make business coverage more interesting yeah i think that's why telling the story is also really interesting because it is um i think i think there's something fascinating about that it's like how much of the uh even in ai how much of the how much of the human experience is just storytelling um is fascinating yeah i look forward to watching this era's pirates of silicon valley uh if you remember the the movie about bill gates and steve jobs and uh i think there will be a lot of interesting stories to tell 20 years from now about this stretch of time we're both in and entering.

48:46Well, Logan, thank you so much for joining us today. We really appreciate it. It's an honor to have you on the Neuron and to get to talk to you about the great products you guys are putting out. Yeah, this was a ton of fun. Thank you for having me. And it was fun to ideate in a bunch of stuff. And hopefully as the sort of vibe coding experience in AI Studio video materializes, we'll be able to help you both do even more and send me suggestions and feedback. I would love to keep making the product better for y 'all. Absolutely. And keep up the fire tweets. Every time you have one, I'm like, let's see it.

49:18Let's go. Let's see it. I appreciate it. That was a great interview. A lot of fun chatting with Logan today, learning about what Google's doing now and maybe what all's around the corner and just kind of seeing more of the big picture of the Google AI plan, even, I would say, not just limited to Studio. And I think we're going to see a lot more good things as time goes, don't you, Grant? Yeah, 100%. I'm really bullish on Google right now. Yeah, they've got a lot going on, and they seem to, after a rough rollout in the AI space, they've really kind of come together and started putting out some big winners.

49:54Yeah, I mean, I say this a lot in the newsletter, but Google was always the one to be in AI. Yes. OpenAI was founded basically to make sure that Google was not an AI monopoly. And so far, they've done a good job of succeeding at that. There's now five major AI labs that are competing directly at Google. That's crazy. But I still think it's Google's to win. I know it's kind of a controversial opinion. Yeah. But I truly do believe that it's Google's race to lose. Let's put it that way. They started further behind, but they were already ahead. But, you know, back in 2016. So I think at this point, it's just like, let's see if they can overcome the bureaucracy and get really smart people like Logan and Demis and let them push through and create something really epic.

50:45Well, you know, and to me, the smartest pivot I've seen in their system was like, they really started out very consumer driven. It was Gemini. Let's get these models out. And I mean, I guess you've got to have the models to continue building out the other things. But while OpenAI has a really strong hold on the consumer market, I mean, Google is growing in other areas. Like in the end, I don't even know that there has to be a winner and a loser. I feel like this is going to be a big space. It's going to continue to be a big space. And the fact is, we're probably going to find, and I hope we do find, that there's room for several companies.

51:26Because, you know, nobody wants the Monopoly situation. But I do believe that really, and not just them, anyone. You know, you don't want any one company controlling all of the AI space. That's true. I think, you know, I think there's still an awful, awful lot of room to grow for, you know, years and years to come. And they're all hanging in there right now. So yeah, we'll see where it goes. I would say the framing with which I said it's Google's race to win or lose is based on the framing of AGI being this like takeoff moment where, you know, all of a sudden some lab has the AGI internally and then they can just continuously self -improve forever.

52:10I think the jury's still out on whether or not that's going to happen anytime soon. But interestingly, I just recently saw Francois Cholet and Dworkesh Patel talking about their AGI timelines when they released the launch of the ARC AGI 3 prize, which for anyone who doesn't know, ARC AGI 3 is basically like the next iteration of the ARC prize to try and prove that, you know, AGI is possible. and that these models aren't just like memorizing things and they're actually generalizing and learning across, you know, large spans. Multiple disciplines and yeah. Exactly. And this one's all games focused, which goes back to our conversation with Logan about gaming and his conversation with Demis on the Google podcast I reference.

52:57We'll include a link to that. Yeah. And they basically are saying, you know, Francois, he's like a major AI skeptic just because he thinks, you know, they're not actually generalizing or at least they haven't up until this point, he's now changed his time frame to be within five years. So that's a big deal. Yeah, it's really interesting because we talk about it an awful lot less and people who were skeptics who have kind of suddenly leaned in and you've got people who are really close who you wouldn't have been shocked if they said tomorrow, who've maybe leaned back just a little bit. So I'm guessing what we're going to look at here is, you know, My money's been on like 2027-ish for a minute.

53:41But we'll see. You know, like you said, it may not even happen. And it may be that some of these people are bright and it's actually kind of irrelevant because it's already good enough it's doing most all of the things. You know, I mean, are we chasing perfection or are we, I mean, obviously we are over the long term. But, you know, we'll just have to see where it lands and how that goes forward. Yeah. But on that note, how about a round robin question for this week? If you had to predict, what will AI development look like for the average knowledge worker in 2026? We'll all be prompt engineers.

54:18We'll all be managing teams of agents, managing teams of agents. Do you think the tools will be invisible and we'll just be talking to our computers via some kind of voice interface? What do you see? hmm this is interesting so i think this is all dependent on what the dominant ux and by ux i mean user experience is and what it becomes over the next couple months so are you thinking 2026 just like broadly like over the course of the year so like the course of the year not like january one because we're you know five months out from january one so i won't i'll uh let's say you know this time next year a year from now okay so august 2026 yeah i think there's a lot of room to run in terms of creating a more optimal or a better ai user experience between now and 2026 august 2026 and so if i had to guess what that form factor is i think it'd be a lot more voice and talking and a lot less hard formatting structured prompts is my guess.

55:27A hundred percent. I think that that part of the process is going to get abstracted away as much as possible. And by that, I mean, it goes to the back end. The, you know, the engineers are doing it so that you, the user, as a regular knowledge worker, don't have to think about, ooh, I got to make sure that I give it my task and my goal and I have to structure it in a certain way. And it's better if I put, you know, my goal first versus my context first versus when do I put things in the order? I think that gets abstracted away as much as possible, unless the model architecture changes a lot. And so then it'll just be you, Corey, basically saying, what's up?

56:03This is what I'm trying to do. What do you need from me to do it? Or the AI even prompts you. It's like, what are you trying to do today, Corey? And here's what I need from you in order to do it. And it makes it a lot simpler. I like that. I'm very, very similar. I would say that really structured prompting is going to be a thing left for building agents. And mostly I say that because, you know, with agents, you want really repeatable results. You're looking for a task that's done over and over in a similar way. And I think that for that purpose, that style of prompting is fine. I will say that I've noticed even with GPT-5 and the little bit of time we've had access to it, I can just throw slop garbage in a chat window now.

56:51It does not have to be structured, grammatically correct. You know, even playing with image gen models lately, like I've been playing with Flux locally, and one of the things I noticed is like there are these sample prompts in there to get different styles of images when you start. and every one of them is just a non-stop line of words there are no commas there are no periods there are no nothing it's just like black and white image nor style three-quarter view from the ground that has a lot to do with how the the images are labeled like if you look at you know the data sets for how they train images you're basically asking a human or in certain cases an AI to go in and describe all of the aspects of the image so that it's properly labeled so that everything that is black and white, you know, is labeled black and white.

57:46And so the model knows, OK, if you use the word black and white, I know what that looks like and vice versa. So for image prompting, it makes perfect sense. I think more that there are no commas or natural breaks included is what threw me off, that it is literally just a long string of words. Word soup. with his word soup like you could grab any chunk and be reading nonsense if he didn't know what you were doing so i say that because it impressed me because what that means is it's still able to grab the context from that and know oh he means black and white he means no are this he means you know he wants a clown and a tutu dancing around on top of a beach ball uh that's not what I was generating images of, to be clear.

58:28But now, with that said, on the end user side, like a typical knowledge worker sitting at a desk, person who's spending their day in Slack or Google Chat or Microsoft Teams or whatever your particular flavor of communication software is, I think it will be much like what you've said. I think we'll get in in the morning and we'll open up our AI of choice. whichever one that happens to be. And it really doesn't matter that much. It matters less every time a new model comes out. Not at this stage, yeah. Yeah, we've really reached a point where the basics are good to go. So, you know, just pick your flavor and be happy with it.

59:12But I picture coming in in the morning and it being like, hey, here are the few things I'm supposed to give you every morning. I got this ready for you. I got this. You've got these meetings. You know, kind of a true sit down and have a five minute stand up with your with your personal assistant kind of thing to the morning, but with AI. And and I think then it'll be things like, oh, they talk to me. I'm going to take over this new task now that Bob has been handling. OK, so I'll holler over here and say, hey, Gemini, chat GPT, whatever. And it'll be like, hey, what's up, Corey? and from there, you know, you'll say, like, I've got this new thing I have to do every morning right now.

59:52I need to be keeping a really close eye on the price of NVIDIA stock, for example. Can you make sure that I have it before market open every day? And that'll be that. And then it just sends it to you, like, the next morning? And then it's just done, and I don't have to go in. I don't have to build an agent. I don't have to go and, you know, set some timer and a schedule. It just knows what every morning means. I mean, it can come back and be like, you mean weekdays or seven days a week? But I think it'll be much more done with either our voices or one or two simple questions. This is really interesting.

1:00:28I think the thing that unlocks this is having an AI operating system. So I'm going to use Apple as an example because Apple, although they are very behind in AI, they can actually make a lot of progress and headway here. And the rumors that have leaked from the company basically suggest as much. So they're working on a new Siri that's supposed to come out in early 2026. Spring 2026 might be summer, you know, could be as late as August or fall of next year. Or December. Or, I mean, don't bet on timelines with them. No, no. But it's allegedly spring of 2026. And they're working on the Siri so that it works with App Intense.

1:01:11So you could be using your phone or your MacBook and you could ask Siri, this new Siri, hey, Siri, can you get me, you know, can you open Uber and order me a pizza through the Uber Eats app? And it will actually do it for you. Now, that's something that Apple can uniquely do because they have the operating system. They like they can, you know, teach their AI to open your app with App Intense, their feature to do this, you know, put in all the data. Yeah. And be able to do this. can an open ai do this at this point they can do it if it's like a staged environment that they control out in the cloud like the chat dpt agent but that's not as convenient for everybody the most convenient form factor will be an ai assisted operating system that where it can actually just use everything locally on your computer for you and i think over under how many companies would you say are working on one of those right now out of the out of just out of the ai labs let's say just out of, and I'm going to include Microsoft here as well.

1:02:11We've got Google, Microsoft, OpenAI, X, Y, Prophec. You know, I think if they're not working on this, they are behind. I think so too. But I think a lot of them, like let's look at Cloud and OpenAI, for example. They don't have the money or attention span to work on this right now because think about who they would be competing with if they were making their own operating system. And I'd say it's not on Meta's radar either. Meta, I think, is closer. I think Meta is. And I think they're working on it for their personal super intelligence for the glasses. So if you think about the world that we just described, where you and I are working, yep, you got them on right there, where you and I are working and we're talking to our AI and the AI is doing things for us, Meta actually has the closest version of that with the super intelligence plan.

1:02:56I personally predicted that if they're able to do this, that means that there's less time, you and me stuck staring at a screen on the computer and more time us doing work as we're living our lives which not everyone agrees with with my read on that and that like you know the forces that be in corporate america will stop that from happening but i think if the everybody is is doing it then the expectation changes and it's not like you have to be glued to your office chair in order to work but that it's just like you're you the expectation is the work is done and i don't think that's wrong. I think it might be a longer leash timeline.

1:03:34And the only reason I say that is because it's the truth. You know, it's kind of happened now. Like, first off, we're both working from home making this podcast right now. Like, that was kind of crazy talk 25 years ago. Correct. Five years ago. Actually, three years ago after COVID when everyone wanted everyone to go back to the office. And I theoretically have the equipment and would be perfectly capable of right now traveling to Southern California and, you know, sitting on a beach and doing my work from a lawn chair or a beach chair. What are they? I don't even know the words. A chair on a beach, chair on a beach.

1:04:14And, you know, I could do that right now. I could go sit on my back porch right now and do my job. I think the key will be, would it cut us loose from the laptop? I think getting away from the big screen is going to be a big step. And it might be some solution like combining AI with an Apple Vision Pro kind of thing where, you know, you could do it from your couch. You could do it from anywhere where you just take a minute and throw your monitors up on the wall when you need to, you know, once every hour or so, check in on Slack and make sure you're caught up with everybody or something, for example.

1:04:47And by the way, Apple is working on this as well. They've waited till, we covered this at one of my old newsletters. They waited, they're waiting until 2026, 2027 for battery life reasons as has been leaked. I think Apple's still a sleeper here. I think they're I think they may do it by acquiring someone, which I don't think is necessarily a bad move. So far all they've done is put out papers that made us all really sad. I'm sorry. But I say that. They've done a great job with the ChatGPT integrations into Apple OS that have been really handy as well. All right. Well, I guess that about wraps us up for today.

1:05:27Special thanks to Logan Kilpatrick and Google DeepMind for joining us today. There are a lot of exciting things coming out of Google right now and all of the industry. And it was really fun to sit down and talk to them and learn some things we could and should be doing with them right now. And you should, too. If you enjoyed today's episode, please take a moment to like, subscribe, and drop us a note below. Also, please don't forget to check out the Neuron's daily AI newsletter, where we break down the latest AI news and developments for a half million people every morning. And that's all we've got for this episode.

1:06:01We'll see you next time. Farewell for now, humans. Bye.