LOGAN KILPATRICK & AMMAAR RESHI (Google) - Google AI Studios

=== LOGAN KILPATRICK & AMMAAR RESHI (Google) - Google AI Studios ===

(04:31:41): Thank you for joining.

(04:31:44): Thanks for having us.

(04:31:46): Yeah, so excited to have you.

(04:31:49): Maybe give people,

(04:31:50): I mean,

(04:31:50): I think people probably know both of you,

(04:31:53): but maybe give people a little bit of background,

(04:31:55): who you are,

(04:31:56): where you're coming from,

(04:31:57): and what you're going to talk about today.

(04:32:01): The same idea.

(04:32:02): Yeah, so Amar, elite product and design on the AI studio team, worked super closely with Logan.

(04:32:08): Love basically all things AI and hacking on AI for the longest time.

(04:32:14): Since 2022,

(04:32:15): first chat GBT moment,

(04:32:16): I think I was messing around with it and then code got better,

(04:32:19): video got better and so on.

(04:32:21): Did you get soft canceled for making a children's book?

(04:32:25): Oh, Dan, you had to bring that up.

(04:32:27): Uh, I thought it was amazing.

(04:32:30): Like I thought, I love, I loved that project.

(04:32:33): I'm on your side.

(04:32:34): I'm on your side.

(04:32:35): Logan's got a copy.

(04:32:38): This copy almost ended up in the trash can about three times in the office.

(04:32:43): And I think every bin,

(04:32:44): he looked at it and was like,

(04:32:45): maybe,

(04:32:45): maybe this one,

(04:32:46): uh,

(04:32:46): but yeah,

(04:32:49): great to,

(04:32:50): great to be here at Logan.

(04:32:51): Yeah.

(04:32:52): And my name's Logan.

(04:32:53): Um,

(04:32:54): I work on AI studio,

(04:32:55): the Gemini API and a bunch of developer products off inside of deep mine.

(04:32:58): So really focused on helping people build with our models, APIs, five coding, all the fun stuff.

(04:33:04): So excited to excited to five code with folks today.

(04:33:09): Fuck yeah.

(04:33:09): What are you going to show us?

(04:33:11): We got some cool stuff, and the best part is that Amar and I love to wing it.

(04:33:14): Amar, I have some examples.

(04:33:16): I think one of the things that's definitely top of mind for me these days is vibe

(04:33:21): coding,

(04:33:22): interactive experiences around data.

(04:33:24): I think there's so many people who practically do this in their job.

(04:33:29): I feel like there's lots of you can build anything,

(04:33:31): things going on in the world right now,

(04:33:33): and lots of tools that help you do that.

(04:33:35): I think this data story is actually really, really interesting and

(04:33:38): helps us continue to bridge into these new areas where there's less saturation in the market.

(04:33:44): So I've got some examples of that.

(04:33:47): I think we have some cool stuff that we want to show for some of the upcoming Vibe

(04:33:50): Coding stuff that we're working on.

(04:33:52): But what else do you have, Coach, Damar?

(04:33:54): Yeah,

(04:33:54): I've got a thing that one thing I think has been pretty awesome is that the

(04:33:58): beginning of this year,

(04:33:59): we've barely been in Figma.

(04:34:00): We've just been making a lot of our prototypes and UI work in AI Studio,

(04:34:05): inside of AI Studio,

(04:34:06): which has been really fun.

(04:34:07): It's really changed how the Teams works as well.

(04:34:10): So I'll show a bit of that,

(04:34:11): like what our workflow is,

(04:34:13): kind of the things you'll see upcoming,

(04:34:15): both in the UI.

(04:34:17): And then to top it off, I saw the ex-algo was recently open source.

(04:34:21): So just before this call,

(04:34:23): I was like,

(04:34:23): let's make an app that just like gamifies that whole thing.

(04:34:29): So we'll also see how that came out while we're on this.

(04:34:33): Amazing.

(04:34:33): And before I guess before we get started,

(04:34:35): one thing that I would love for you to you guys to level set us on.

(04:34:38): So we've we've been going for we've gone for a few hours now.

(04:34:40): We've gone through claw.

(04:34:41): We've gone through codecs.

(04:34:42): We've gone through a conductor,

(04:34:44): a bunch of different tools like tell us about AI studio for people who have maybe

(04:34:48): used a little bit or maybe are new to it.

(04:34:50): How do you think about is like at Google,

(04:34:53): a studio as as an entry point into AI when you use it,

(04:34:55): when you don't use it,

(04:34:56): what do you think it's good at all that kind of stuff?

(04:34:59): Yeah, I would take, and then Logan, I'd love your take as well.

(04:35:02): Yeah, so I think AI Studio, I think about it in a couple of ways.

(04:35:05): Like one, it is the place to discover all of Google DeepMind's latest models.

(04:35:09): And then the playground is effectively where we have bespoke experiences so that

(04:35:13): you can see the best capabilities,

(04:35:15): you can effectively tweak every parameter,

(04:35:17): and then you should be able to take any of that to code.

(04:35:20): So our principle in the playground is whatever you do,

(04:35:22): you should take it to code and you can take it to code.

(04:35:25): Then the vibe coding side of it is,

(04:35:27): Super easy to make AI powered apps with Gemini with just a single prompt.

(04:35:32): So we automatically stitch up the API for you.

(04:35:34): We figure out how to set up like Nano Banana or any of these other things in your app.

(04:35:38): And all you have to do is just ask for it.

(04:35:40): So you don't have to worry about any of that kind of setup and stuff.

(04:35:43): And that is only going to get better.

(04:35:45): How do I think about using it versus like anti-gravity, for example?

(04:35:50): Yeah,

(04:35:51): the way I think about that,

(04:35:52): and I'm curious to get your take on this one,

(04:35:54): is I think of that as like,

(04:35:56): that's the full-blown IDE.

(04:35:57): You want all the bells and whistles of an IDE,

(04:35:59): and you want something closer to cursor and all of that control.

(04:36:03): Like, I think antigravity is like a great set of things for that.

(04:36:06): for us it's like we're really going after folks who are uh you know entering by

(04:36:10): coding for the first time or are more casual builders uh and we're simplifying it

(04:36:15): you know we're making it approachable and easy and we're bringing you all the power

(04:36:19): without the complexity uh we don't ever want to necessarily be a full-blown ide uh

(04:36:23): look

(04:36:24): Yeah,

(04:36:25): and I think one tactical comment is we'll make it super easy as something we're

(04:36:28): working on to go from AI Studio to Antigravity,

(04:36:30): which we're really excited about.

(04:36:31): So as you build the initial version of what you want,

(04:36:34): if you're a developer,

(04:36:35): you want to continue in an IDE,

(04:36:37): we'll let you do that very easily,

(04:36:38): single click.

(04:36:40): We're also,

(04:36:41): and I think we'll show a little bit of this on the horizon,

(04:36:44): but also the forthcoming version of AI Studio is actually powered by the

(04:36:48): anti-gravity coding agent.

(04:36:49): So you'll be able to,

(04:36:51): from how advanced you can build some of these things,

(04:36:55): you'll be able to start going a lot farther than what we let you do in AI Studio

(04:36:58): today,

(04:36:59): which is really exciting.

(04:37:00): And I think the gap between

(04:37:03): the professional developer and what they can build and the average person.

(04:37:08): I think our goal is to help continue to compress that so that people can actually

(04:37:12): build the things that they want to build.

(04:37:15): Sweet.

(04:37:15): Well, let's get into some demos.

(04:37:18): I love it.

(04:37:19): Amazing.

(04:37:20): Amar, you want to go first?

(04:37:21): You want me to go first?

(04:37:21): Whatever you prefer.

(04:37:23): Happy to jump in.

(04:37:24): I think the first thing I'll jump into is how it's changing the way we're working.

(04:37:30): Let me share my screen.

(04:37:34): I'm going to riskily share my entire screen, so hopefully.

(04:37:38): We're doing it live, folks.

(04:37:39): We're doing it live.

(04:37:40): That's the only way.

(04:37:42): So, okay.

(04:37:43): So we're in this.

(04:37:44): Okay.

(04:37:44): So this is Figma, right?

(04:37:46): And we were exploring what a new homepage would look like for AI Studio,

(04:37:50): something a bit more utilitarian.

(04:37:52): It's got this kind of omnibar situation going on,

(04:37:55): your recent builds and stuff,

(04:37:57): and then some charts.

(04:38:00): So wanted to see what this would feel like.

(04:38:02): A Figma only really goes so far.

(04:38:05): And so to get back into it,

(04:38:07): took that over and you can see it in AI studio here where we basically have a live

(04:38:12): version of it.

(04:38:13): And what was really nice about this was you could actually feel out all the interactions.

(04:38:17): So if I click into this and I wanted to start a new chat or whatever,

(04:38:20): I can quickly pick my models.

(04:38:22): you know, type something and it just simulates what that send would be like.

(04:38:26): Or maybe I want to create a new key so I can actually just say a new key and then

(04:38:31): it goes ahead and creates a key that is a fake API key for everybody on this live

(04:38:35): stream.

(04:38:36): But what was really cool about it was like,

(04:38:38): Immediately, we were able to make it real just by giving it a screenshot and some information.

(04:38:44): And I just asked, what does the new user experience look like?

(04:38:48): Because I haven't designed that yet.

(04:38:49): And it came up with something like that.

(04:38:51): So I think this gets back to our workflow is changing because it's so much faster

(04:38:55): to just ask for that new user experience and then go and tweak it here and there.

(04:38:59): Wait, that was so fast.

(04:39:00): I want to, I want to back up and like really understand that.

(04:39:03): So, so you started in, you started in a Figma.

(04:39:06): Yeah.

(04:39:06): Right.

(04:39:07): Yep.

(04:39:07): We started basic version.

(04:39:09): Right.

(04:39:09): Basic version of Figma.

(04:39:10): And then what was the, what was the step after that?

(04:39:12): I think I, I like blinked and missed it.

(04:39:14): Um, how did you get it into AI studio?

(04:39:16): Yeah, so two things.

(04:39:17): We basically just like copy a screenshot.

(04:39:20): There'll be something like this,

(04:39:21): you know,

(04:39:21): and I just go go back to a studio and in build mode and I paste it.

(04:39:26): And then we also have this Figma plugin that we like to use a lot that just

(04:39:31): basically grabs everything you need.

(04:39:33): So for instance, if I wanted

(04:39:35): this chart.

(04:39:36): And I wanted it to look exactly like that.

(04:39:39): We have essentially a plugin which gets you all the JSON code and you basically

(04:39:44): just pass that over.

(04:39:45): And so what this ensures is that the model has effectively the exact styles,

(04:39:50): the right typography,

(04:39:51): everything.

(04:39:51): So it really looks and feels like your Figma file.

(04:39:53): Because I think that's the biggest thing designers wrestle with, right?

(04:39:57): It's like that precision.

(04:39:59): And so once you get that in, you basically just like fire away your prompt, make changes.

(04:40:03): And this was the end result of that.

(04:40:05): I see.

(04:40:06): Okay.

(04:40:07): And basically,

(04:40:08): what it looks like you have in here is there's a sidebar where you're doing your

(04:40:12): prompting,

(04:40:13): and then it has a preview panel that you can kind of look at the result and go back

(04:40:19): and forth with the model,

(04:40:20): right?

(04:40:21): Exactly.

(04:40:21): Exactly.

(04:40:22): So yeah, and that's maybe just a level set on the UI so that it's easier to see that.

(04:40:26): So this is basically the start screen of where you

(04:40:29): you go in and start prompting um and then once you're in you look at something like

(04:40:33): this where you have the left chat chat you know that we're all used to seeing now

(04:40:37): um and then and then your app preview on the right and you did that from and and is

(04:40:42): it hooked up to any sort of project where it knows anything about what you're doing

(04:40:45): or just doing it like raw from scratch raw from scratch okay yeah that's pretty

(04:40:49): cool

(04:40:51): Yeah, we see this all over Google, which is really exciting.

(04:40:54): I think it's like some double digit percentage points of all Googlers are using AI

(04:41:00): Studio to go and do this type of workflow.

(04:41:05): really exciting so uh we're seeing more more of these demos flying around and and i

(04:41:10): think it's just letting us like explore more ideas like i think the team is

(04:41:14): basically like more creative now because we can just riff on random ideas and the

(04:41:19): cost of doing that is not like a month or a week it's like you know an hour of just

(04:41:24): like coming up with crazy ideas and seeing new work and i feel like the omni barth

(04:41:28): uh proposal that amar is showing i think is a is one of these examples and also

(04:41:33): this is a great way to test it folks have feedback on this

(04:41:35): Omnibar experience.

(04:41:37): If you think this is the time before we go and build it all to get the feedback.

(04:41:43): Exactly.

(04:41:44): Very cool.

(04:41:45): Yeah.

(04:41:46): Logan, what do you got?

(04:41:48): Yeah, let me show.

(04:41:48): This is one of the experiences that I love the most,

(04:41:53): which is this was inspired by somebody on our team.

(04:41:56): I went online.

(04:41:57): I found a bunch of it was a CSV file with a bunch of like YouTube shorts analytics

(04:42:02): data from Kaggle.

(04:42:04): And I just dropped that in.

(04:42:05): And for whatever reason, it's hiding my original message.

(04:42:08): But I promise it was not like a crazy prompt.

(04:42:10): It was just like, hey, here's the CSV file.

(04:42:12): Build me an interactive visualization to look at all this data.

(04:42:16): So it built me this interactive visualization,

(04:42:18): which is just like parsing through the data and then showing me a bunch of

(04:42:21): different information on the YouTube Shorts.

(04:42:24): The cool thing,

(04:42:24): and this is what was really interesting,

(04:42:26): was I was able to add a bunch of AI features to this.

(04:42:29): So I was able to say, now I want an option to automatically use VO3.

(04:42:33): One,

(04:42:34): to create YouTube shorts that mirror the performance of those videos riffing on the

(04:42:38): ideas using AI.

(04:42:40): So it's taking in like the real actual content that was in this example,

(04:42:44): which is like food hacks and daily habits that changed my life and all these things

(04:42:49): which are like,

(04:42:49): you know,

(04:42:50): have comments and likes and other data about them.

(04:42:53): And then I can literally just one click,

(04:42:55): create a new version of that video,

(04:42:58): which is really exciting.

(04:42:59): We'll see if it works.

(04:43:01): I don't know if this is a demo of that.

(04:43:13): Hopefully it works.

(04:43:16): Okay, this part is working.

(04:43:18): Please.

(04:43:18): This is actually, I'll make a quick comment.

(04:43:22): One of my biggest gripes,

(04:43:23): and I'm curious what you two think about this,

(04:43:25): is that I end up building UIs these days where I don't actually know how the UI

(04:43:30): works and I just end up fumbling through like this.

(04:43:33): I was expecting something to happen when I click on this video and it's actually

(04:43:36): this tiny little gem right here,

(04:43:39): which does it.

(04:43:40): I don't know if that's a me problem or if that's something that other folks

(04:43:43): experience as well,

(04:43:44): but I'm always running into this.

(04:43:47): I need like a direction.

(04:43:51): Here's the thing that I just built because I'm always confused.

(04:43:53): I think that's the AI.

(04:43:54): It's like, you know, it has to do a better UX for you.

(04:43:57): You know, that's not your fault.

(04:43:58): Yeah.

(04:44:00): It is funny though,

(04:44:00): because I have seen people who have started asking for in their prompts,

(04:44:04): like create a walkthrough of the feature as well.

(04:44:07): And it actually then just goes and clicks through things and shows you stuff.

(04:44:11): And I think that's a pretty interesting way to do it too.

(04:44:13): This has been a big thing is like, okay, how do you validate something that has been vibe coded?

(04:44:18): And I do think that

(04:44:21): just actual testing, either human or AI testing is the new test suite in a lot of ways.

(04:44:28): And when someone sends a PR for one of my Vibe Coded projects,

(04:44:33): I just want to see a loom of them going through the feature a few different times

(04:44:38): rather than a code diff,

(04:44:40): honestly.

(04:44:40): I don't know if you guys have been noticing that,

(04:44:43): but I think it's a new unit of proof of quality.

(04:44:45): I walk through it a lot.

(04:44:48): Oh wow.

(04:44:54): I love it.

(04:44:55): So a fun,

(04:44:56): just like a fun way of like going from some random data source,

(04:45:00): which like I literally never even opened up the CSV file.

(04:45:03): It was able to go and take that information.

(04:45:05): It was able to process it, build me a visualization.

(04:45:07): And then actually like the next step of like taking action based on that

(04:45:12): information and like build me a bunch of like new content and things like that is a

(04:45:17): fun example.

(04:45:19): That's awesome.

(04:45:23): Should we tease some upcoming stuff?

(04:45:26): Let's do it.

(04:45:26): Let's see what we're talking.

(04:45:29): Yeah.

(04:45:29): So I think one of the other things that we're doing,

(04:45:31): uh,

(04:45:32): in AI studio is like I said,

(04:45:33): we're still making our vibe coding experience even more capable.

(04:45:36): And the thing we're doing there is it's going to support full stack apps.

(04:45:40): It's going to have amazing integrations.

(04:45:42): Um,

(04:45:43): and part of it is just seeing how good the new,

(04:45:45): uh,

(04:45:45): anti-gravity agent is as well inside of AI studio,

(04:45:48): making those apps.

(04:45:49): So, uh, the one I wanna show, and I'm gonna paste it in our chat in the studio as well.

(04:45:54): So we can see if this works live, um, and you know,

(04:45:58): We'll see.

(04:45:59): I love risking it with these demos.

(04:46:00): So what we're going to do is I'm sending you all a link to a multiplayer FigJam clone.

(04:46:09): And so that should hopefully work.

(04:46:12): And meanwhile, I'm just going to jump right in.

(04:46:14): I hope you guys show up in there.

(04:46:17): Let me see.

(04:46:18): I just want to make sure it's actually.

(04:46:21): Where did you send the link?

(04:46:23): I think I sent it in private chat.

(04:46:25): Yeah, yeah, yeah.

(04:46:26): I'm there.

(04:46:26): Hopefully that works.

(04:46:27): Did you just do that Logan?

(04:46:29): Hello?

(04:46:29): Okay.

(04:46:30): Nice.

(04:46:30): Okay.

(04:46:31): So you guys are in there, right?

(04:46:32): We are.

(04:46:33): That's awesome.

(04:46:33): Yeah.

(04:46:34): So the whole the whole the whole chat, everybody.

(04:46:37): Why not?

(04:46:37): Let's see.

(04:46:38): Get in there.

(04:46:40): Click the link.

(04:46:41): Oh, boy.

(04:46:49): uh and also you know don't don't put anything horrible yeah please yeah please

(04:46:55): don't yeah i'm gonna be nice people be nice you can put in feature requests we'll

(04:47:00): accept those this is so cool look at this there's so many people and so the crazy

(04:47:05): thing is this was one shot like this was literally a screenshot of the fig jam

(04:47:10): website and i was like make me this app

(04:47:13): Um,

(04:47:13): and then obviously like I added more and more features to it and was like fixing on

(04:47:16): the left,

(04:47:17): but the one shot experience had the,

(04:47:19): you know,

(04:47:20): had people joining in.

(04:47:21): This is wild to see, um, had people joining in and, uh, you could do a bunch of the basic stuff.

(04:47:27): So these are the kinds of apps that were really excited that people will be able to

(04:47:30): make in just one shot.

(04:47:32): Um, and, uh, yeah, more of that coming.

(04:47:34): So easy multiplayer apps, games, all sorts of things and great UI, right?

(04:47:39): Um,

(04:47:40): All in one.

(04:47:41): I love it.

(04:47:41): Yes.

(04:47:42): Make me this app.

(04:47:43): Oh, is it always a top tier prompt?

(04:47:45): Good.

(04:47:45): A good reminder.

(04:47:46): Don't don't make it too complicated.

(04:47:48): Keep it simple.

(04:47:49): I am curious as we're as we're watching,

(04:47:52): like,

(04:47:53): I don't know how many people,

(04:47:54): but at least 10,

(04:47:56): 50,

(04:47:59): 100 people like,

(04:47:57): you know,

(04:47:57): use this use this app that you one shot it.

(04:47:59): I am kind of curious.

(04:48:02): For people who are thinking about, okay, how do I, how does this fit into my workflow?

(04:48:07): When would I wanna use this versus any other tool out there?

(04:48:11): First thing is,

(04:48:12): are you thinking about this as being an alternative to like the Claude codes of the

(04:48:16): world or more of like an alternative to the maybe like bolt lovables,

(04:48:21): maybe like artifacts in Claude,

(04:48:24): or I can't remember what Chachupete calls it,

(04:48:26): Canvas and Chachupete.

(04:48:28): Like, where do you think this whole thing sits?

(04:48:31): Yeah,

(04:48:32): I definitely see it as like bold,

(04:48:34): lovable,

(04:48:35): that world where you're not jumping into an ID per se,

(04:48:38): you're not necessarily looking at CLI.

(04:48:41): But the idea here is that it's effectively a powerful way to make

(04:48:46): you know,

(04:48:46): full blown apps that you could totally deploy without much of the complexity of the

(04:48:51): setup and having to go through the pains of,

(04:48:54): you know,

(04:48:54): getting integrations and it's really easy for teams and it's easy to share these

(04:48:59): apps and remix them.

(04:49:01): So I kind of think it's all what I really like about it is like it's all in the

(04:49:05): cloud,

(04:49:06): super easy for you to share with your team as an artifact.

(04:49:08): Anybody can go and remix it.

(04:49:10): So yeah, both indie builders and teams, I think are great fits.

(04:49:13): To that point, and I know Logan, you're about to jump in.

(04:49:15): If you,

(04:49:16): the first person to do a copy of this app fully from any agent of their choice,

(04:49:22): throw it in the chat and we'll send you every hat.

(04:49:25): I want to see the first person to do it well.

(04:49:27): Logan, go for it.

(04:49:28): Yeah,

(04:49:28): I think then the beauty of this and actually like your example of just copying this

(04:49:32): app,

(04:49:32): like the real,

(04:49:35): thing that we're trying to solve is like the code quality generation across all

(04:49:39): these products is actually like somewhat similar assuming you can make the agent

(04:49:42): harness it's like if you're in IDE you then there's like four or five hops that you

(04:49:47): have to do to then like take the thing that's running locally on your computer and

(04:49:50): actually like go and make it available to more people other than running like a

(04:49:53): little

(04:49:53): local dev server and deploy it and do all the other things, spin up a production database.

(04:49:58): So we're definitely, the goal is make all of that as simple as possible.

(04:50:01): Make it so that you have the freedom to build.

(04:50:04): You don't need to worry about how do I set up this integration access to a database

(04:50:09): or auth or sign in or payments or whatever it is.

(04:50:14): And one of the cool things, and this is where I think we have

(04:50:17): our team and Google has an interesting edge and advantage is we have this whole

(04:50:21): Google ecosystem that we can tap into.

(04:50:24): With one click, you'll be able to connect to Google Workspace.

(04:50:27): With one click, you can add Google Calendar.

(04:50:29): With one click, you can add Gemini, which already works today.

(04:50:32): I think you'll see more of us going in that direction of trying to just make it

(04:50:38): really easy to connect to all the things which everyone wants to connect to across

(04:50:42): the Google ecosystem.

(04:50:43): Cool.

(04:50:44): You guys have both been involved in AI for a long time,

(04:50:48): at least since I've been paying attention to it in the GBT three days.

(04:50:52): And I'm sort of curious how things feel to you right now versus then in terms of

(04:50:59): like what your hopes back then,

(04:51:01): where we are now,

(04:51:02): where we're going,

(04:51:03): like just reflect a little bit for me.

(04:51:04): I'm really curious.

(04:51:06): yeah biggest thing is like you you are more free um i felt like back then you could

(04:51:11): like you could like taste it a little bit where it was like i i'm like kind of

(04:51:14): getting close to being able to make the thing that i want like a mars crappy

(04:51:18): children's book i was like you know it's kind of close it wasn't crappy it was so

(04:51:22): good yeah there you go i'm gonna get dan a copy immediately my copy

(04:51:30): No,

(04:51:30): but I think it's just like you're,

(04:51:33): you,

(04:51:34): there's early signs of like one day you'd be able to like do something ambitious.

(04:51:38): And I feel like the moment that we're in and especially the last six months is

(04:51:40): like,

(04:51:41): you can now do the ambitious thing,

(04:51:42): which I think is like what all of us originally wanted to do.

(04:51:45): And I feel like that's been my like personal biggest differences.

(04:51:48): I don't need to be like worried about trying to build something too complex.

(04:51:52): I can just build that thing that I wanted.

(04:51:54): do you have a sense for like what changed like to me it's it's obviously like the

(04:51:58): models just got better but is there something about the models like something that

(04:52:01): we learned over the last six months that made us be able to make the models better

(04:52:05): that would sort of cross the line anything different

(04:52:08): Tool calling is the, is the, is the fruits, the basket of fruits that bears all the gifts.

(04:52:13): Um,

(04:52:14): and so as you make tool calling better,

(04:52:16): um,

(04:52:16): and specifically over long running tasks,

(04:52:18): it's the holy grail.

(04:52:20): I agree with you,

(04:52:20): but I think it's probably unintuitive to a lot of people listening or watching why

(04:52:24): that would be,

(04:52:25): why is that?

(04:52:26): Yeah, I think, I mean,

(04:52:29): It's because, and like editing code is a good example of this.

(04:52:32): Like you're not actually,

(04:52:34): there's like a,

(04:52:35): and I think the UI products actually do a reasonable job of like abstracting this,

(04:52:39): but like there's all of these micro decisions being made.

(04:52:41): Like, should I look through these files?

(04:52:43): Should I go and edit this thing?

(04:52:44): Should I use this tool to search for documentation so that I better understand what

(04:52:48): this package or this random thing is doing?

(04:52:51): in every one of your user return requests of I want to do this thing in some

(04:52:56): software environment,

(04:52:57): there's many,

(04:52:58): many,

(04:52:58): many subsequent requests of the model basically doing on-the-fly context

(04:53:03): engineering to make sure that it has the right stuff and it's making the right

(04:53:07): edits and it's doing all that.

(04:53:09): From a model capability perspective, it's tool calling that is actually what's enabling that.

(04:53:14): yeah i think the other thing i i think is like when we were jumping into it we were

(04:53:19): all still so so early in terms of understanding what the practical applications of

(04:53:24): this will be and i think it's like finally becoming mainstream enough i think 2026

(04:53:28): we're seeing more and more teams and non-technical teams actually move into

(04:53:32): building things with these products um and i think that'll become the norm by the

(04:53:36): end of the year for sure and i also think my hot take is that now we're in the in

(04:53:40): the era of like there's

(04:53:41): going to be now you can make more and more things that are not slop now you could

(04:53:45): argue the book was slop because it was like you know there were issues with the

(04:53:49): hands and things like that but that's all easy argument but like the point is

(04:53:55): that's gone now it's actually much easier to make high quality looking things and i

(04:53:59): think um i think the slop era is uh honestly behind us it's just if you have taste

(04:54:05): you're you can go and make the thing

(04:54:07): What what is slop like?

(04:54:09): OK, here's something I've been I've been thinking about.

(04:54:11): So I don't know if you saw this this stream at the beginning and I don't blame you

(04:54:15): if you haven't been watching for the last five hours,

(04:54:18): but I've been demoing this.

(04:54:20): This Markdown editor I made is an agent native Markdown editor.

(04:54:23): It's just a very simple one where it's a text editor,

(04:54:27): all the main features you want,

(04:54:28): but you can attribute,

(04:54:29): hey,

(04:54:29): this is written by me,

(04:54:30): this is written by AI,

(04:54:32): and also it has a built-in agent,

(04:54:33): and any agent cloud code or codex or demini or whatever can go and write in it,

(04:54:37): and you can start your collaborating back and forth.

(04:54:39): So it's a really good thing to use with plan mode is you look at your plan in this editor.

(04:54:44): And it's...

(04:54:46): I wrote it in like less than two weeks in between meetings and I have no idea how

(04:54:51): it works,

(04:54:51): but it works really well.

(04:54:54): And so it's forcing me to reevaluate a little bit of like what counts a slot

(04:54:59): because I don't know how the code works,

(04:55:02): but the app itself works well.

(04:55:03): So does it matter that if I if I probably looked underneath the hood,

(04:55:09): I would probably be like, there are ways I would make this a lot better.

(04:55:12): But because software is so cheap to do, is it even worth it?

(04:55:15): Like, I feel like it's changing.

(04:55:17): The bar is changing for what craft looks like between are the actual lines of code

(04:55:23): good to like,

(04:55:23): is the output sort of what I want?

(04:55:25): I don't know.

(04:55:25): What do you guys think?

(04:55:26): Yeah, I mean, my immediate reaction to when I hear the word slop is like, it's lazy.

(04:55:30): It's like you look at it and you're like, oh, that was this first generation.

(04:55:34): You didn't really iterate on it.

(04:55:36): You basically are copying all the AI-isms of the thing and you didn't refine it and

(04:55:41): actually make it great.

(04:55:42): I mean, the app you described, you made it with intention.

(04:55:44): You basically designed around a workflow.

(04:55:46): It works really well for you, right?

(04:55:48): Now,

(04:55:48): the code may be considered by engineers at Slop,

(04:55:51): and I think they hold a different craft bar,

(04:55:53): right?

(04:55:54): The same way we would hold a different craft bar to any media we would see

(04:55:57): generated with AI if it was necessarily not great or whatnot.

(04:56:00): So I think it's a bit of that.

(04:56:02): It's like,

(04:56:02): how much effort and tension did you do to actually refine the thing and make it

(04:56:06): great?

(04:56:07): And the point being that that refining and making the thing great was actually

(04:56:10): really hard to do a while ago.

(04:56:13): You still had to put a ton of effort, and it still looked like crap.

(04:56:16): Like now I think we're at a point where no, it actually works well.

(04:56:19): You can deploy the thing.

(04:56:20): You can, you know, be confident about like using it daily.

(04:56:23): Um, those things kind of stick out to me, but Logan, do you have a take?

(04:56:27): Yeah.

(04:56:27): My quick take is like,

(04:56:28): it's,

(04:56:29): it's where your thing that you built is in distribution to like the,

(04:56:34): the.

(04:56:37): things that are like most in distribution for the model i would consider to like

(04:56:41): most b-slop right like the easiest thing to make the model do it's like the default

(04:56:45): and so the goal is like you get out of that by like pushing on the edges of the of

(04:56:49): getting the model to do things that it can like kind of barely do um and i think

(04:56:53): it's more or less the models can like barely generate high quality like really

(04:56:58): aesthetic apps with high high code quality etc but i think amar to your point i

(04:57:02): think it's like

(04:57:03): that is a transitory effect.

(04:57:05): I think the expectation you would imagine is that as we're sitting here in 12

(04:57:08): months,

(04:57:09): like that's actually no longer the case in a lot of ways,

(04:57:11): which is really interesting.

(04:57:12): So you're thinking like in 12 months from what you're seeing internally,

(04:57:17): like that's going to be the case.

(04:57:20): Yeah,

(04:57:20): I think the hard part,

(04:57:21): though,

(04:57:22): is if then you just took my definition of slop at face value,

(04:57:25): it's like the thing that looks nice and has all the right aesthetics,

(04:57:29): et cetera,

(04:57:29): et cetera,

(04:57:29): and high code quality becomes the thing that's mostly in distribution.

(04:57:33): The definition of slop will change.

(04:57:37): It's like I was,

(04:57:39): if you read,

(04:57:39): if you read like Kierkegaard,

(04:57:40): he has a lot of like,

(04:57:41): if it's not X,

(04:57:43): it's Y isms.

(04:57:44): And you're like, Kierkegaard, you're using Chachibiti.

(04:57:46): Like what the fuck?

(04:57:47): But it's actually,

(04:57:48): that was a,

(04:57:49): just a very good rhetorical move for a while,

(04:57:52): but it got overused by Chachibiti.

(04:57:53): So it sounds like, yeah, the, the, the slop target is always moving.

(04:58:00): that's why i say delve is just to sort of just see but i'm like i feel like

(04:58:04): everyone should be able to use these words just because it's in distribution for

(04:58:08): for the models for whatever random reason doesn't mean bring back delve you heard

(04:58:12): it here first um also m dashes we can't let them take that away from us no i

(04:58:17): totally everywhere yeah um guys thank you so much for joining if people want to

(04:58:23): find you or ai studio where should they find you

(04:58:27): AI.studio slash build.

(04:58:28): Go make stuff.

(04:58:29): And then don't find Amar on the internet.

(04:58:31): He doesn't.

(04:58:32): More followers.

(04:58:34): Look for the book.

(04:58:35): Great.

(04:58:35): Thanks, guys.

(04:58:36): Have a good one.

(04:58:37): Yeah.

(04:58:42): All right, everybody.

(04:58:43): So that was that was amazing.

(04:58:45): Super fun.

(04:58:46): I'm really excited for our next guest because I think he's going to have a lot of

(04:58:49): thoughts on code quality and slop and craft in this new world.

(04:58:53): We will get to that in one second.

(04:58:55): Before we do, if you are here, if you just got here, this is vibe code camp.

(04:59:01): We have been live for the last five hours showing you the best five coders in the

(04:59:06): world doing their thing,

(04:59:07): showing their workflows,

(04:59:08): what they built.

(04:59:08): We've seen some amazing stuff.

(04:59:10): If you want to get access to this whole stream turned into really consumable

(04:59:17): content,

(04:59:18): like guides that you can kind of go through and learn from all these people,

(04:59:21): subscribe to Every,

(04:59:22): every.to slash subscribe.

(04:59:24): Just put it in your email.

(04:59:25): We will be dripping those out over the next couple of weeks.

(04:59:28): As we have time to cut the stream up,

(04:59:29): we'll turn them into guides and videos that you can put into Claude or put into

(04:59:33): your LLM of choice.

(04:59:33): You can read.

(04:59:34): And you can apply all the things that you're learning.

(04:59:36): So please go to every.to slash subscribe.

(04:59:39): And you also get a bunch of other stuff as well.

(04:59:41): There's a free tier and then there's a paid tier.

(04:59:43): And you'll get a lot of stuff on the free tier.

(04:59:45): But I highly recommend, highly recommend the paid tier.

(04:59:50): With that, I am very excited to add to the stage my friend Jeffrey Lit.

(04:59:54): Jeffrey, welcome.