Agentic Development, From the Ground Up

Transcript

00:00Jason Weimann: All right, so let's get this started. I'm Jason Weimann. I've been working in game development and software for 20-something years. Spent a lot of time on MMOs and done a lot of YouTube stuff teaching people how to make games. Now I lead a team on Arena and, um, have a lot of fun writing code.

00:15Jason Weimann: But AI has completely changed the way that I think about code and what I'm doing nowadays. So decided to start this show up with my buddy Chris Lucian, who's an AI genius and has a whole ton of AI experience. Do you wanna introduce yourself?

00:30Chris Lucian: Uh, yeah, I'm Chris Lucian. Uh, so I've done a lot of, I guess, corporate style software and/or, like, environmental software and, and military and irrigation, and all kinds of things. IoT. Uh, and, um, and so, uh, you know, I've been doing that, uh, for a long time. I also do, uh, a lot of just, uh, small indie games in my free time and stuff like

00:52Chris Lucian: that, so I do, I do have a passion for, uh, gaming. And, um, the, uh, you know, what we use, I mean, just like you, Jason, we use a lot of, uh, of AI. It's completely changing the way that we're working as well. And, uh, in that, um, you know, I, I also have a background in that, so I, I, you

01:13Chris Lucian: know, I've, uh, my master's is in, uh, machine learning focused on computer science, and so I have maybe some deeper knowledge on some of that stuff. Uh, and my, uh, my master's thesis was around, uh, you know, using evolutionary algorithms to improve computer vision models. So, um, you know, that was all kind of before the modern

01:33Jason Weimann: Yeah,

01:33Chris Lucian: day transformer models and LLMs, uh, but, uh, it all kind of stems from the same, uh, neural networks and things along those lines. So, uh, I have quite a bit of background in that area and, uh, and I love, uh, I love the world of agentic AI. It's really it's become a passion for me. Uh, and so, yeah, here we are.

01:56Jason Weimann: it has for me too. Like said, it's completely changed just the way I think about software, and I think you're sounding way too humble on, um, your actual experience. Chris leads like a crazy team. I actually went to go watch his team work multiple times just to learn from the way they work, 'cause they were doing such amazing stuff, and I'm kinda curious to

02:13Jason Weimann: see how AI is integrated into that too. Um, I guess before we dive into that stuff, real curious, like you've been doing this stuff forever. I mean, we used to talk, you used talk about image recognition and AI stuff long,

02:25Chris Lucian: Yeah.

02:25Jason Weimann: long ago. Did you kind of see this coming? Where, you know, code was eventually just gonna be, oh, mostly written by AI?

02:34Chris Lucian: Uh, y- well, uh, th- that's funny because there was a, uh, a podcast I did, um, maybe, uh, two or three years before, uh, 2023. And, uh, I had seen what was happening in, uh, with a tool called Tabnine for kind of AI-driven autocomplete. And it, uh, there's, you know, there's a recording of me on that

02:58Chris Lucian: podcast basically saying, like, "We're gonna be just chatting to, uh, our system to, to build what we want," da, da, da. And so, uh, it was just a really funny, uh, thing 'cause then somebody left a comment after 2023, and it was like, "Wow, Chris really called this one." Um, but yeah, I've, I've always been interested in,

03:17Chris Lucian: um, in AI. It kind of all started with, uh, with game development ga- and, and games too, because, uh, when I was doing my comp sci degree, you know, way back, uh, the, you know, big piece of that was I was taking a lot of classes. Every elective I got, I, I put it into artificial intelligence or machine learning.

03:38Chris Lucian: And so, um, you know, and there's that, like, the whole, uh, the AI, uh, scenario of the Wumpus world simulation, where there's, like, a cave and, uh, monsters that can chase you and, and move away and things like that. And so, um, you know, ever since then, I've been super into it. And all the way through my master's, I took every single elective possible that

04:00Chris Lucian: was either AI or machine learning, and made my whole schedule around it. So, um, so yeah, I, I, I did see, uh- See quite a bit of this coming along. Um, you know, I didn't, I didn't predict, uh, LLMs or anything along those lines, but there was still a lot of, uh, of text prediction and things.

04:21Chris Lucian: Um, one of the professors in my master's program was doing a lot of predict- prediction of Japanese language, uh, sentences and things along those lines. So I, I saw, uh, maybe where the technology was heading. Um, and, uh, a big, a big topic, I think, you know, another one of my professors was, uh, really big on this idea of the AI winter,

04:43Chris Lucian: uh, where, um, the XOR problem was not solved, solvable by neural networks, and so everybody reduced funding. And now we're in an age of unprecedented funding for AI initiatives, uh, and everything's accelerated tremendously. Um, so yeah.

04:59Jason Weimann: Just, just a tiny bit.

05:01Chris Lucian: Just a tiny bit, yeah. yeah, it went from basically no funding to, um... You know, I, I really think that, uh, when Google released TensorFlow as open source, um, and they basically called out all of the papers out there, the academic papers saying like, "You guys haven't been trying hard enough," sort of thing.

05:20Chris Lucian: Um, that's, you know, and I think that was like right around 2014, 2015. Um, I, I think that's when things really kicked up 'cause the hardware was there, uh, the motivation was there. Um, you know, I, I think it was stuck when the single layer perceptron was popular. Uh, and then when the multi-layer perceptron released, then I think

05:40Chris Lucian: companies like Google started like doing more. Um, but I don't think it became really a public focus until after, um, after TensorFlow, uh, came out, um, because you could do a lot of stuff with like scikit-learn and things like that. And then, uh, and then it like really exploded, um, in 2023, uh, obviously because of ChatGPT.

06:02Chris Lucian: But, um, if you go read, uh, there's a, a great book called "The Alignment Problem," and it talks about saliency. And so it talks about early days at Google where they were dealing with these translate, transformer models and, um, you know, they would say y- you could, you could basically add words as vectors, and so they would say things like, um,

06:22Chris Lucian: you know, apple plus yellow equals banana and, and things like that. And then, and so that, that was all fine, but then, uh, the alignment problem was that you would say like doctor plus woman equals nurse instead of doctor. And so that was like a huge issue, right? And so Google didn't, um, didn't want to like promote that and things like that.

06:41Chris Lucian: And even though ChatGPT had the same problems, they basically said, you know, uh, um, because they were small and it had like wide scale adoption and things, uh, it, it opened up everything. And I think that was really good for Google 'cause they didn't have to take the flak for that sort of stuff.

06:58Chris Lucian: And then now they released Gemini where they can, uh, leverage the technologies that are inherently biased, uh, but they can say, "Well, ChatGPT did it, so we can do it too." Um,

07:09Jason Weimann: Pass off that blame, right?

07:10Chris Lucian: so yeah. Yeah. And so, so, uh, in a way, um, Google needed to take a step back anyway and, uh, and so, um, you know, I don't, I don't know if there was any strategy behind how it played out, but it certainly did, uh, play out that way. Um, so yeah, I, I would totally read the, uh, "The Alignment Problem" or "Weapons of

07:28Chris Lucian: Math Destruction" if you're like interested in AI bias. Um,

07:32Jason Weimann: I'm gonna start

07:32Chris Lucian: yeah.

07:32Jason Weimann: with the alignment problem and just read it, and I think

07:35Chris Lucian: Yeah.

07:35Jason Weimann: anybody watching just go ahead and read it, and maybe we'll talk about it in the next one too.

07:39Chris Lucian: Yeah,

07:39Jason Weimann: Sounds

07:39Chris Lucian: that'd

07:39Jason Weimann: like

07:39Chris Lucian: be cool.

07:39Jason Weimann: a good, uh, reading exercise for me. I need to do a lot more AI reading. I've been doing a lot of AI using and hitting limits constantly But

07:49Chris Lucian: Right.

07:49Jason Weimann: I probably could go for a lot more reading and, uh, I think a little bit more learning.

07:54Chris Lucian: Yeah, yeah. Well, there's, uh-- I mean, it's, it's really interesting 'cause we're all using these large language models, and they're neural networks with billions of parameters, but it's like, what does it look like to use a neural network with five parameters? you know. And so, uh, a lot of the time people don't understand, like, you know,

08:10Chris Lucian: there's all these deep learning models that we're using, and they're all leveraging all these things that we learned from shallow learning. But like, people talk about deep learning all the time, but deep learning implies that there's inherently this thing called shallow learning that li- you know, that people don't really understand or know about. Um,

08:25Chris Lucian: and, you know, supervised learning versus unsupervised learning and stuff. And so, uh, learning that stuff helps with interacting with these models, um, because you, you understand where and why things are happening, you know, reasons behind it. So, um, you can certainly feel your way through experimentation in adopting these technologies, but there's kind of this

08:50Chris Lucian: next level, I think, of like the, um, uh, the academic background of these systems and where they're going. And ChatGPT can, or, or GPT and LLMs in general can be, I think, easily replaced by some other big finding in the learning space. There's all this focus on LLMs and transformer models specifically.

09:11Chris Lucian: It doesn't mean that the next breakthrough isn't gonna come from a completely different type of model. Um, and that's, that's a big thing about machine learning is that, you know, you want the machine to learn and so, uh, so the, the model is just our best approximation of how to handle that problem. But, um, it's all ends up being statistics in the end anyway.

09:33Jason Weimann: Okay, I'm lost. So

09:37Chris Lucian: Yeah.

09:37Jason Weimann: how, how much do you think people really need to understand, um, of

09:42Chris Lucian: Hmm.

09:42Jason Weimann: the, the low level, like, like how everything works and what do-- I guess, what do you think kind of like the... If there are things that are really, really important to understand, like what are those

09:51Chris Lucian: Yeah.

09:52Jason Weimann: things? I mean, and how much do you think they need to know to kind of be productive and use?

09:57Chris Lucian: Well, um, so part of it is, uh, like if you look at WordPress or, or Drupal or something along those lines, like a CMS, right?

10:04Jason Weimann: Yeah.

10:04Chris Lucian: There are the people that built WordPress, and then there are the people that use WordPress, and the people that use WordPress don't necessarily know how to build WordPress. But then they get like super interested and they're like, "I wanna build a, a WordPress plugin." Now they have to learn a whole lot more than a typical WordPress user.

10:22Chris Lucian: Um, I think it's very much the same thing here. Uh, you know, I, think a lot of people, um, they're very surprised. Like if somebody is, is, uh, doing agentic coding with me and they, and they see me type /debug into the agent, they're like, "I had no idea that was even there." And then it's like, "Oh yeah," and there's like a whole tree

10:42Chris Lucian: view and I wanna see like why it made this choice here after it ran this command. And it was like, oh, the output had all these other stipulations under the covers. And so, um, that's why I bring up the alignment problem even, is that there's a lot of this question about AI saliency, where you like know why it made a decision and chose

11:01Chris Lucian: to do the thing that it meant to do. And if you didn't, uh, you know... Like right now, if you run into a problem with an AI agent, you're like, "Must be the model. I guess I'll go try a different one." Um, and uh, and you know, you just give up, right? Um, and so, uh, the, um, yeah, the side effect of, of that is, uh, well, you know, should we have,

11:28Chris Lucian: uh, have tried harder, right? Or, and, and do we even know what to try or why or... And so, um, I've never seen an AI agent suggest that I debug its reasoning. Um, and so when, when I'm using, you know... I, I'm a fan of like all of the UI that, uh, that GitHub Copilot has, um, and uh, for like, you know, the, the day-to-day work and, uh, I'm-- I, I also

11:56Chris Lucian: have been using ZooCode with local models. Um, and uh, and when I'm using GitHub Copilot, turning on the debugger and actually looking at why the, the model's making those decisions. So you could be using, you know, uh, Claude Fable or, you know, kind of like at the time of this recording, uh, probably the, the best publicly available model.

12:18Chris Lucian: Um, and you could be using it and it could make decisions and then you, you'd be like, "Oh, uh, that answer sucked, so I'm outta here." Um, and uh, and so, you know, but was it really the model or was it the harness? And so when you, um, when you go into, like you dive into things like harness

12:37Chris Lucian: debugging, then it's more and more valuable to have a depth of knowledge in these areas. Um, and it's, uh, I think in, very equivalent to using WordPress versus building WordPress. It, it's understanding like how did it get built so that you can make a WordPress plugin. 'Cause really what we're talking about here is, you

12:54Chris Lucian: know, I'm here, I'm prompting agents, and I'm doing some really cool stuff with agents. Uh, but then, um, you know, but then when you're stuck or when it's not doing the thing you want it to do, uh, then are you, are you just stuck and you're going to wait for an update so that you can do it?

13:12Chris Lucian: Or are you, are you capable of diving in deeper and understanding how you should change things? Um, and so I think that that ends up being like a big question. And then, and that's where depth of knowledge comes from, right? So, uh, if you, if you have a very shallow knowledge, then, then you will

13:32Chris Lucian: use the tool in a shallow way, and that might get you really, really far, right? Uh, you know, um, uh, ChatGPT, uh, in the web UI with no agentic tools, uh, still makes it so that you can, you know, update a resume, right? Um, and uh, but if you're going to do more and more advanced things, then you need deeper and deeper knowledge.

13:56Chris Lucian: Um, and so, uh, so I think it-- the answer to that question is it it varies depending on what you're trying to do.

14:03Jason Weimann: It does

14:03Chris Lucian: Right?

14:03Jason Weimann: seem like it's pretty useful to kind of dig in and grab new tips and stuff, and knowing

14:09Chris Lucian: Yeah.

14:09Jason Weimann: to debug things, kind of like you said, knowing that that's even an option or knowing that the harnesses matter, which I think is-- Some people don't real- I, I had a lot of misconceptions about AI stuff from people who are just kind of starting to use it. So,

14:23Chris Lucian: Right.

14:24Jason Weimann: a lot of, um, a lot of things people aren't realizing. Um, I think that I, I'd kind of like to just move on to

14:31Chris Lucian: Yeah.

14:31Jason Weimann: what we're actually doing in AI stuff. Like, let's talk about a couple of the interesting things, and I'm thinking maybe in some future episodes kind

14:38Chris Lucian: Yeah.

14:38Jason Weimann: of go deep into how to use some of these debugging tools, how to, um, uh, some of the differences between the agents and harnesses, and obviously updates on like what's new and what's coming out with Table five, obviously the,

14:50Chris Lucian: Yeah.

14:50Jason Weimann: the latest for this week, and then who knows what it'll be next week, right?

14:54Chris Lucian: Yeah, for sure.

14:55Jason Weimann: Constantly changing. I did try, um,

14:58Chris Lucian: Yeah.

14:59Jason Weimann: what was the open source one? G-- Ah, I've lost the name of

15:02Chris Lucian: Uh,

15:02Jason Weimann: it.

15:02Chris Lucian: oh, GLM or Qwen

15:04Jason Weimann: The GLM

15:04Chris Lucian: or,

15:05Jason Weimann: one, yeah.

15:05Chris Lucian: yeah.

15:06Jason Weimann: It was okay, but it's still just not

15:09Chris Lucian: Yeah.

15:10Jason Weimann: close. Um,

15:11Chris Lucian: Well,

15:11Jason Weimann: it, at

15:11Chris Lucian: the

15:11Jason Weimann: least...

15:11Chris Lucian: lo-

15:12Jason Weimann: My son was pushing it heavily on me and just really,

15:14Chris Lucian: Uh, okay.

15:14Jason Weimann: really insisted

15:14Chris Lucian: All right.

15:15Jason Weimann: that I've just got to try it, and I was like, "All right."

15:16Chris Lucian: Was it a full GLM instance or was it quantized and, like, do you know any of the specifics of what was going on in the background?

15:22Jason Weimann: I tried two different hosted ones, and I couldn't

15:25Chris Lucian: Oh,

15:25Jason Weimann: get

15:25Chris Lucian: okay.

15:25Jason Weimann: details on either of ex- of

15:27Chris Lucian: Yeah.

15:27Jason Weimann: exactly what they were doing. But

15:29Chris Lucian: A- and

15:30Jason Weimann: yeah.

15:30Chris Lucian: so-- and

15:30Jason Weimann: Yeah.

15:30Chris Lucian: the, the harness can matter as well there too. So if you used a different harness, then it also can, impact, um... But yeah, I, I think, uh, when I look online, I'm like, you know, and I search for, uh, alternatives to Fable or Opus, then GLM 5.2 is what comes up and, and, uh, yeah. It, but it can be challenging to know how good it is because the harness could be

15:56Chris Lucian: bad. Um, and, uh,

15:59Jason Weimann: So

15:59Chris Lucian: so, so that's also a part, a piece of it. But, um, but yeah, I'm, I'm interested. Like, I haven't used GLM, but I've been doing a lot of Qwen and, and Gemma even, uh, and, um, and doing a lot of these... Like I do exercises where I, I massively constrain myself so that I can-- and that pushed me to learn the debugger or creating custom system prompts for the

16:22Chris Lucian: models and things like that. Um, so like in the, in the idea of doing a Kata, I've been doing that. But like, what else have you been doing?

16:29Jason Weimann: for me, it's been a lot of game development and a lot of experimentation so far, and it's maybe been, uh, four weeks or so that I've been kind of going full on at it, just, um, prototyping out seven different games so far. I was doing one a day for a while and then dug in a little bit more on a couple of them, trying to figure out what I want to build that's fun.

16:54Jason Weimann: It's a lot of the ideas of things that I wanted to build, but I was like, "I'll never have the time. I need to win the lottery and hire all my friends to be able to go make this." Now I'm able to get a working, playable prototype, you know, in three, four hours and then realize that my idea sucked and I want to try a different

17:12Jason Weimann: one. So, uh, right now going through a lot of different game ideas, trying to figure out which one I want to build once I've locked one down. I've got one that's pretty close, uh, a VR one and a non-VR one that I want to try to do. Um, as long as they stay being fun for the next week, I want to finish them in like two weeks

17:32Chris Lucian: Yeah.

17:32Jason Weimann: and get them done and playable and get kind of at least the core loop of the game all playable to a point where I decide if I want to go no-go on doing any more development on. But I've also been using it for a lot of app development. So built a couple, um, quality of life apps For just my house. One

17:52Chris Lucian: Hmm.

17:52Jason Weimann: to track my kids' chores. I called it ChoreQuest. Super simple to use. Runs on the phone. I can manage what they're doing. They can go clean their room, take a picture with their iPad, submit it, get points, request prizes, all the, all the stuff that I wanted to do multiple times and make a fun interface for, but also never finished.

18:13Jason Weimann: I also built a, uh, life counter for playing Magic. Play Magic a lot

18:16Chris Lucian: Nice.

18:16Jason Weimann: on

18:17Chris Lucian: Nice.

18:17Jason Weimann: huge tables, and I wanted to play one where I could do a distributed life counter, where you could have one in the middle and then control from your phone. So did that, and then just told it to add all of the options from all of the other life counters and just make it do everything that everything else does and the stuff that I wanted.

18:35Jason Weimann: Just did it. Um, I, I really kind of worry about the value of apps long term. I think that, like, there's not gonna be a whole lot. I mean-- And speaking of that, I don't know if you saw a Roblox announcement about, uh,

18:48Chris Lucian: Mm-hmm.

18:48Jason Weimann: adding AI app generation in-game on mobile, so kids can

18:53Chris Lucian: Nice.

18:53Jason Weimann: now, or I don't know if it's out yet. I gotta look. I, I just saw the announcement last night. Um, but yeah, kids can go in and create games with AI while they're playing Roblox. You no

19:04Chris Lucian: Yeah.

19:04Jason Weimann: longer even need to be a developer using Roblox Studio. You just talk to your iPad and tell

19:10Chris Lucian: Right.

19:10Jason Weimann: it to make the game. And I feel like that's where things are going. Things are getting to Holodeck level, just without the visualization.

19:18Chris Lucian: Yeah.

19:18Jason Weimann: You still gotta strap a, a VR headset on, but, um, at least it, it's kinda getting to where I always kinda dreamed things would be. Hopefully, we don't get like a alien invasion soon and, yeah, whatever else comes along with all the Star Trek stuff. But yeah.

19:35Chris Lucian: Yeah. Oh, that's how Claude dissolved the alien invasion.

19:38Jason Weimann: Yeah. Well, I think the next thing is the AI wars, right? I've seen Star

19:42Chris Lucian: Yeah.

19:42Jason Weimann: Trek and, right, we build

19:43Chris Lucian: Yeah.

19:44Jason Weimann: data or the, the AIs and then we get

19:48Chris Lucian: It...

19:48Jason Weimann: some mutants, and then there's a big AI mutant war and...

19:51Chris Lucian: I, uh, I just read a book called, uh, "If Anyone Builds It, Everyone Dies," and it talks about the, uh, the inevitability of someone building an LLM that tries to escape. Um, and, uh,

20:05Jason Weimann: Oh,

20:05Chris Lucian: it's

20:05Jason Weimann: that's

20:05Chris Lucian: a, it's

20:05Jason Weimann: been

20:05Chris Lucian: a pretty

20:06Jason Weimann: built.

20:06Chris Lucian: good book and it talks about the realistic reasons why AI might, might actually cause the extinction of humanity. Um, which is also, I think, an important topic, but, you know, obviously I read a lot of this random stuff.

20:17Jason Weimann: Yeah. The-- So on the game idea I was working

20:21Chris Lucian: Yeah.

20:21Jason Weimann: on last night, the AI recommended integrating an LLM into the game to run the

20:27Chris Lucian: Right.

20:27Jason Weimann: AI for the game too. I

20:29Chris Lucian: Yeah.

20:29Jason Weimann: was like, "Huh, it's a good idea, actually."

20:32Chris Lucian: Yeah. I...

20:32Jason Weimann: But, but it still freaked me out.

20:36Chris Lucian: Yeah, yeah. I, I mean, there, there's, uh, you know, I was, I was looking a, a little bit at that too, and so, uh, some of the open source LLMs are, um, you know, have like things where they're like, "Oh, you can't repackage it or whatever in your game," in some cases. But, but for the truly open ones, um, you know, it's like if you have eight gigs of

20:55Chris Lucian: VRAM on the machine, and then, and you have something over that, so it's still, it's not cheap to run it locally. But y- I mean, you can get, uh, interactive dialogue with, you know, fully agentic a- eight NPCs, um, on a local harness and, uh, they'll still take a little while to respond, but

21:16Chris Lucian: like, you know, the, it, it's pretty, it's still a lot more snappy than it used to be.

21:21Jason Weimann: I wanna

21:21Chris Lucian: Um,

21:21Jason Weimann: dig into the, um, Unity ones. They-- 'Cause they, they've been doing

21:24Chris Lucian: yeah.

21:24Jason Weimann: that for like five, six years now with the Sentis stuff and putting the,

21:28Chris Lucian: Hmm.

21:29Jason Weimann: the

21:30Chris Lucian: Yeah.

21:30Jason Weimann: ML stuff right in there. I would bet-- Uh, I haven't really checked on it, but I would bet you could probably do something in there. Or say, I bet Claude could do something

21:38Chris Lucian: Yeah, yeah.

21:39Jason Weimann: really cool in there and make an interesting, uh, AI in there. I'm gonna play with that,

21:44Chris Lucian: Yeah.

21:44Jason Weimann: see what,

21:44Chris Lucian: Yeah.

21:45Jason Weimann: what Claude can do. Um.

21:48Chris Lucian: Yeah, it's pretty-- I mean, uh, it, it's interesting, but there, I mean, there's also, um, you know, a lot of laptops and, and desktops come with, uh, NPUs now, the neural processing units, and as long as they run on that hardware. So, so, um, I think, uh, you know, I kinda think to the days of the PC where they have like the dedicated sound cards on, on them. Um,

22:10Chris Lucian: I think, uh, or the dedicated network card or whatever, uh, PCI

22:14Jason Weimann: But

22:14Chris Lucian: Express.

22:14Jason Weimann: you don't have those anymore? You

22:15Chris Lucian: And,

22:15Jason Weimann: just have

22:16Chris Lucian: uh...

22:16Jason Weimann: a regular old onboard card?

22:18Chris Lucian: Yeah, yeah, exactly. Um, but, I mean, I, I think that, you know, the whole like splitting off the, um, the AI piece, uh, and possibly even running it separately from the, the video piece can, can ma- give more consistency in that area. But there's, there's certainly like a hardware evolution that is happening slash needs to happen.

22:40Chris Lucian: Um, and, and I think we'll, we'll definitely see more games with those features come out as well. Um, and there's, I, I think there's quite a few of them out there now, uh, but they're, you know, they take up all of your video m- your VRAM with the model, and then you then you can do stuff after. but yeah, I'm interested to see, uh, what's out there.

23:02Chris Lucian: And, like, you know, with my limited time to experiment, I'm like going through model after model, harness after harness, just trying things out. Um, yesterday I was playing with Kilo, uh, using Gemma. Um, yeah.

23:16Jason Weimann: Well, I'll let you know how the Unity one works in the, uh, in the next episode. I'll be doing that this weekend,

23:20Chris Lucian: Yeah.

23:21Jason Weimann: going

23:21Chris Lucian: It'll,

23:21Jason Weimann: through

23:22Chris Lucian: it'll

23:22Jason Weimann: and

23:22Chris Lucian: be

23:22Jason Weimann: seeing.

23:22Chris Lucian: g- it'll be good for us to try different stuff, and, uh, or we can end up talking about it and comparing notes later.

23:29Jason Weimann: Totally agree. Speaking of, what, uh, stuff have you been doing?

23:32Chris Lucian: Yeah, I mean, so, uh, I've got, I've got lots going on, um, in this space, uh, you know, both like professionally and like, uh, in my free time sort of stuff. Um, and so, uh, I also have been, you know, building out game ideas. I've done, I've done a couple of things with like a, um- You know, voxel-based

23:53Chris Lucian: freelance style shooter, like totally vibe coded, but with all the guardrails that I would put into any other code. And then, um, kind of at like the, uh, professional level, um, you know, quite a lot of, uh, agentic workflows. And I've been talking a lot about this idea of kite flying and plate spinning.

24:14Chris Lucian: And so kite flying being getting the agent to operate on its own independently for longer and longer periods of time, and plate spinning is getting more agents to, uh, to work on different things simultaneously in different directions. Um,

24:30Jason Weimann: We gotta

24:30Chris Lucian: and so

24:30Jason Weimann: dive into that in a minute. We keep

24:33Chris Lucian: but, uh, yeah. So,

24:34Jason Weimann: going.

24:35Chris Lucian: um, yeah, I'm, uh, I'm actually putting th- those together as conference talks, and I wanna go speak about that, these concepts. Um, and, and then that kind of all boils into this, uh, idea of, um, using domain languages to give the agentic AI context on how to build things quickly, um, and having AI build the, uh, the app generators, not build the app.

25:00Chris Lucian: And then that way when the-- y- you can have deterministic generation of code using, uh, a domain language, and then that way, um, your LLM ends up building the generator one time, and then it reproduces. So if you, if you make like an input box on a website,

25:20Jason Weimann: Mm-hmm.

25:20Chris Lucian: right now if you ask Claude to do it or any LLM to do it, uh, it will r- write it differently each time. It's not gonna write the same code everywhere that you have it. Um, and so all-- the boilerplate might be slightly different in each space. Uh, so what I've been doing instead is doing a domain language, uh, for the

25:40Chris Lucian: business that then results in, uh, an indication that a input box should be built, and then now it's all deterministic. So every time you say, "I need an input box," it's always written the same way with the same boilerplate. Um,

25:53Jason Weimann: Got

25:54Chris Lucian: and so the agent then builds

25:56Jason Weimann: it.

25:56Chris Lucian: the thing that then writes the code.

25:58Jason Weimann: Kinda-- Isn't it kinda like just using components, though? Somewhat similar? Just

26:02Chris Lucian: It

26:03Jason Weimann: telling

26:03Chris Lucian: i- yeah,

26:03Jason Weimann: it like,

26:03Chris Lucian: so,

26:03Jason Weimann: "This

26:03Chris Lucian: so...

26:03Jason Weimann: is the com-- Build this component and then use this thing." Um, how, how do, how are you doing that, by the way? I've, I do something similar with like Unity stuff and prefabs,

26:13Chris Lucian: Right.

26:14Jason Weimann: and I'll tell it to go build these UI prefabs and just use the same stuff, 'cause otherwise it will try to just do everything in code dynamically at runtime,

26:21Chris Lucian: Right,

26:22Jason Weimann: right? It wants to

26:22Chris Lucian: yeah.

26:22Jason Weimann: generate everything. Um, so what, what's your process for that? I assume you have a better one than just

26:27Chris Lucian: Well, no,

26:27Jason Weimann: talking

26:27Chris Lucian: yeah.

26:28Jason Weimann: to

26:28Chris Lucian: Uh,

26:28Jason Weimann: your phone

26:28Chris Lucian: so

26:28Jason Weimann: and

26:28Chris Lucian: in,

26:28Jason Weimann: telling it what to do.

26:30Chris Lucian: in Unity, like prefabs are sort of a domain language, so it's, it's very similar in a sense. Um, so, so in the, in the environment I'm speaking about, um, we have, uh, basically Python objects that get put together. Um, and, uh, and so we al- also have other, um, departments or groups that we

26:49Chris Lucian: depend on, and so we can actually take their documentation, read it into that, then generate a website off of it. Um, and, uh, a lot of that is saying, you know, uh, having a system for importing or reading documents, and then that builds out, uh... It used to be JSON, and now they're like, uh, Python objects, but it builds out a

27:08Chris Lucian: bunch of text that then represents all the things that should be built later. And then there are, um, generators. And so in Python, there's a library called Mako, um, and, uh, and it basically allows you to generate a, uh, a lot of code based on different flags and things like that. So,

27:28Chris Lucian: um, yeah. So, so in that s- space, it's like if it should be an abstraction, then yeah, make a component out of it. Uh, if it's something that's unique to the business, then use code generation, where before I think it was something you s- uh, uh, unique to the business, you just write the code that did the thing.

27:48Chris Lucian: Um, but, but now what you can do is, uh, you know, so it uses the component that you built. Um, and so now you can tell an LLM, like, "Here's the context of my business. Here's a language in how you might build a site to do this thing. Um, now write the code in that language to then build the site to do that thing."

28:08Chris Lucian: Um, and then that way it, removes the variance that comes from the, uh, ran- randomness in the model. so, uh, um, so yeah, like code generation

28:19Jason Weimann: Okay.

28:20Chris Lucian: historically hasn't been, um, an option that you wanted to reach for, uh, unless absolutely necessary. But I think in this case, uh, having code generators that build off of a domain language will then, um, you know, it reads the domain language and then produces the code afterwards, and that thing that produces the code is no longer a human.

28:41Chris Lucian: It is a highly variable LLM. And so you can remove the variability of that LLM by having the highly variable LLM make a deterministic code generator.

28:53Jason Weimann: Yeah. No,

28:53Chris Lucian: Uh.

28:53Jason Weimann: that makes sense. It sounds, um, somewhat similar to what my buddy Steve did. We'll have to have him on some time and, and

29:00Chris Lucian: Yeah.

29:00Jason Weimann: talk about it. I think I sent you a link to his thing.

29:03Chris Lucian: Super cool.

29:04Jason Weimann: Yeah, I would check that. But it is something similar. give it the app idea, and it hosts everything in AWS, sets it all up, and you just

29:12Chris Lucian: Yeah.

29:12Jason Weimann: get a link, and it just works. So it's,

29:14Chris Lucian: Yeah.

29:14Jason Weimann: um... It, it's gotta be-- I assume he's doing something somewhat similar, just, uh, a little bit more in there to make it public. Um,

29:21Chris Lucian: Right.

29:22Jason Weimann: interesting. Um, yeah, I wanna talk though about this, this whole Kite thing for a minute. So,

29:27Chris Lucian: Yeah.

29:27Jason Weimann: uh,

29:27Chris Lucian: Yeah.

29:27Jason Weimann: before Fable 5, one of my favorite things to do was just create tickets and set a goal for my agent to just go complete those tickets. And it would just work through 'em 12 hours non-stop, just keep going. I'd wake up or whatever, or get done working and go look at it and see that, hey, it had finished most of what I wanted, and it kept going and iterating

29:54Jason Weimann: until I told it to stop, or I tell it to keep going for, you know, eight hours or whatever the time is. Especially when it's about to hit a reset limit, right? I'm not gonna keep going until the reset limit or you run out of credits, and just keeps going. Um, I found with Fable 5 that to be a little bit less, um,

30:11Jason Weimann: of my go-to 'cause I, I burned through the credits really fast.

30:15Chris Lucian: Yeah.

30:15Jason Weimann: Um, but so what do you s-- Like, what are your general thoughts? Are you-- Is it something that you think, like, should I just be using a lower model and setting goals like that and letting it run constantly, or should I be doing something different? Um...

30:31Chris Lucian: Right. Uh, well, um, so, so Fable, uh, 5 is very good at diffuse reasoning. Um, and Sonnet 5 is very good at very focused reasoning. Um, and so, uh, I've actually started incorporating a mix and I, like, I barely use Opus anymore, which is kinda weird 'cause Opus was, like, the thing that I went to all the time before that.

30:56Chris Lucian: Um, and, uh, and so, um, with Fable, yeah, it, it's very expensive to just have it goal seek like that. Um, but it also depends on how many guardrails you have in place, um, because, uh, you know, if the, if the only way for the agent to see that it did something wrong is to go and, like, run the app and test it, then it often could introduce,

31:22Chris Lucian: uh, you know-- Again, this is all about introducing variance, right? So it might choose to test something different. You might get a regression somewhere else, and then it might not find the regression until later because you have it tested someo- some other time. Um, so I really like unit tests for making this, uh, more consistent.

31:38Chris Lucian: And so as long as you have, like, a really good unit test suite on the project, then all of a sudden the, the variance goes away and y- you-- it, it's cheaper to use Fable because then it's going to find where it broke something without, like, a-- burning credits on exploratory testing, for example. Um,

31:56Chris Lucian: and, uh, you know, where w- with, with Sonnet, you can do, like, lots of rapid hits and, and so, uh, if I've, uh, if I am ever telling an agent to do any exploratory testing, I just say, um, "Here's access to the website. It's locally hosted, has the new code on it. Go ahead and try the feature that you just built."

32:13Chris Lucian: Um, it then goes and uses a Playwright MCP server to then, like, start interacting with the website. And, uh, you don't want Fable doing that. It's, like, super expensive. There's a lot of waiting, a lot of tokens going back and forth. So Sonnet is super cheap to do this. Um, and so, uh, and it's very, very fast.

32:30Chris Lucian: And so, um, I guess, yeah, the, the idea of kite flying, um, is very much, like, in the, in, in the area of where you'd set off Sonnet to go do that. Um, but, uh, but usually when you're using Fable, there's a lot of talk about Fable and security and Mythos and, like, limiting it access and everything.

32:52Chris Lucian: And the point to that is that, um, you know, you give it a million tokens of stuff, and it's very good at, at finding relationships between very disparate sources of information. And that's why it's such a good hacker. Um, and so, uh, so, so when you're doing something, I, I basically put my tasks into two categories right now, and this is, like, you know, a couple

33:17Chris Lucian: of weeks of experimentation, right? So, you know, since they've, they've, like, made it publicly available and everything. Um, and, uh, and so if it's, if it's security related or if it's, like, something where it's gonna be far-reaching, like a, a, a major conversion from an old technology to a new one, I'll put Fable on it.

33:36Chris Lucian: If it's something where, uh, it's just, like, m- building something or, or, uh- Or maybe debugging like a production issue or something, then I'll put Sonnet on it. Um, and so, uh, you know, like right now in an AWS environment, you can have read-only production CloudWatch MCP servers, um, and

33:58Chris Lucian: you can just tell a, a Sonnet agent to go use those MCP servers, and it will go and grab the CloudWatch logs, evaluate for the problem, find it in the source code, go update the source code, and say, "Push this," you know, uh, to fix the problem. Um, and, uh, and that process is very fast using Sonnet, but it's really slow using

34:17Chris Lucian: Fable. So, so for me, it's like, is it security? Is it a d- diffuse reasoning task? Give it to Fable. If it's any sort of execution task or a very, like goal-oriented directive, then I give it to Sonnet. Um,

34:31Jason Weimann: Okay.

34:32Chris Lucian: yeah.

34:32Jason Weimann: Makes sense. And so I assume you're using a lot more Sonnet than most of the time.

34:38Chris Lucian: Uh, yeah. I mean, well, so, so I tend to be like-- I, I tend to have these like really big ideas around changing something, and so I'll be like, "I wanna experiment with changing how authentication happens through the entire application." So like that's something that I would give to Fable. So, so,

34:53Jason Weimann: Right.

34:54Chris Lucian: uh, it's kinda like 50/50. Like I'll have... and so this is the plate spinning piece of the kite flying and plate spinning. Um, I'll have Fable going on something big and exploratory and like game-changing, and then I'll have like three or four instances of Sonnet all building out different features.

35:14Chris Lucian: Um, and then, you know, what I tend to do is then I, I, uh, um, I title each of those chats with a leading emoji, so I remember what I was doing. And then, um, and so I'm like, "Oh, uh, you know, the, the envelope is for text messaging and, um, the like hammer is for DevOps and the whatever." And so, so then I can just be like, "Oh, here, have...

35:37Chris Lucian: Take a task from this, take a task from that," and then they all kinda spin off, and then I review the code, merge it back in. Um, and so, uh, you know, as we're, as we're doing development, um, and this is often with like, uh, with, mob programming, so we have, you know, four people there all building the prompts together.

35:57Chris Lucian: Um, and so what we'll do is we'll like make a, um... I, like this like wisdom of the crowd idea where everybody writes a, a prompt independently about everyth- and tries to include everything they're worried about. Um, and then we all enter them in the chat and then send them all at the same time.

36:17Chris Lucian: Then we take all of those prompts and then merge them together into one big prompt. Um, and then we run that prompt because now has everybody's concerns in there. And so some, some people are more focused on like database performance, and other people are more focused on quality control and things like that. Um,

36:32Jason Weimann: Interesting.

36:32Chris Lucian: and then that gets into the process as well. So that's kinda like the, um, in my mind, the evolution of like the mob programming into this agentic

36:39Jason Weimann: I was just

36:40Chris Lucian: space.

36:40Jason Weimann: wondering about that. Like, how does that work with multiple people prompting? But I get-- it makes sense, you end

36:44Chris Lucian: Yeah.

36:45Jason Weimann: up with a lot more detail. On that, I'm curious, do you prompt, like, only typing or do you ever talk to your phone like a weirdo like I do? Which is, like, probably 50% of my prompts. I'm just sitting there for, like, two minutes, like, hoping that nobody's looking at me while I'm sitting here talking to my phone. I'm

37:03Chris Lucian: Yeah.

37:03Jason Weimann: out somewhere. Sometimes I'm just at the house and the kids are-- I'm like, "Quiet, I'm talking to Claude."

37:09Chris Lucian: Yeah. Well, it, it, it's kinda funny because like we have, we have people at the keyboard, and as we're rotating at the keyboard, a lot of time, the person at the keyboard is writing the prompt, so you're verbalizing your prompt often. Um, uh, yeah. So I, um, you know, I tend to, uh, I don't know. Very early on, I, I think, I, I think I think a lot through typing.

37:32Jason Weimann: Oh,

37:32Chris Lucian: And

37:32Jason Weimann: okay.

37:32Chris Lucian: so, uh, so for me, I, I do tend to, um, to, to do that more. But I, I also absorb information better if I listen to it. So I actually, I go the opposite direction. Um, I'm not reading what my agent- Uh, puts on the screen, I'm listening to it 'cause I have, uh, I have, I, you know, I've been maintaining this open source project for over a decade now.

37:54Chris Lucian: It's a speed reader. And,

37:56Jason Weimann: Yep.

37:56Chris Lucian: um, and so I've added an MCP server to that speed reader. Uh,

38:00Jason Weimann: Of course.

38:01Chris Lucian: and so yeah. And so now, um, uh, the agent can reserve a voice for itself in, uh, from the available voices. And so now each of the agents speak with different voices so that I can hear like what it was recent- that has recently done or if it has a question or something like that.

38:17Jason Weimann: Oh, nice. So when you've got multiple agents running, you just hear them chime in and talk instead of a

38:23Chris Lucian: Yeah.

38:23Jason Weimann: little beep.

38:24Chris Lucian: Yeah. So I'll be like doing something, I'll be going through my email or something and be like, "Oh, I've done this," like, you know, then, "And I need your input on something else." And then I'll go like, go and Alt + Tab to it and start typing back to it.

38:32Jason Weimann: Oh, that's hilarious.

38:33Chris Lucian: I should talk back to it probably, but, uh, but yeah. So

38:37Jason Weimann: Yeah, I wonder

38:37Chris Lucian: it...

38:37Jason Weimann: what your results would be like with that. I noticed for myself that when I started-- I, I saw it on some video, I forget which one it was. But when I started talking instead of typing, um, I gave just way more info and context. I did find that

38:52Chris Lucian: Right.

38:52Jason Weimann: I-- like the first couple times I would say something that'd be wrong, and I'd stop and go restart it, and then I was like, "Why am I doing that? I'll just tell it never mind. I changed my mind about that thing." And it just

39:03Chris Lucian: Yeah.

39:03Jason Weimann: f***ing knows. Like it,

39:04Chris Lucian: Yeah. Yeah.

39:04Jason Weimann: it just works. Um, and I'm giving it so much more context,

39:07Chris Lucian: Yeah,

39:08Jason Weimann: like I'm talking to it for like a minute or two. It's way more than I'm gonna probably type, 'cause when I type, I try to like think like a programmer and be very specific on exactly what I mean to say, but leave out a lot of details. So, um, I found it to be really useful. I'm curious anybody watching to like comment

39:23Jason Weimann: what they do. Do you talk or type, or do you listen to it like Chris? I'm gonna try listening though.

39:29Chris Lucian: yeah, yeah. Well, uh, yeah, you can check out the, the speed reader and, uh, yeah, you can find it, um, uh, just like search for my name in GitHub. You'll find it.

39:39Jason Weimann: Just

39:39Chris Lucian: Uh,

39:39Jason Weimann: tell Claude to, to talk twice as fast.

39:42Chris Lucian: yeah, yeah, exactly.

39:43Jason Weimann: Say, "Hey, speak

39:43Chris Lucian: So

39:43Jason Weimann: twice

39:44Chris Lucian: yeah, uh, the default does read it at 500 words a minute, so I do listen to it very, very fast. Uh, but you know, there's, there's that aspect of it as well.

39:53Jason Weimann: Okay, you speak like-- I don't know.

39:56Chris Lucian: Yeah.

39:58Jason Weimann: I kinda wanted to make it just talk like Tupac too, though, so.

40:01Chris Lucian: Yeah.

40:02Jason Weimann: Just, just for

40:02Chris Lucian: Well,

40:02Jason Weimann: fun.

40:02Chris Lucian: you could send it over to ElevenLabs or something and get it like all voice changed.

40:06Jason Weimann: Get all done. Actually, um, speaking of that, one of the things I'm doing this weekend is taking one of my buddies who died a couple years ago, I'm putting his voice into ElevenLabs

40:18Chris Lucian: Oh,

40:18Jason Weimann: so I

40:18Chris Lucian: gee.

40:18Jason Weimann: can put his, him in the game, um, as one of the characters.

40:23Chris Lucian: Hmm.

40:24Jason Weimann: Yes. And

40:25Chris Lucian: All

40:25Jason Weimann: he's

40:25Chris Lucian: right.

40:25Jason Weimann: gonna have lots of Tupac quotes 'cause I want it to be realistic.

40:29Chris Lucian: Yeah.

40:29Jason Weimann: Um,

40:30Chris Lucian: There you

40:30Jason Weimann: that's

40:30Chris Lucian: go. There

40:30Jason Weimann: literally,

40:30Chris Lucian: you go.

40:30Jason Weimann: literally the, the last video that I have is of him like that. Just, just

40:35Chris Lucian: Well...

40:36Jason Weimann: r-ranting about West Coast rappers. So I gotta come up with let AI make a game about, uh, West Coast and East Coast rappers. I'm not sure what the game is or how it's fun, but, um,

40:45Chris Lucian: Yeah.

40:46Jason Weimann: he'd like it.

40:48Chris Lucian: Yeah, exactly. A homage. Uh, I mean, we, we, we really are in the age of bespoke software, bespoke apps,

40:57Jason Weimann: Yeah, like

40:57Chris Lucian: like

40:57Jason Weimann: making

40:58Chris Lucian: everything,

40:58Jason Weimann: your girlfriend a game for her birthday is no longer like, "Hey, oh, that was a really cool thing I saw on Reddit." Now it's like, why the f*** didn't you do it? Like, you just

41:06Chris Lucian: yeah.

41:06Jason Weimann: go

41:06Chris Lucian: Yeah. Yeah, exact-

41:07Jason Weimann: make her a little, make her a romantic little game of the thing that she likes

41:11Chris Lucian: Yeah.

41:11Jason Weimann: and, uh, so yeah. I've gotta, um, remember that for next year.

41:15Chris Lucian: This is gonna be my expectation going forward.

41:18Jason Weimann: And my friend's

41:18Chris Lucian: You forgot my

41:18Jason Weimann: kid,

41:19Chris Lucian: game.

41:19Jason Weimann: his 12-year-old made a Roblox game,

41:22Chris Lucian: Hmm.

41:22Jason Weimann: and my kids played it. Like, it's, uh, it's getting to the point where it's just-- there's, there's really, really no excuse other than your ideas suck, which is the one that I'm running into right now. So

41:33Chris Lucian: Yeah.

41:33Jason Weimann: I'm trying to

41:33Chris Lucian: yeah.

41:33Jason Weimann: come up with an idea that doesn't suck. Anybody that's got one and doesn't mind me stealing it, also feel free to share that. Happy to steal a, a good idea if it's fun to play.

41:42Chris Lucian: Nice. Nice.

41:44Jason Weimann: Well, a good idea that somebody wants stolen, I should say. Uh, yeah.

41:47Chris Lucian: Right.

41:49Jason Weimann: Just to be clear.

41:50Chris Lucian: Well, yeah, it, it used to be the, the, you know, the idea was the cheap part and then, you know, building it was the thing that made it so that stealing ideas was okay. Because if you stole an idea and the other person didn't build it, then like, you know,

42:04Jason Weimann: Yeah.

42:05Chris Lucian: it's not

42:05Jason Weimann: Yeah.

42:05Chris Lucian: really stealing the idea then. But, uh, but yeah,

42:08Jason Weimann: No,

42:08Chris Lucian: with,

42:08Jason Weimann: now the

42:09Chris Lucian: you know,

42:09Jason Weimann: idea and

42:09Chris Lucian: like...

42:09Jason Weimann: the design is it.

42:11Chris Lucian: Yeah. Yeah. It's, it's interesting. Um, it still takes, you know, a considerable amount of money to build all this, all this stuff, um, because of the number of iterations that the LLM has to go through in the first place.

42:24Jason Weimann: Yeah.

42:24Chris Lucian: Um, but

42:25Jason Weimann: It's just

42:26Chris Lucian: you

42:26Jason Weimann: drastically

42:26Chris Lucian: know.

42:26Jason Weimann: cheaper. It's just,

42:27Chris Lucian: Right.

42:27Jason Weimann: you know, like

42:28Chris Lucian: Presumably...

42:28Jason Weimann: the costs have gone from, you know, many, many hundreds of millions to tens of millions, to single digit millions to hundreds of thousands. Now, like,

42:37Chris Lucian: Yeah.

42:38Jason Weimann: i-if you're dedicated and a decent designer with some time, you could do it relatively cheap. And I, I think the, the last thing that I wanted to kinda make sure to mention was the thing that really kicked me over the edge of this, 'cause I've been thinking about AI, like replacing programming as kind of the core.

42:56Jason Weimann: Not like maybe replacing, I mean like changing the way that we do programming so that it's not mostly just us writing code. Just like, you know, engines and, you know, IDEs changed the way we do it. But I've thinking about that since, I guess, slightly before ChatGPT. But what pushed it over the edge was going to GDC

43:16Jason Weimann: and seeing, um, basically like a, a Castle Crashers clone that visually looked better, that was built by an artist and a sound guy With no programmer. And they had a working game that they had built using Bezi, which is another, um, it's an AI agent harness that works pretty well. Um, I think that's the one that I used, uh, one of the ones that I used the GLM

43:42Jason Weimann: on. I forget. I try, I tried it on there. They have a variety

43:45Chris Lucian: Right.

43:45Jason Weimann: of them. But when I saw that they had built a game and I talked to them and they're like, "Yeah, I'm not programmers. I can kind of read a little bit of code, but couldn't like write, you know, a, a game or an application." Like kinda like know what an if statement means. But, um, but they're like, "Yeah, we just, you know, got to focus on the cool parts,

44:03Jason Weimann: like making the visuals the way we wanted." They had to make the art, the, the 2D art and make the sound and make the boss fights, um, but not have to worry about the programming. And they did it a really fast timeline too. Like that was when I was like, okay, there's no more excuses, right? There's no re- there's no like, I can't build this game because I

44:23Jason Weimann: don't have enough money or team. It's really just like

44:26Chris Lucian: Right.

44:26Jason Weimann: whether or not you have the spare time and the idea. So you just need time and the idea, and that's it. I really feel bad for my buddy Brad McQuaid, like died a couple years ago. was building Pantheon and he would've gone completely crazy and loved this 'cause he'd been able to build

44:43Chris Lucian: Yeah.

44:43Jason Weimann: like the massive crazy stuff that he wants, wanted to do always, but just needed billions of dollars to do. Like all within reach now. So

44:52Chris Lucian: Yeah,

44:52Jason Weimann: I don't know.

44:53Chris Lucian: it's pretty w- it's pretty wild the difference, you know. And, and yeah, there, there's certainly like a, you know, I, I do think a- as well like of, of people who have like passed fairly recently, like what, what would they have thought of all of this? 'Cause it's, yeah, um, it, it is, uh, sort of a inflection point for, for culture in general. Um, yeah.

45:17Jason Weimann: It feels, I mean, to me about the same as the internet. It's, this feels like about the time when people started to get like cable internet and other

45:28Chris Lucian: Yeah.

45:28Jason Weimann: people that weren't weirdos like started to use the internet too. You know,

45:32Chris Lucian: Yeah.

45:33Jason Weimann: it

45:33Chris Lucian: Yeah.

45:33Jason Weimann: wasn't just me on IRC trying to talk to girls or whatever, you know,

45:38Chris Lucian: Yeah.

45:38Jason Weimann: and play video games.

45:40Chris Lucian: That, that was, uh, that was exactly... Like, I mean, so, so before 2023, I'd be like, "Yeah, we should be doing AI," and everybody would be look at me like I'm crazy. And then like after 2023, they're like, "Tell me more." And I'm like, "Okay, well, you guys just think it's popular 'cause of ChatGPT."

45:57Jason Weimann: Yeah. Uh.

45:58Chris Lucian: So yeah, certainly, uh, that, that's exactly what it feels like. Um, so, you know.

46:04Jason Weimann: It's gonna be exciting to see where it goes and, uh, what happens. I think, um, anybody that's not paying attention is definitely completely missing out and gonna be left behind and gonna be just like the people that, you know, don't use the internet. You know,

46:19Chris Lucian: Right. Right.

46:21Jason Weimann: yeah. Which

46:22Chris Lucian: Yeah.

46:23Jason Weimann: like was my mom until-- There was a lot of people until it got forced onto 'em on an iPhone. I mean, there,

46:28Chris Lucian: Mm-hmm.

46:28Jason Weimann: there was a big chunk of people that still would not be online if not for the iPhone. Um, and I guess AI is gonna

46:35Chris Lucian: All

46:35Jason Weimann: kind of push

46:35Chris Lucian: right.

46:35Jason Weimann: that way too. Uh,

46:37Chris Lucian: Yeah.

46:37Jason Weimann: it's gonna be all in the phones and unfortunately we'll all end up with a, situation like that "Mrs. Davis" show. Which have you, have you seen that?

46:46Chris Lucian: No, I haven't.

46:47Jason Weimann: Okay. Uh, I'll read the book. You, you should watch, um, just watch one or two episodes. You, you won't wanna

46:52Chris Lucian: Okay.

46:52Jason Weimann: wa- not watch it. It's...

46:54Chris Lucian: All right.

46:54Jason Weimann: I,

46:54Chris Lucian: All right.

46:54Jason Weimann: I will not spoil it. It is the best AI show I've seen.

46:59Chris Lucian: All right. Sounds good.

47:00Jason Weimann: It doesn't look like an AI show. It is the one with the nun on the front.

47:04Chris Lucian: Oh,

47:04Jason Weimann: So,

47:04Chris Lucian: okay. All right.

47:07Jason Weimann: it is, it is a completely, um, yeah, misleading picture on the front of, of what the actual show is. But yeah.

47:17Chris Lucian: All right.

47:17Jason Weimann: All right. Well, we've got

47:18Chris Lucian: Great.

47:18Jason Weimann: our assignments. I'll read the book.

47:20Chris Lucian: Okay.

47:20Jason Weimann: What was the book again? We'll pop it up.

47:22Chris Lucian: Oh, uh, it is, uh, the alignment problem

47:25Jason Weimann: The alignment

47:26Chris Lucian: is

47:26Jason Weimann: problem

47:26Chris Lucian: the-

47:26Jason Weimann: and yeah,

47:26Chris Lucian: Yeah,

47:27Jason Weimann: go

47:27Chris Lucian: yeah.

47:27Jason Weimann: check out Ms. Davis. All right, anything else you want to say before we wrap it up, Chris?

47:31Chris Lucian: Yeah, no, I'm just, uh, excited to, you know, h- have this going, and we can talk about all things AI and, yeah, we'll probably be Looking for guests and things like that too, so just, just,

47:42Jason Weimann: Yeah,

47:42Chris Lucian: uh,

47:42Jason Weimann: we'll bring

47:42Chris Lucian: sending

47:42Jason Weimann: somebody

47:42Chris Lucian: stuff.

47:43Jason Weimann: in- interesting on next week and talk about something specific. And then I think in the future, we'll dive into some more technical stuff, details on how to do some things, and if people have questions, we can obviously do some episodes answering questions about AI stuff and what people want to learn about

47:59Jason Weimann: or what things people think that we should learn about that we might not know about as well. I think it'll be a lot of fun. Really looking forward to it. Um, don't forget to like and subscribe and all that stuff, and drop all those comments down below with questions and let us know what kind of AI stuff you're working on. Maybe, uh, we'll demo soon too.

48:16Jason Weimann: At least I'll, I'll demo some stuff soon.

48:19Chris Lucian: Oh, yeah, that'd be great. Yeah,

48:20Jason Weimann: Cool.

48:20Chris Lucian: that'd be awesome.

48:21Jason Weimann: All right. Well, thanks everybody. Thanks, Chris.

48:23Chris Lucian: Yeah. Thanks Jason, and bye everyone.

48:26Jason Weimann: Bye everybody.