TanStack Start, AI, and the Future of Frontend Architecture - JSJ 701의 전체 녹취록

00:03:49.080 --> 00:03:52.159 is a lead, and that lead is not in that case, 81 00:03:52.439 --> 00:03:53.479 uh Tanner. 82 00:03:53.199 --> 00:03:55.520 Okay, Yeah. The first time I interviewed him on I 83 00:03:55.520 --> 00:03:57.639 asked him where the name Tansdak come from. I'm in, like, 84 00:03:57.719 --> 00:03:59.719 did you get a good really sun Tan one time or. 85 00:03:59.639 --> 00:04:00.280 Something like that? 86 00:04:01.159 --> 00:04:02.000 You know what my name is? 87 00:04:02.039 --> 00:04:02.240 Said? 88 00:04:02.520 --> 00:04:07.479 Oh, yeah, yeah, there you go. 89 00:04:07.719 --> 00:04:11.120 Right, Yeah, I guess it's indicative of success when you 90 00:04:11.159 --> 00:04:14.159 become the big boss and you don't need to work 91 00:04:14.199 --> 00:04:15.360 on the nitty gritty anymore. 92 00:04:15.759 --> 00:04:16.920 He's definitely in the nitty gritty. 93 00:04:18.279 --> 00:04:20.560 You can't get tenor out of the nitty gritty. No. 94 00:04:20.759 --> 00:04:24.079 In fact, actually the way that I got into tan 95 00:04:24.160 --> 00:04:27.839 Snak was I built the create tan Stack app c 96 00:04:28.079 --> 00:04:31.720 l I. So right when create React app was phased out, 97 00:04:31.839 --> 00:04:34.519 I was like, hey, Tanner, let's go and like I'll 98 00:04:34.560 --> 00:04:37.560 just go and make like a create React app variant 99 00:04:37.759 --> 00:04:39.199 where all they got to do is just create a 100 00:04:39.319 --> 00:04:41.959 change React to tan Stack and I'll give you basically 101 00:04:42.040 --> 00:04:44.600 the same app with our stuff underneath, and he's like, yeah, 102 00:04:44.600 --> 00:04:47.240 that's a great idea, and so I built that out 103 00:04:47.279 --> 00:04:49.759 and then it became this kind of like huge thing 104 00:04:49.839 --> 00:04:52.800 that you know, so we added add ons and examples 105 00:04:52.879 --> 00:04:55.759 and you know, added support for solid which we can 106 00:04:55.839 --> 00:04:58.279 now support, and we're we're going to add support review 107 00:04:58.360 --> 00:05:02.199 and all that. And actually just recently he uh he 108 00:05:02.319 --> 00:05:05.120 kind of basically took it, took it over because he 109 00:05:05.240 --> 00:05:08.000 was going and integrating it onto tansdak dot com. So 110 00:05:08.040 --> 00:05:11.040 you can now do tanstack dot com slash builder and 111 00:05:11.399 --> 00:05:14.639 basically get like a checkbox list of all the stuff 112 00:05:14.639 --> 00:05:17.399 that you want in your app and you just say, hey, cool, 113 00:05:17.519 --> 00:05:20.800 export this and they and will give you a zip 114 00:05:20.879 --> 00:05:21.560 that's got it all. 115 00:05:21.600 --> 00:05:24.639 It's pretty cool cool. So basically, instead of running it 116 00:05:24.759 --> 00:05:28.439 locally via your CLA, you can also just go online 117 00:05:28.480 --> 00:05:30.879 fit in a form and get exactly what you need. 118 00:05:31.079 --> 00:05:33.839 Exactly

and we're going to be doing a lot of 119 00:05:33.839 --> 00:05:35.000 cool stuff in that space. 120 00:05:35.240 --> 00:05:36.920 Now, what was if you support you just said you had, 121 00:05:36.959 --> 00:05:37.639 I'm a view guy. 122 00:05:37.759 --> 00:05:42.240 Just well nice. Yeah. So basically tan sak start is 123 00:05:42.399 --> 00:05:47.279 really just tansdac Router with some extra stuff for like 124 00:05:47.319 --> 00:05:51.759 service side rendering and SERO functions and API routes and 125 00:05:52.399 --> 00:05:56.879 both of those things. Router and Start are actually essentially 126 00:05:56.959 --> 00:06:00.759 an agnostic core, like Router is an agno core, and 127 00:06:00.800 --> 00:06:03.839 then there's a React variant of the router that then 128 00:06:03.959 --> 00:06:07.560 uses that agnostic core. And so we've got two different 129 00:06:07.680 --> 00:06:09.879 versions of well, we got much a bunch of different 130 00:06:09.920 --> 00:06:11.319 versions of Router, and now we've got a bunch of 131 00:06:11.319 --> 00:06:14.959 different versions of Start. So we've got a React Start, 132 00:06:15.240 --> 00:06:19.040 but we also have Solid Start, so you can actually 133 00:06:19.079 --> 00:06:23.480 build on Tan sac Start using Solid and View is 134 00:06:23.639 --> 00:06:27.120 just just now kind of getting into like the alpha phase. 135 00:06:27.519 --> 00:06:29.920 So we're going to have a view version of Start 136 00:06:30.000 --> 00:06:33.240 and probably Angular at some point. So yeah, and we're 137 00:06:33.240 --> 00:06:36.000 doing the same exact approach with AI. Basically we have 138 00:06:36.160 --> 00:06:40.360 agnostic core for an AI client and then we wrap 139 00:06:40.399 --> 00:06:42.800 that for view and. 140 00:06:42.839 --> 00:06:45.120 Spelt and so for the router, then would you use 141 00:06:45.160 --> 00:06:46.920 that instead of view router. I guess I'm trying to 142 00:06:47.000 --> 00:06:47.439 understand it. 143 00:06:47.600 --> 00:06:52.040 Yeah, yeah, you'd use that instead of view router. And yeah, 144 00:06:52.360 --> 00:06:54.480 so then you get file based routing, you get code 145 00:06:54.480 --> 00:06:57.240 based routing. We want to play it so you wouldn't 146 00:06:57.240 --> 00:06:57.399 have to. 147 00:06:57.519 --> 00:06:59.800 Use like view versus nuxts then to get the too, 148 00:06:59.839 --> 00:07:02.000 you could use either or with View. 149 00:07:03.199 --> 00:07:06.160 So I would say Tansac start for View would be 150 00:07:06.160 --> 00:07:09.160 a competitor with next. How it like in terms of 151 00:07:09.240 --> 00:07:12.360 like replacement and all that. I can't say that for sure, 152 00:07:12.360 --> 00:07:14.160 because I do know that you can actually do some 153 00:07:14.160 --> 00:07:18.319 interesting configurations, like for example, you can use Astro. You 154 00:07:18.319 --> 00:07:22.639 can actually have Tanstax start for whatever React or Solid 155 00:07:22.680 --> 00:07:25.800 or whatever you want routes within an Astro app. So 156 00:07:25.839 --> 00:07:27.800 you can actually have like an Astro app that has 157 00:07:27.879 --> 00:07:30.240 like your homepage and all that, and then when you 158 00:07:30.240 --> 00:07:32.120 want to get into the app, like you got like 159 00:07:32.199 --> 00:07:35.519 slash app, and then that's all tan sak start routes 160 00:07:35.720 --> 00:07:38.079 with all its routing. So it's actually quite nice. You 161 00:07:38.079 --> 00:07:42.839 get like combinations. What's that cloud Flair bought Astro. Yeah, 162 00:07:42.839 --> 00:07:45.480 that was a surprise to me. I mean, not really 163 00:07:46.519 --> 00:07:47.000 eat in. 164 00:07:47.000 --> 00:07:48.560 But we got to get Fred on. I haven't talked 165 00:07:48.560 --> 00:07:49.279 to him in a while. 166 00:07:49.399 --> 00:07:49.839 Oh. 167 00:07:49.879 --> 00:07:52.079 Fred's great and I would love to hear his perspective. 168 00:07:52.319 --> 00:07:57.120 Yeah, I had a version four, version five. We have 169 00:07:57.399 --> 00:07:59.519 both View and here. We've had one here. 170 00:07:59.800 --> 00:08:01.800 Yeah, we've had them on here a couple of times. 171 00:08:01.800 --> 00:08:04.199 We definitely should have him back if you know. 172 00:08:04.240 --> 00:08:08.519 Yeah, yeah, oh yeah, it's his job to do the 173 00:08:08.560 --> 00:08:15.519 cloud Feah. Well, I don't know. I'm I'm happy for him. 174 00:08:15.560 --> 00:08:16.560 That's that's clearly easy. 175 00:08:16.639 --> 00:08:21.839 Yeah, look he was he was trying. He did transform 176 00:08:22.120 --> 00:08:26.199 it into a business, so effectively selling your business or 177 00:08:26.879 --> 00:08:29.600 uh is an exit. You know, it's a success story. 178 00:08:29.720 --> 00:08:32.080 Yeah. I don't don't know what that means for the ecosystem, 179 00:08:32.159 --> 00:08:35.559 but I'm it's cool and Cladla is a great company, 180 00:08:35.600 --> 00:08:37.360 so I'm a very happy for him on that front. 181 00:08:37.759 --> 00:08:40.480 Like I know, like Sneil over cloud Flair, we have 182 00:08:40.480 --> 00:08:43.679 a great time hanging out. He's fantastic. Yeah, love love 183 00:08:43.720 --> 00:08:44.840 love the class flight folks. 184 00:08:44.919 --> 00:08:47.879 So going back to the topic at hand, Oh you 185 00:08:47.960 --> 00:08:50.000 know what before then, so one more thing. So you 186 00:08:50.000 --> 00:08:54.639 were saying that you you're introducing ten stack start as 187 00:08:54.679 --> 00:08:57.919 the basis is kind of a meta framework on top 188 00:08:57.960 --> 00:09:02.159 of a variety of front and frameworks you mentioned of 189 00:09:02.200 --> 00:09:05.120 course React where it's a kind of an alternative to 190 00:09:05.639 --> 00:09:09.759 next GS. I'm saying kind of because it has a 191 00:09:09.799 --> 00:09:13.759 bit of a different philosophy. We've actually spoken about it 192 00:09:13.799 --> 00:09:17.799 with

Tanner. Next GS, you would say, is back end 193 00:09:17.840 --> 00:09:21.600 first and can do front end whereas ten stack is 194 00:09:21.759 --> 00:09:24.480 like front end first and then you can enrich it 195 00:09:24.519 --> 00:09:26.200 with back end is the way that I was. 196 00:09:26.320 --> 00:09:30.279 Yeah, that's actually that's perfect, although the next thing certainly 197 00:09:30.279 --> 00:09:33.039 gets more complicated because you've got like literally two different 198 00:09:33.080 --> 00:09:36.120 types of routers there. You've got the pages router in 199 00:09:36.159 --> 00:09:39.720 the app router, and they work on two entirely different models. 200 00:09:39.879 --> 00:09:42.559 That being said, Tan sac Start is coming out with 201 00:09:42.720 --> 00:09:46.159 RSC support sometime soon, so excited. 202 00:09:45.840 --> 00:09:48.080 About Yeah, he was talking about that as well. 203 00:09:48.360 --> 00:09:52.039 Yeah, and it's a different model than the app rider model. 204 00:09:52.039 --> 00:09:54.240 Don't expect that you're going to get the same exact thing, 205 00:09:54.559 --> 00:09:57.320 and you're going to get a lot of different approaches 206 00:09:57.440 --> 00:09:58.799 and different benefits. 207 00:09:59.120 --> 00:09:59.279 Yeah. 208 00:09:59.320 --> 00:10:01.759 Again me who don't know what RSC support is. Can 209 00:10:01.799 --> 00:10:02.440 you elaborate on that? 210 00:10:02.679 --> 00:10:06.559 Sure? This could React Server component support. So this is 211 00:10:07.080 --> 00:10:11.600 it's a feature of React that we kind of really 212 00:10:11.639 --> 00:10:17.000 saw canonically in next JS as their app router. And 213 00:10:17.080 --> 00:10:21.120 the idea is that everything is essentially server side rendered first, 214 00:10:21.399 --> 00:10:23.960 and if you don't actually say that this one that 215 00:10:24.000 --> 00:10:25.919 you want this to run on the client, then we 216 00:10:25.919 --> 00:10:30.559 don't actually send the component code over to the client 217 00:10:30.840 --> 00:10:34.519 for rehydration on the client. So potentially, if you've got 218 00:10:34.559 --> 00:10:38.279 a very static app, you get a big performance boost 219 00:10:38.519 --> 00:10:41.960 out of rscs. But in reality, the reality on the 220 00:10:41.960 --> 00:10:45.080 ground is somewhat different because the page size, the actual 221 00:10:45.320 --> 00:10:49.720 physical bites of the page are actually like three times 222 00:10:49.759 --> 00:10:52.799 the amount of size coming out they would see coming 223 00:10:52.840 --> 00:10:54.840 out of the pages router. So there's a big difference there. 224 00:10:55.399 --> 00:10:58.679 So so interesting actual choices there. 225 00:10:59.159 --> 00:11:01.519 We kind of digre think from the original topic. But 226 00:11:01.639 --> 00:11:03.960 I want to put on that because you brought it up. 227 00:11:04.360 --> 00:11:08.320 It I have an interesting oscillation with how I

define 228 00:11:08.720 --> 00:11:14.120 or explain r SC's. I had an original explanation which 229 00:11:14.159 --> 00:11:19.159 I kind of was, which was my original explanation if 230 00:11:19.200 --> 00:11:21.799 you've ever if you've gone back in time and listened 231 00:11:21.879 --> 00:11:26.200 to my series of episodes on the Virtual Dome here 232 00:11:26.240 --> 00:11:28.519 on the podcast, I think I kind of mentioned it 233 00:11:28.559 --> 00:11:31.919 toward the end, and then after speaking having Dan Abram 234 00:11:31.960 --> 00:11:34.480 of it and friends on the podcast, I kind of 235 00:11:34.600 --> 00:11:37.559 veered away from it, and now I find myself kind 236 00:11:37.559 --> 00:11:40.919 of going back to it. And that's that my explanation 237 00:11:41.080 --> 00:11:45.919 of four reacts over components is now again v doom 238 00:11:46.039 --> 00:11:46.720 over the wire. 239 00:11:46.960 --> 00:11:51.080 Yes, I agree, Yeah, I agree completely. And the nice 240 00:11:51.120 --> 00:11:54.080 thing about the tan sac approach is we give you 241 00:11:54.120 --> 00:11:56.840 pretty much exactly that, like if you know, so it's 242 00:11:56.879 --> 00:11:59.519 up to you how you want to deploy that. So 243 00:11:59.600 --> 00:12:01.639 if you if you want the raw stuff, you can 244 00:12:01.679 --> 00:12:04.279 basically you can you can put it in reacquery if 245 00:12:04.279 --> 00:12:07.320 you want, you can have it, you know wherever we've 246 00:12:07.399 --> 00:12:09.519 you know, I've sent it over the wire on you know, 247 00:12:09.559 --> 00:12:14.240 service side event service events, which is really cool. So yeah, 248 00:12:14.240 --> 00:12:15.679 I know there's a lot of cool things that you 249 00:12:15.720 --> 00:12:17.879 can do when you think about it from that perspective. 250 00:12:18.039 --> 00:12:20.600 Yeah, I kind of veered away from it because the 251 00:12:20.639 --> 00:12:24.039 first essential implementation, and for a long time, the only implementation, 252 00:12:24.159 --> 00:12:29.240 which was next GS, really tied it to the routing behavior. Yeah, 253 00:12:29.240 --> 00:12:32.600 it wasn't exactly vedom. It was more of a way 254 00:12:32.679 --> 00:12:35.759 to get the pages in a sense, kind of a 255 00:12:35.840 --> 00:12:39.480 replacement for loaders in a sense. But now that I'm 256 00:12:39.600 --> 00:12:43.840 seeing alternative implementations like the one in ten stack, I'm 257 00:12:43.919 --> 00:12:46.559 kind of going back to that original definition. 258 00:12:46.840 --> 00:12:52.039 Yeah, I agree. Yeah, WAKU was very analogous to next GS, 259 00:12:52.320 --> 00:12:54.559 and I think that that's I think those two things 260 00:12:54.919 --> 00:12:57.919 were there because the React team was like, this is 261 00:12:57.960 --> 00:13:01.919 our vision of r SC kind of thing, and I 262 00:13:01.960 --> 00:13:05.639 remember talking with Diishi pretty early on about Waku stuff, 263 00:13:05.679 --> 00:13:08.080 and he's like, oh, don't use it like I was 264 00:13:08.159 --> 00:13:09.720 using it for microphone end to the stuff. I was 265 00:13:09.919 --> 00:13:14.159 using it so like I would change a component property 266 00:13:14.320 --> 00:13:17.279 and it would automatically just go grab another instance from 267 00:13:17.480 --> 00:13:20.000 the server and rerender it. And he's like, no, no, no, 268 00:13:20.039 --> 00:13:21.639 don't use it that way. That's not the way they 269 00:13:21.720 --> 00:13:23.720 want to want it to be used. But it works 270 00:13:23.759 --> 00:13:26.679 perfectly fine that way if you want to do that. 271 00:13:27.639 --> 00:13:30.320 And that's the great thing about the ten Sac approach 272 00:13:30.360 --> 00:13:32.919 is we give you the flexibility so you don't have 273 00:13:32.960 --> 00:13:35.000 to go and render the whole route and pay all 274 00:13:35.000 --> 00:13:38.279 that flight data cost for all of it. You can 275 00:13:38.320 --> 00:13:42.399 go and you have that Okay, I just want to 276 00:13:42.440 --> 00:13:45.679 have this component rerender off the server in place, and 277 00:13:45.720 --> 00:13:49.000 then you know, bring that data back down and it works. 278 00:13:49.159 --> 00:13:49.960 It's really cool. 279 00:13:50.159 --> 00:13:53.279 And again it has to do with flipping the direction 280 00:13:53.559 --> 00:13:56.240 because in the next years everything is server first, and 281 00:13:56.279 --> 00:14:00.240 then you explicitly need to buy into being Clyde side, 282 00:14:00.320 --> 00:14:04.159 whereas with ten stack it basically says, and again I'm 283 00:14:04.200 --> 00:14:06.840 saying this off of what Tenner said when he was 284 00:14:06.840 --> 00:14:10.799 a guest here. He realizes that the majority of React 285 00:14:10.799 --> 00:14:14.679 applications out there that exists, especially in larger organizations and 286 00:14:14.759 --> 00:14:20.639 larger deployment, are basically client side rendered applications like sixty forty. 287 00:14:20.720 --> 00:14:23.879 You can exactly and then you can enrich client side 288 00:14:23.879 --> 00:14:29.440 application rendered applications with server side rendered components. You don't 289 00:14:29.480 --> 00:14:32.799 have to flip your entire architecture in order to adopt it. 290 00:14:33.000 --> 00:14:35.519 Yep, yep, absolutely. I mean, how many times have you 291 00:14:35.759 --> 00:14:38.399 started off with a VT app right, and it's it's 292 00:14:38.440 --> 00:14:41.759 really cool, it's great, it's got all the functionality, and 293 00:14:41.799 --> 00:14:43.440 then you've got to like actually call off to an 294 00:14:43.440 --> 00:14:46.600 API to do something right. Well, either you've got to 295 00:14:46.639 --> 00:14:50.320 run a different API server and decors, or you've got 296 00:14:50.360 --> 00:14:53.039 to do some hackery inside the Vican fig to go 297 00:14:53.080 --> 00:14:55.679 and add some API rounds or

things like that. That's 298 00:14:56.080 --> 00:14:59.759 essentially what React, what start is giving you is giving 299 00:14:59.799 --> 00:15:03.240 you nice formalized way of creating those API routes and 300 00:15:03.240 --> 00:15:06.120 having them integrated in the app, creating server functions, which 301 00:15:06.120 --> 00:15:08.799 is something you don't get in the pages router, uh, 302 00:15:09.000 --> 00:15:11.200 and then making it very easy to call stuff on 303 00:15:11.240 --> 00:15:14.559 the back end. But yeah, and at the end as 304 00:15:15.159 --> 00:15:18.559 the primary vehicle, and then saying Okay, we can augment 305 00:15:18.600 --> 00:15:19.200 with stuff. 306 00:15:19.080 --> 00:15:21.759 And still encapsulated as reusable components. 307 00:15:21.879 --> 00:15:22.240 Yeah. 308 00:15:22.320 --> 00:15:27.200 Absolutely, Yeah, And before we end on this topic, it 309 00:15:27.279 --> 00:15:30.080 was really interesting for me two aspects. For the first 310 00:15:30.360 --> 00:15:34.120 is that you're also addressing view and in fact, kind of, 311 00:15:34.200 --> 00:15:37.960 like you said, in a way that's kind of an 312 00:15:38.000 --> 00:15:40.919 analogous to the way that you're kind of positioning ten 313 00:15:40.960 --> 00:15:44.600 stacks start for React versus next GS, you're kind of 314 00:15:44.600 --> 00:15:49.879 positioning ten stack start for view versus let's say next 315 00:15:50.120 --> 00:15:52.919 in a sense, so giving a bit of an alternative 316 00:15:52.960 --> 00:15:53.440 to next. 317 00:15:53.440 --> 00:15:57.240 I guess, I guess, yeah, I mean I only know 318 00:15:57.679 --> 00:16:00.559 of the view stuff, you know a little bit. I 319 00:16:00.600 --> 00:16:02.840 went in actually to try and see if I could 320 00:16:02.840 --> 00:16:05.440 get create tan sac app to go and build out 321 00:16:05.440 --> 00:16:07.039 some view apps, and I was like, oh, okay, this 322 00:16:07.120 --> 00:16:09.279 is this is still a little early days, but I 323 00:16:09.279 --> 00:16:10.279 think it's very promising. 324 00:16:10.679 --> 00:16:15.000 And finally, solid start is effectively wholly built on ten 325 00:16:15.080 --> 00:16:15.679 stack start. 326 00:16:15.759 --> 00:16:21.159 Now, well there's you said, yeah, yes, sorry, there is 327 00:16:22.080 --> 00:16:24.960 very confusing. There's the original like solid start from the 328 00:16:25.000 --> 00:16:27.759 Solid team, and then there's ten stack start for a 329 00:16:27.879 --> 00:16:31.120 solid ad different. Yeah, no, there is a difference. Yeah, 330 00:16:31.200 --> 00:16:31.679 it's two. 331 00:16:31.519 --> 00:16:33.360 Different Okay, that's interesting. 332 00:16:33.600 --> 00:16:35.759 I want to Hey, I don't know, talk to talk 333 00:16:35.799 --> 00:16:39.840 to Ryan. Yeah maybe maybe maybe two Oh, is like, hey, 334 00:16:40.039 --> 00:16:42.679 we'll take the we'll take the service side stuff for you. 335 00:16:42.759 --> 00:16:45.679 Yeah, that's the thing I wanted to say. We had 336 00:16:45.679 --> 00:16:48.759 them both on the show together, and there there's so 337 00:16:48.879 --> 00:16:51.600 much overlap between you know, the thing that they're doing, 338 00:16:51.679 --> 00:16:58.080 and and given that Ryan's primary interest is not that 339 00:16:58.240 --> 00:17:03.399 whole back yeah, front and art a part, I can 340 00:17:03.480 --> 00:17:07.920 definitely see the collaboration increasing. His is the whole thing 341 00:17:07.960 --> 00:17:13.640 about you know, reactivity and uh signals stuff and signals 342 00:17:13.720 --> 00:17:15.400 and and the sink signals. 343 00:17:15.920 --> 00:17:19.519 By the way, that's exciting. I'm excited about Solitude. 344 00:17:19.680 --> 00:17:21.960 I have to tell you. I literally told him that 345 00:17:22.160 --> 00:17:24.680 from the way that I see it, he should probably 346 00:17:24.720 --> 00:17:26.119 be in the academia. Yeah. 347 00:17:26.319 --> 00:17:28.599 Maybe, but you know, I mean having him out in 348 00:17:28.680 --> 00:17:30.839 the real world is you know nice and that like 349 00:17:30.920 --> 00:17:36.559 he's he's he really understands like ours, like the fully 350 00:17:36.720 --> 00:17:38.519 the full totality of our space. You can ask him 351 00:17:38.519 --> 00:17:41.359 pretty much about any server and any framework, and he 352 00:17:41.440 --> 00:17:43.200 just keeps up with all of it. He knows all 353 00:17:43.200 --> 00:17:47.200 the different approaches. It's just it's really he's a fantastic guy. 354 00:17:47.279 --> 00:17:49.759 He actually lives pretty close to me now, so hopefully 355 00:17:49.799 --> 00:17:50.960 I can't get him out for lunch again. 356 00:17:51.000 --> 00:17:52.559 But yeah, I know that they've moved. 357 00:17:52.680 --> 00:17:53.960 So yeah, they moved up to Oregon. 358 00:17:54.680 --> 00:17:56.119 Oh cool, where are you at. 359 00:17:56.000 --> 00:18:00.400 Jack, I'm in Portland, Oregon. So there you go. We 360 00:18:00.440 --> 00:18:01.319 need to meet up. 361 00:18:01.599 --> 00:18:03.160 Yes we do. I'm out east. 362 00:18:03.519 --> 00:18:06.799 I mean, I mean I'm out east, I mean inn 363 00:18:06.839 --> 00:18:07.400 Heavy Valley. 364 00:18:07.480 --> 00:18:09.759 But dude, you're like five miles from me. 365 00:18:10.119 --> 00:18:11.480 Dude, we need to meet up. 366 00:18:12.440 --> 00:18:14.079 I didn't he like right around the corner. 367 00:18:16.000 --> 00:18:17.000 It's a small world. 368 00:18:17.559 --> 00:18:20.559 It is a small world. I would say. Portland is 369 00:18:20.640 --> 00:18:22.319 kind of up and coming, like if you're on the 370 00:18:22.400 --> 00:18:26.680 West coast. Portland is a great, great city. It really is, 371 00:18:26.759 --> 00:18:29.279 Like I but don't come here because we don't, you know, 372 00:18:29.279 --> 00:18:30.359 we don't want more more people. 373 00:18:30.440 --> 00:18:32.319 All the construction going around right around my

house. 374 00:18:34.319 --> 00:18:37.519 Oh my god, dude getting around though, Like there's always 375 00:18:37.519 --> 00:18:40.359 a flag guy I've got, there's always Like. 376 00:18:40.359 --> 00:18:44.599 It might surprise you, but Tel Aviv is undergoing, like 377 00:18:44.960 --> 00:18:47.079 the whole city is one. We basically say that the 378 00:18:47.079 --> 00:18:49.319 whole city is on wheels, Like all the streets are 379 00:18:49.359 --> 00:18:53.880 dug up, it's all filled with cranes. Uh, they're we're 380 00:18:53.920 --> 00:18:57.480 building a kind of like a metro system, so everything 381 00:18:57.559 --> 00:19:00.000 is like dug up high rises going up a little 382 00:19:00.039 --> 00:19:02.920 of the place. So it's yeah, but the problem is 383 00:19:02.960 --> 00:19:05.000 is trying to get from point A to point B 384 00:19:05.119 --> 00:19:07.480 in the side the city is becoming literally impossible. 385 00:19:07.559 --> 00:19:09.440 Yeah. If you want to find history lesson, go check 386 00:19:09.440 --> 00:19:12.799 out the history of Chicago. They actually raised up a 387 00:19:12.839 --> 00:19:15.160 lot of buildings up on these jacks when they're adding 388 00:19:15.200 --> 00:19:17.559 in like sewage and plumbing lines and all that kind 389 00:19:17.599 --> 00:19:19.880 of stuff. And there were like there were stories of 390 00:19:20.119 --> 00:19:22.839 folks waiting in an intersection as a house would come by. 391 00:19:23.000 --> 00:19:26.160 Well, from my understanding, it's kind of built like on 392 00:19:26.160 --> 00:19:28.079 on the basis of a swamp or something. 393 00:19:28.880 --> 00:19:33.200 Well, I think that's DC Actually, yeah, that's maybe Also yeah, 394 00:19:33.359 --> 00:19:35.319 we're like Mexico City, which is built on a lake. 395 00:19:35.920 --> 00:19:38.319 Anyway, I was just going, but I think we're kind 396 00:19:38.319 --> 00:19:44.400 of digressing. That's okay, let's that's a podcast for exactly. 397 00:19:44.799 --> 00:19:47.759 So let's try to get back to the original topic 398 00:19:48.000 --> 00:19:50.160 that I thought would be great to discuss with you, 399 00:19:50.400 --> 00:19:54.319 And that's a new ten stack component that you've mentioned, 400 00:19:54.680 --> 00:19:57.960 which is ten stack AI. Yeah, so you said that 401 00:19:58.039 --> 00:20:01.519 it's kind of like the verse V zero whatever they 402 00:20:01.559 --> 00:20:01.960 call it a. 403 00:20:02.240 --> 00:20:04.880 No, it's not V zero. V zero is their application builder. 404 00:20:05.119 --> 00:20:07.680 I'm talking about the AI s d K, which is 405 00:20:07.839 --> 00:20:13.799 a library that gives you some an agnostic wrapper around 406 00:20:14.039 --> 00:20:19.839 the AI, the l m uh chat you know, APIs 407 00:20:19.839 --> 00:20:22.400 for all these back end primarily although you know they 408 00:20:22.400 --> 00:20:24.720 can now support like you know we and we can 409 00:20:24.759 --> 00:20:28.119 do this too, image generation, video generation content. 410 00:20:28.240 --> 00:20:30.559 So what would so what would I use it for? 411 00:20:30.640 --> 00:20:33.720 What is it exactly? Like, let's talk about your stuff, 412 00:20:33.759 --> 00:20:37.759 So what's exactly ten stack AI and what would I 413 00:20:38.119 --> 00:20:38.640 use it for? 414 00:20:38.880 --> 00:20:40.279 Right, So let's say that you want to go and 415 00:20:40.319 --> 00:20:44.279 add a chat functionality to your app. Right, you want 416 00:20:44.319 --> 00:20:45.920 to have the user be able to you know, have 417 00:20:46.000 --> 00:20:48.160 a bot and talk to the bot and get like 418 00:20:48.240 --> 00:20:52.319 product information that kind of stuff. So tansac AI that 419 00:20:52.400 --> 00:20:55.640 library is going to be the server side of that equation. 420 00:20:56.039 --> 00:21:01.880 It gives you a vendor agnostic way of connecting to lms, 421 00:21:01.920 --> 00:21:07.200 so anthropic open AI, you know, O LAMA, they all 422 00:21:07.240 --> 00:21:13.079 have the same basic API, and then you can connect 423 00:21:13.119 --> 00:21:17.480 it to our chat function which is an agentic chat, 424 00:21:17.480 --> 00:21:20.839 which means that it supports service side tools with approvals 425 00:21:20.960 --> 00:21:23.559 and all that kind of fun stuff, so you can 426 00:21:23.559 --> 00:21:27.039 get an agentic workflow out of the app, and then 427 00:21:27.079 --> 00:21:29.079 you can connect to that that actually gives you a 428 00:21:29.079 --> 00:21:32.240 bunch of tokens that you can then send out via 429 00:21:32.359 --> 00:21:34.359 any way you want. We give you out of the 430 00:21:34.359 --> 00:21:37.400 box HTP streams and service and events, and then the 431 00:21:37.519 --> 00:21:41.319 AI client stuff, which we also wrap as AI react 432 00:21:42.519 --> 00:21:45.839 is a generic way to connect to that endpoint and 433 00:21:45.880 --> 00:21:50.279 then get the message stream and put in the chunk 434 00:21:50.319 --> 00:21:52.880 stream and then convert that into messages so you get 435 00:21:52.880 --> 00:21:55.759 the like the full on chat experience. And it supports 436 00:21:55.759 --> 00:21:58.319 also client side tools, so you can have things like 437 00:21:58.799 --> 00:22:01.279 a good example would be I'm on a e commerce 438 00:22:01.319 --> 00:22:04.240 site and I want to, you know, talk with the 439 00:22:04.319 --> 00:22:06.519 chatbot about what a motorcycle I want to buy or 440 00:22:06.599 --> 00:22:08.440 something like that, and it says, oh, you know, here's 441 00:22:08.440 --> 00:22:10.240 the motorcycle for you, blah blah blah blah blah, and 442 00:22:10.279 --> 00:22:11.799 you can say, cool, add to my cart, you know, 443 00:22:12.119 --> 00:22:13.920 not I don't know if you do that on the motorcycle, 444 00:22:13.960 --> 00:22:16.279 but you know, and then it would w use a 445 00:22:16.279 --> 00:22:19.599 client some people are a lot more

wealthier than me, 446 00:22:20.039 --> 00:22:23.759 would then add it to your cart on client side 447 00:22:23.880 --> 00:22:25.079 or on the service side. How are you going to 448 00:22:25.119 --> 00:22:27.160 play it. So that's where that that kind of tools 449 00:22:27.200 --> 00:22:29.720 integration and agentic workflow come in here. 450 00:22:29.839 --> 00:22:34.960 So your primary audience would be a website looking to 451 00:22:35.079 --> 00:22:39.039 integrate AI functionality or literally building something like a cursor. 452 00:22:39.160 --> 00:22:42.319 Oh, that's an interesting you could build cursor. And in fact, 453 00:22:42.319 --> 00:22:45.680 actually I believe open code uses the versaal Ai SDK, 454 00:22:45.839 --> 00:22:48.960 but doesn't use it all that much. They just use 455 00:22:49.000 --> 00:22:51.039 it as like a neutral layer so they can go 456 00:22:51.079 --> 00:22:53.319 and have like a single code base that connects to 457 00:22:53.359 --> 00:22:58.440 open ai models and anthropic models and all that and 458 00:22:59.240 --> 00:23:01.960 more than welcome to talk to them about about moving 459 00:23:02.000 --> 00:23:05.240 to tanzac ai. So yeah, you could use it in 460 00:23:05.240 --> 00:23:07.480 a CLI, you can use it in web, you can 461 00:23:07.599 --> 00:23:11.240 use it in Native, you can use it on you 462 00:23:11.240 --> 00:23:13.400 can use it in electron if you want. You can 463 00:23:13.440 --> 00:23:16.400 actually run the tokens across IBC. Right, it's really just 464 00:23:16.440 --> 00:23:18.640 a question of you know, you've got one thing that's 465 00:23:18.680 --> 00:23:20.720 going to go and give you an agnostic way to 466 00:23:21.119 --> 00:23:26.400 hit your lms for various services like JAT or image 467 00:23:26.440 --> 00:23:29.880 generation or whatever whatever they support. And then the client side, 468 00:23:29.880 --> 00:23:32.559 which is about like getting the tokens back and formatting 469 00:23:32.599 --> 00:23:35.119 them into messages. So that they're easy to display. 470 00:23:35.279 --> 00:23:40.480 So just so if on the higher level, it seems 471 00:23:40.480 --> 00:23:42.599 to me, let's see if find the center correctly, that 472 00:23:42.680 --> 00:23:50.400 you're offering two main services. One is an LLM agnostic 473 00:23:50.799 --> 00:23:55.240 way to connect to the llms, and you're saying that 474 00:23:55.319 --> 00:24:00.000 they have a similar but not identical API, so you 475 00:24:00.920 --> 00:24:04.640 overlay like a facade on top of that that basically 476 00:24:05.240 --> 00:24:08.119 irons out the wrinkles that make them makes them actually 477 00:24:08.160 --> 00:24:09.279 identical in a sense. 478 00:24:09.480 --> 00:24:12.279 Yeah, basically, yeah. 479 00:24:11.319 --> 00:24:14.720 And in addition, you also get a chat client that 480 00:24:14.960 --> 00:24:17.799 works with this infrastructure. 481 00:24:18.720 --> 00:24:23.319 Yeah, and we're just in the process of finalizing that 482 00:24:23.400 --> 00:24:27.400 we'll be moving. We have our own internal standard for chunks, right, 483 00:24:27.440 --> 00:24:29.799 that'd be, oh, this is the LM and it just 484 00:24:29.799 --> 00:24:32.400 put out a chunk with you know, this particular token 485 00:24:32.440 --> 00:24:35.319 in it or whatever. And we're going to be moving 486 00:24:35.359 --> 00:24:40.279 to what's called AGUI, which is a standard for chunked tokens. 487 00:24:41.000 --> 00:24:43.799 And the cool thing there is that you'll be able 488 00:24:43.839 --> 00:24:49.720 to use our server AGUI exported chunks with anything on 489 00:24:49.799 --> 00:24:51.839 the front end that supports that standard, So you don't 490 00:24:51.880 --> 00:24:53.799 have to use our hooks and our client if you 491 00:24:53.839 --> 00:24:57.119 don't want to, and vice versa. If you have if 492 00:24:57.119 --> 00:24:59.319 you use our client and our hooks, you could talk 493 00:24:59.359 --> 00:25:01.599 to any server doing a g u I on the 494 00:25:01.640 --> 00:25:07.720 back end. And we actually also have currently Python and 495 00:25:07.839 --> 00:25:12.000 PHP libraries for the server, which is something that is 496 00:25:12.200 --> 00:25:15.640 something that the versaal AI s DK isn't currently supporting. 497 00:25:15.839 --> 00:25:20.599 It's a typescript only ecosystem where we are pretty interested 498 00:25:20.640 --> 00:25:24.400 in making sure that like Python fast API as an example, 499 00:25:24.519 --> 00:25:27.440 as well positioned is able to use it as well 500 00:25:27.480 --> 00:25:30.440 as PHP. You know, there are a few sites out 501 00:25:30.440 --> 00:25:31.640 there that we're on PHP. 502 00:25:31.880 --> 00:25:37.680 Interesting. I'll actually speak to the AI engineers at size Sense, 503 00:25:37.720 --> 00:25:40.000 the place where I work, and you know, to let 504 00:25:40.079 --> 00:25:44.359 them know about it, because we have AI chats embedded 505 00:25:44.400 --> 00:25:51.799 within our within our software. We also do AI based services, 506 00:25:51.880 --> 00:25:56.240 so we do data analytics. You can you know, ask 507 00:25:56.640 --> 00:25:59.359 you can actually we call it semantic enrichment. You can 508 00:25:59.400 --> 00:26:05.880 actually tell the system, hey, I want AI to provide 509 00:26:05.920 --> 00:26:09.519 additional meta data on top of the meta data that 510 00:26:09.599 --> 00:26:12.720 I got out of the databases that I'm working with. 511 00:26:13.039 --> 00:26:18.759 Sure then and you know it's it's really it's really 512 00:26:18.920 --> 00:26:23.000 a useful functionality. It makes reasoning with your data much 513 00:26:23.039 --> 00:26:27.160 easier and and definitely I don't know which models we're 514 00:26:27.240 --> 00:26:30.599 currently limited to, but if we're able to easily open 515 00:26:30.640 --> 00:26:33.759 it up to additional models, that that can be really 516 00:26:33.839 -->

00:26:37.400 really interesting because very often our customers say, hey, we 517 00:26:37.440 --> 00:26:40.319 want we prefer working with this model or that model 518 00:26:40.480 --> 00:26:45.000 because that's the one that standardized within our organization, that's 519 00:26:45.039 --> 00:26:49.480 the one that we have you know, coupons for and 520 00:26:49.519 --> 00:26:53.279 stuff like that. Yeah, we we go we use Ralph, 521 00:26:53.359 --> 00:26:59.680 so now we run through coupons like like, uh exactly. 522 00:27:00.000 --> 00:27:02.599 I'll probably talk about that maybe later on. 523 00:27:02.640 --> 00:27:02.880 But. 524 00:27:05.359 --> 00:27:08.720 Yeah, So in terms of compatibility, just so folks know, 525 00:27:09.720 --> 00:27:15.079 we originally came out with open Ei, Anthropic, Gemini, and Olama, 526 00:27:15.279 --> 00:27:17.960 and then we added GROC and now we've just recently 527 00:27:17.960 --> 00:27:22.319 added open Router, so quite a bit of support there, 528 00:27:22.359 --> 00:27:25.640 and then we're also obviously looking at Azure and Bedrock 529 00:27:26.119 --> 00:27:30.000 is next big primary targets in terms of the back 530 00:27:30.119 --> 00:27:32.119 end support. And the reason it takes a little bit 531 00:27:32.160 --> 00:27:33.960 of time is we spend a lot of time actually 532 00:27:34.039 --> 00:27:36.160 going through and taking a look at all the different models, 533 00:27:36.480 --> 00:27:39.359 cataloging them, getting all the typescript support in there. Because 534 00:27:39.640 --> 00:27:44.519 when you select a model from a vendor, then you 535 00:27:44.559 --> 00:27:47.759 get all the usual properties, but you also like a 536 00:27:47.880 --> 00:27:51.039 temperature and this and that. But you also get proprietary 537 00:27:51.079 --> 00:27:54.079 properties for each model, and they're all strongly typed. So 538 00:27:54.119 --> 00:27:56.400 if you've got things like you know, some sort of 539 00:27:56.640 --> 00:27:59.039 I don't know, conversation ID or something like that that's 540 00:27:59.079 --> 00:28:03.200 going to go to Anthropic for storage, you know, then 541 00:28:03.599 --> 00:28:06.519 you can add that in and it's all type safe. 542 00:28:06.559 --> 00:28:10.000 And how how do you align through the differences between 543 00:28:10.039 --> 00:28:12.759 the various models, because you said that they're very similar 544 00:28:12.799 --> 00:28:15.480 but not identical, so I assume that there might be 545 00:28:15.559 --> 00:28:18.279 some interesting differences in kinks to work out. 546 00:28:18.640 --> 00:28:21.720 Yeah, I mean yeah, not so much, to be honest. 547 00:28:21.720 --> 00:28:24.160 I mean they're like, yeah, there's there's like little you know, 548 00:28:24.240 --> 00:28:28.359 there's proprietary flags for sure, but in most of most 549 00:28:28.359 --> 00:28:30.200 of the l l ms, you know, they're pretty much 550 00:28:30.240 --> 00:28:34.839 standardized on on at least in terms on the options. 551 00:28:35.240 --> 00:28:37.920 It's the chunk format really that comes back like from 552 00:28:38.039 --> 00:28:41.119 the actual vendor that can get a little That's actually 553 00:28:41.160 --> 00:28:43.680 where we end up spending a bunch of time is like, yeah, 554 00:28:44.359 --> 00:28:48.400 massaging the chunks coming back from the l M into 555 00:28:48.440 --> 00:28:51.720 our standard and now obviously a a g u I standard, 556 00:28:51.799 --> 00:28:55.200 and then from there we senator for the client. 557 00:28:55.400 --> 00:28:59.720 So given the AGUI standard. Theoretically, in an ideal world 558 00:29:00.200 --> 00:29:02.720 two years from now, you won't actually be needed. 559 00:29:05.720 --> 00:29:10.000 I yeah, well, okay, I think that. I guess there's 560 00:29:10.039 --> 00:29:13.079 some really cool stuff that we're working on that's way, way, 561 00:29:13.119 --> 00:29:16.000 way beyond what people are used to seeing out there. 562 00:29:16.200 --> 00:29:18.759 So I kind of consider where we are currently in 563 00:29:18.759 --> 00:29:22.799 that published state, sort of like table stakes, and then 564 00:29:23.200 --> 00:29:25.079 you know, the kind of stuff that we'll be putting 565 00:29:25.119 --> 00:29:28.480 on top of that is going to be groundbreaking. So 566 00:29:28.480 --> 00:29:29.799 I'm really really excited. 567 00:29:30.039 --> 00:29:32.640 That'll be interesting. Also, you know, again, when you're talking 568 00:29:32.680 --> 00:29:39.119 about extended media types like audio, video images, you're adding 569 00:29:39.160 --> 00:29:42.240 a whole new additional layer of complexity, which I'm sure 570 00:29:42.680 --> 00:29:47.359 needs better alignment between the different models, the type of 571 00:29:47.400 --> 00:29:50.319 stuff that they're bringing, you know, handing out back and 572 00:29:50.359 --> 00:29:52.440 how yeah you know it to the user. 573 00:29:52.599 --> 00:29:56.000 Yeah, like video was a kind of a an entirely 574 00:29:56.039 --> 00:30:00.240 different interaction models. It's like a job based model, fire 575 00:30:00.240 --> 00:30:01.559 off a job to get back a job, I d 576 00:30:01.839 --> 00:30:03.519 because it just takes so long. It takes so long 577 00:30:03.519 --> 00:30:06.599 to generate video, so it's like an entirely different API surface. 578 00:30:06.680 --> 00:30:10.160 In fact, actually that's kind of an interesting digression, is 579 00:30:10.920 --> 00:30:15.279 around we actually spend a bunch of time going from 580 00:30:17.359 --> 00:30:20.039 kind of like an initial release to the second release 581 00:30:20.079 --> 00:30:23.359 that had like all the support for image and video 582 00:30:23.640 --> 00:30:26.519 and structured output and that sort of stuff. And the 583 00:30:26.559 --> 00:30:29.799 reason was because we kind of did an AI an 584 00:30:29.839 --> 00:30:32.319 API review and we saw that we were just going 585 00:30:32.400

--> 00:30:35.400 to keep on adding new functions to the export of 586 00:30:35.440 --> 00:30:39.200 AI and you know, like generate image and generate video, 587 00:30:39.240 --> 00:30:40.839 and we kind of ask ourselves, you know, is that 588 00:30:40.839 --> 00:30:42.680 what we want to do, because you know, as the 589 00:30:42.720 --> 00:30:44.839 field expands, we just kind of want to keep adding 590 00:30:44.880 --> 00:30:48.839 functions to it. And so we initially decided, well, we'll 591 00:30:48.920 --> 00:30:50.599 just kind of try to see if we can do 592 00:30:50.720 --> 00:30:53.319 a polymorphage design where you have it just like basically 593 00:30:53.359 --> 00:30:57.319 one export coming out of Tansac AI, one real export, 594 00:30:57.319 --> 00:30:59.799 which would be the AI function, and then based on 595 00:31:00.039 --> 00:31:02.920 the model that we send in, we would then do 596 00:31:02.960 --> 00:31:04.960 all the typing and all that kind of stuff to 597 00:31:05.039 --> 00:31:08.200 like say, oh, okay, cool, this is a you know, 598 00:31:08.319 --> 00:31:10.839 a video model. Therefore, you know, get to get back 599 00:31:10.839 --> 00:31:13.400 a job i'd or something like that. And what we 600 00:31:13.640 --> 00:31:17.200 found after six days was that it just wasn't a 601 00:31:17.240 --> 00:31:20.960 great model, and it was actually working with the AIS 602 00:31:20.960 --> 00:31:23.680 themselves that allowed me to kind of go in and 603 00:31:23.880 --> 00:31:27.319 and and say, hey, between these two different APIs this 604 00:31:27.480 --> 00:31:31.000 kind of generic AI function versus generate image. You know, 605 00:31:31.119 --> 00:31:34.680 with that, which what what would you prefer us to 606 00:31:34.759 --> 00:31:37.200 do and write and what what what do you think 607 00:31:37.240 --> 00:31:40.839 would be it would increase accuracy of an l M 608 00:31:41.039 --> 00:31:45.440 and one across the board. All the frontier models are like, ah, dude, 609 00:31:45.759 --> 00:31:46.519 generate image. 610 00:31:46.680 --> 00:31:50.799 Duh. Oh yeah, well that's that's not actually really surprising. 611 00:31:51.039 --> 00:31:52.720 No, no, it's not surprising at all. But it's one 612 00:31:52.720 --> 00:31:54.200 of those things where you're like, oh, you know, you 613 00:31:54.240 --> 00:31:57.039 should have done this earlier, so you know it. 614 00:31:57.160 --> 00:31:59.880 You know, that's it's it's kind of like actually learning. 615 00:32:00.079 --> 00:32:02.519 So it's kind of like ECHA mean, having the different 616 00:32:02.559 --> 00:32:06.720 tags for the very different media types rather than having 617 00:32:06.839 --> 00:32:11.440 like a generic media type element that does everything right exactly. 618 00:32:11.519 --> 00:32:14.519 You know. It's it's so much more verbos And if 619 00:32:14.559 --> 00:32:17.799 you have just a single function I right, you're getting 620 00:32:17.799 --> 00:32:23.400 away like the primary communications vehicle for what the thing does, 621 00:32:23.519 --> 00:32:27.200 which is the function name. It's like pretty obvious once 622 00:32:27.200 --> 00:32:29.119 you think about it. The other thing that I thought 623 00:32:29.160 --> 00:32:32.519 was interesting about that experience was that all of the 624 00:32:32.559 --> 00:32:35.400 agents like Claw and all that were really having a 625 00:32:35.440 --> 00:32:38.640 hard time actually trying to get the polymorphic stuff to work. 626 00:32:38.839 --> 00:32:41.440 You know, one you get it right, it would work 627 00:32:41.480 --> 00:32:44.440 for one thing, and then it would break something else. 628 00:32:44.759 --> 00:32:46.240 And we had all these type tests and all that 629 00:32:46.319 --> 00:32:49.200 kind of stuff, and what I really learned was that 630 00:32:49.559 --> 00:32:52.039 AI is not going to complain. Man, They're going it's 631 00:32:52.039 --> 00:32:54.640 gonna be like, okay, yeah, no problem, Man, I can 632 00:32:54.680 --> 00:32:57.960 do whatever you want. And that's a problem because if it, 633 00:32:58.079 --> 00:33:01.920 if it had beened me trying to manually code this 634 00:33:02.000 --> 00:33:04.200 new interface, I would have been like an hour into 635 00:33:04.279 --> 00:33:06.720 and been like, man, this doesn't work, Like this is 636 00:33:06.759 --> 00:33:10.400 really hard, Like it shouldn't be this hard. And that's 637 00:33:10.440 --> 00:33:13.480 always been like those pain points have always been like 638 00:33:13.519 --> 00:33:16.119 a like an engineering smell to like see in the 639 00:33:16.440 --> 00:33:19.599 junior dev senior devs like you run into a wall 640 00:33:19.799 --> 00:33:22.119 and like C plus plus is not happy with how 641 00:33:22.160 --> 00:33:24.960 you're doing your stuff. That's an indication that you probably 642 00:33:24.960 --> 00:33:28.519 are doing it the wrong way or some thing. You know, 643 00:33:28.960 --> 00:33:31.640 typescripts and types. If you if you're having a heck 644 00:33:31.640 --> 00:33:34.319 of a time trying to type it correctly, and maybe 645 00:33:34.359 --> 00:33:37.200 that's not great, maybe you want to do something else. 646 00:33:37.400 --> 00:33:39.839 Yeah, that's kind of the thing about the eyes. From 647 00:33:39.839 --> 00:33:44.440 my from what I've been experiencing, is that AI. Even 648 00:33:44.480 --> 00:33:47.119 when you tell it, you know, we were talking about 649 00:33:47.119 --> 00:33:49.720 it before, about the rule files, before we start to the show. 650 00:33:50.079 --> 00:33:53.079 Even when you tell it you're an experienced developer, whatever, 651 00:33:53.440 --> 00:33:55.960 its still very much works like a junior in terms 652 00:33:56.000 --> 00:33:57.160 of its attitude. 653 00:33:57.319 --> 00:33:58.920 Whenever you tell me to do boss, I'm. 654 00:33:58.799 --> 00:34:04.160 In, Yeah, you're very They're very compliant, you know they Yeah, 655

00:34:04.200 --> 00:34:06.400 too compliant. They'll do it you tell them to do, 656 00:34:06.480 --> 00:34:10.239 even if it's totally stupid, or they'll try to do it. 657 00:34:11.719 --> 00:34:13.519 Yeah, unless you ask it the review and it will 658 00:34:13.519 --> 00:34:17.880 do a good good review. Yeah, I've actually asked it, like, hey, 659 00:34:17.960 --> 00:34:22.039 does this make sense? Yeah, but I understand, Like at 660 00:34:22.119 --> 00:34:25.599 least in Cursor, there's four different modes. There's agent mode, 661 00:34:25.880 --> 00:34:29.320 do what I tell you to do. There's a plan mode, 662 00:34:29.519 --> 00:34:32.920 the under underutilized plan modes I use the heck out 663 00:34:32.960 --> 00:34:36.440 of Basically, it'll ask you like, okay, before we do this, 664 00:34:36.519 --> 00:34:40.280 I've got some questions. And then there's ask and debug 665 00:34:40.360 --> 00:34:43.679 and ask is also I would say, like an underrated 666 00:34:43.880 --> 00:34:47.039 S tier mode where you can just ask the thing 667 00:34:47.119 --> 00:34:48.840 like how did you do that? How? 668 00:34:49.000 --> 00:34:51.639 I've used all of it except I've used all of 669 00:34:51.679 --> 00:34:52.719 them except the bug. 670 00:34:52.920 --> 00:34:53.599 Yeah, me too. 671 00:34:53.679 --> 00:34:55.400 I don't. I don't really get the bug. 672 00:34:55.800 --> 00:34:59.480 No, yeah, but you know there, but I haven't. I'm 673 00:34:59.519 --> 00:35:02.280 in love with the new the browser stuff, Oh my god, 674 00:35:02.360 --> 00:35:05.239 it's so great. What do you mean the browser integration 675 00:35:05.320 --> 00:35:08.280 and cursor. It's like, hey, this page doesn't look right, go, 676 00:35:08.360 --> 00:35:10.719 you know, go go fix it. And it's like okay, 677 00:35:10.880 --> 00:35:12.679 and pops up a browser takes screenshot. 678 00:35:12.800 --> 00:35:16.239 Like, oh yeah, yeah, I've seen the automated for you 679 00:35:17.639 --> 00:35:23.199 that I've been playing. You know. We we spoke personally 680 00:35:23.239 --> 00:35:26.159 a lot about mcps I've been playing with. I had 681 00:35:26.159 --> 00:35:29.480 a project. It's an interesting project. Again, we're digressing all 682 00:35:29.480 --> 00:35:33.159 over the place. I had an interesting project that I've 683 00:35:33.159 --> 00:35:39.519 been working on. Looking at work, we had a whole bunch, 684 00:35:39.719 --> 00:35:43.800 like we are. Scienceence has been around the block the 685 00:35:43.840 --> 00:35:49.480 company and for years. Their approach to testing when they 686 00:35:49.679 --> 00:35:54.199 kind of automated away from manual QA is to build 687 00:35:54.320 --> 00:35:57.079 a ton of end to end tests, so most of 688 00:35:57.119 --> 00:35:59.960 the coverage is achieved by end to end tests rather 689 00:36:00.079 --> 00:36:05.840 than by unit tests or integration tests. So there's like 690 00:36:05.840 --> 00:36:09.119 like thousands of end to end tests out

there, and 691 00:36:09.280 --> 00:36:11.679 they were built because they were built years ago. They 692 00:36:11.719 --> 00:36:14.920 were built using uh selenium. 693 00:36:14.599 --> 00:36:18.639 And oh dude, no. 694 00:36:18.920 --> 00:36:22.360 Actually, actually, even more accurately, they were built using if 695 00:36:22.400 --> 00:36:25.920 you've know if you heard about it about it using Serenity. 696 00:36:27.480 --> 00:36:28.159 No, I don't know that. 697 00:36:28.760 --> 00:36:32.960 So it's a layer above selenium that uses cucumber or 698 00:36:33.360 --> 00:36:39.079 Garca cucumber tests. I remember those exactly. So Serenity is 699 00:36:39.199 --> 00:36:44.880 basically Selenium plus cucumber. Yeh. So I was looking at 700 00:36:44.920 --> 00:36:52.400 the possibility of migrating these tests to play Right, because 701 00:36:52.440 --> 00:36:54.880 we write all our new tests and Playwright and play Right, 702 00:36:55.320 --> 00:37:00.719 despite all its own challenges, is much better Selenium, and 703 00:37:00.760 --> 00:37:03.320 most of our coding these days is done in typescript 704 00:37:03.440 --> 00:37:07.639 rather than Java, so we wanted to align on that stuff. 705 00:37:07.880 --> 00:37:11.559 But nobody is going to manually migrate thousands and thousands 706 00:37:11.559 --> 00:37:14.360 of tests. So I was looking at the possibility of 707 00:37:14.519 --> 00:37:18.159 using AI to automate as much of the effort as possible, 708 00:37:18.239 --> 00:37:20.800 and I kind of show that it was possible to 709 00:37:20.840 --> 00:37:26.679 reach somewhere between eighty to ninety percent automatic migration, which 710 00:37:26.760 --> 00:37:29.880 is not one hundred percent. It still needs a whole 711 00:37:29.920 --> 00:37:33.360 bunch of manual effort, especially that sometimes it's the challenging 712 00:37:33.400 --> 00:37:37.519 manual effort. But you know, getting rid of ninety percent 713 00:37:37.559 --> 00:37:42.760 of the work is definite plus. Yeah, And so I 714 00:37:42.880 --> 00:37:45.599 was talking about that because one of the things that 715 00:37:45.639 --> 00:37:48.679 I was trying to do is whenever it failed, I 716 00:37:48.880 --> 00:37:53.679 basically told it to try to use the playwright MCP 717 00:37:54.280 --> 00:37:57.679 server to try to figure out what the problem was 718 00:37:57.719 --> 00:38:01.519 and see if fix it automatic. Yeah. Yeah, yeah, sometimes 719 00:38:01.519 --> 00:38:04.519 it could, sometimes it couldn't, but it was interesting to 720 00:38:04.599 --> 00:38:05.800 watch the attempt. 721 00:38:05.880 --> 00:38:08.239 Yeah, and you're Ralphing basically Ralphing. 722 00:38:08.639 --> 00:38:12.199 Yeah, yeah, I'm Manual ralphingman. Yeah, you know what we 723 00:38:12.320 --> 00:38:15.679 digressed it to that, Let's talk about Ralph. Let's talk 724 00:38:15.719 --> 00:38:21.320 about I don't know, probably we would probably probably want 725 00:38:21.360 --> 00:38:25.159 to bring Mitt Pocock to speak about that, but nowhere anyway, 726 00:38:25.280 --> 00:38:27.320 So let's talk about Ralphing for a minute. 727 00:38:27.239 --> 00:38:27.400 Yeah. 728 00:38:27.440 --> 00:38:29.599 I was just watching a video his last night there 729 00:38:29.599 --> 00:38:29.800 you go. 730 00:38:30.000 --> 00:38:30.760 Was it about Ralphing? 731 00:38:31.320 --> 00:38:37.199 Uh No, it was about uh the open the SDK 732 00:38:37.320 --> 00:38:38.039 that you were talking. 733 00:38:37.800 --> 00:38:41.559 About, the yeah, the versl SK yeah versaill. 734 00:38:41.559 --> 00:38:41.840 Thank you. 735 00:38:41.920 --> 00:38:43.679 Yeah, he is he's got a full course on that, 736 00:38:43.800 --> 00:38:45.440 like a big course on that. 737 00:38:45.559 --> 00:38:47.519 So yeah, yeah, because I'm using it for a demo 738 00:38:47.599 --> 00:38:49.400 app so yeah, pretty cool. 739 00:38:49.639 --> 00:38:53.440 Yeah, Well, anyway, sorry, you sandseck and so by the way, 740 00:38:53.519 --> 00:38:58.960 so if your viewers or listeners uh hands like AI, 741 00:38:59.239 --> 00:39:02.119 all the tanzacle or is are completely independent of each other, 742 00:39:02.199 --> 00:39:04.199 so you don't you don't have to be on tanstac 743 00:39:04.239 --> 00:39:07.239 start to use tanac AI. You can use it on 744 00:39:07.639 --> 00:39:08.559 anything you want. 745 00:39:08.639 --> 00:39:10.039 Yeah, I used tanstack table. 746 00:39:11.480 --> 00:39:15.239 Yeah, the o G the OG library. 747 00:39:15.320 --> 00:39:16.760 Yeah. 748 00:39:16.800 --> 00:39:18.320 Okay, so ralphing you. 749 00:39:20.960 --> 00:39:24.159 Since you mentioned ten stack table. One of the things, 750 00:39:24.199 --> 00:39:27.199 one of the things, one of the things that I 751 00:39:27.360 --> 00:39:30.239 ran into problem with in that project was I don't 752 00:39:30.239 --> 00:39:36.360 remember which virtualized table library it was actually being used. 753 00:39:36.639 --> 00:39:42.159 But it's a real big hassle to do e t E. 754 00:39:42.159 --> 00:39:49.679 Effectively visual testing with with virtual table because the screen exist, 755 00:39:50.199 --> 00:39:54.239 you know, Chris, you kind of dependent. So I had 756 00:39:54.280 --> 00:39:57.679 tests failed. You know, my greated tests failed because it 757 00:39:57.760 --> 00:39:59.960 didn't make the virtual screen big enough. 758 00:40:00.199 --> 00:40:00.679 Yeah. 759 00:40:00.840 --> 00:40:07.760 Yeah, talk about brittle uh brittle actually and fragile tests 760 00:40:07.760 --> 00:40:10.199 and stuff like that. You know, your screen is too small. 761 00:40:10.320 --> 00:40:14.039 So that makes a virtual screen sixteen thousand pixels. 762 00:40:14.440 --> 00:40:18.960 Twenty exactly exactly, so that all the column headers fit in, 763 00:40:19.079 --> 00:40:23.679 because otherwise it's just the element is just not there. 764 00:40:23.719 --> 00:40:24.480 It's insane. 765 00:40:25.360 --> 00:40:28.960 Yeah, man, Like I I used

to do mac os 766 00:40:29.039 --> 00:40:31.960 programming back in like the version five days. Man, I 767 00:40:31.960 --> 00:40:33.679 I've been through it all, and we did a lot 768 00:40:33.679 --> 00:40:36.519 of virtual table stuff back then. I'm like, uh no, never, 769 00:40:36.719 --> 00:40:39.480 never again, never, Nope, nope, I don't want to man 770 00:40:39.519 --> 00:40:41.800 the scroll bar. I don't want to write virtual table code. 771 00:40:42.320 --> 00:40:44.840 It's a good business. Yeah. 772 00:40:44.880 --> 00:40:47.039 Well, so it was like, you know, maintaining cobal code. 773 00:40:47.119 --> 00:40:49.840 I guess nowadays, I wonder if like LM is going 774 00:40:49.880 --> 00:40:50.360 to put those. 775 00:40:50.239 --> 00:40:53.079 Guys out of Oh yeah, for sure. I well, I 776 00:40:53.079 --> 00:40:55.599 don't know. I mean, they need guys are literally dying 777 00:40:55.639 --> 00:40:57.360 off true. 778 00:40:57.440 --> 00:40:59.719 I guess yeah. If we all want to get paychecks, 779 00:40:59.719 --> 00:41:02.679 better better make sure the lmsn't want to understand that code. 780 00:41:02.800 --> 00:41:05.920 Yeah. Well, I would think that COBYL is like the 781 00:41:06.239 --> 00:41:10.519 optimal language for LMS, giving its verbosity. You know, ad 782 00:41:11.119 --> 00:41:17.119 add one to the two three in order increased valuable 783 00:41:17.239 --> 00:41:17.880 by four. 784 00:41:18.280 --> 00:41:20.199 Yeah, you're talking about digressions. 785 00:41:20.199 --> 00:41:20.360 Man. 786 00:41:20.400 --> 00:41:22.719 I was, I think watching some football game and I 787 00:41:22.760 --> 00:41:24.960 was like, oh, man, what does COBYL look like. It's 788 00:41:25.039 --> 00:41:27.320 like one of the programming language. I've never coded it, 789 00:41:27.480 --> 00:41:29.440 and I just got like three minutes and I'm like, nope, 790 00:41:29.719 --> 00:41:32.280 I'm good. I don't know, there's some things that I 791 00:41:32.360 --> 00:41:32.800 just don't need. 792 00:41:33.000 --> 00:41:35.960 There's a story about Ellen k the guy who invented 793 00:41:36.000 --> 00:41:40.159 object oriented programming, oh yeah, and small talk and whatnot, 794 00:41:40.239 --> 00:41:43.559 that he was walking around Stanford and suddenly he saw 795 00:41:43.599 --> 00:41:46.840 a crowd of people, so, uh, he walked over to 796 00:41:46.880 --> 00:41:49.360 see what it was. And it turns out that Admiral 797 00:41:49.440 --> 00:41:54.480 Grace Hopper and and you know, somebody mentioned she invented 798 00:41:54.559 --> 00:42:01.480 cobol and he said something like, oh, my condolences, and. 799 00:42:01.480 --> 00:42:05.159 She probably got a laugh out of that too. Yeah. 800 00:42:05.639 --> 00:42:08.280 Anyway, so back back to ten stack Ai. 801 00:42:08.519 --> 00:42:15.039 Right, let's go. Yes, we're almost talking about Ralph loops. 802 00:42:15.280 --> 00:42:15.519 Yeah. 803 00:42:15.920 --> 00:42:18.679 Sorry, but Jack you'll probably get this more than Dan. 804 00:42:18.760 --> 00:42:21.039 But sorry. When I started hearing about ralphing, I'm thinking 805 00:42:21.039 --> 00:42:22.960 about a Christmas story and a Red Rider be again. 806 00:42:23.159 --> 00:42:25.400 But that's I was thinking about Rolfing. I remember like 807 00:42:25.519 --> 00:42:27.960 there was a trend there for a while, like a 808 00:42:28.039 --> 00:42:31.639 massage technique where they like really crunch your bones and stuff. 809 00:42:32.599 --> 00:42:33.199 That's what it was. 810 00:42:34.039 --> 00:42:35.719 Okay, sorry, anyway, go ahead, Dan. 811 00:42:35.880 --> 00:42:41.719 So, so Ralph loop is is named after uh well, 812 00:42:42.280 --> 00:42:43.760 uh Simpsons Ralph. 813 00:42:44.199 --> 00:42:50.760 Yeah, it's yeah, because William is a persistent kid, just 814 00:42:50.760 --> 00:42:54.199 just grinds. He grinds. So that's what a Gralph loop does. 815 00:42:54.320 --> 00:42:54.920 It grinds. 816 00:42:55.079 --> 00:42:58.239 It's basically a wild loop. It's fine, Just correct me, 817 00:42:58.360 --> 00:43:02.440 correct me if I'm wrong. It's basically I forget who 818 00:43:02.480 --> 00:43:03.800 it was who invented it. 819 00:43:04.639 --> 00:43:07.559 He's some dude in some rancher in. 820 00:43:07.400 --> 00:43:12.119 Australia exactly h he was. He was trying AI and 821 00:43:12.159 --> 00:43:14.679 he was annoyed with the fact that he would a 822 00:43:14.920 --> 00:43:19.480 I would do some tasks for him, uh, produce an 823 00:43:19.480 --> 00:43:23.320 incorrect results and stop and and he just. 824 00:43:23.239 --> 00:43:24.760 Had to kick it in the button be like, no, 825 00:43:24.880 --> 00:43:28.719 that's wrong. Next, you know, try eventually you would get it, you. 826 00:43:28.599 --> 00:43:31.840 Know, yeah, Like on the fifth try, the AI would 827 00:43:31.840 --> 00:43:34.320 get it. So basically, if I under stand correct, the 828 00:43:34.400 --> 00:43:37.800 Ralph Wigham loop is basically running AI in a loop 829 00:43:38.039 --> 00:43:41.519 and every time you you run a sort of a 830 00:43:41.559 --> 00:43:45.239 test suite to check if you've reached your end condition 831 00:43:45.440 --> 00:43:48.960 or not, and if it didn't, you take everything that 832 00:43:49.039 --> 00:43:52.159 you got and basically use it in as an input 833 00:43:52.519 --> 00:43:54.239 to the looping all over again. 834 00:43:54.320 --> 00:43:57.199 So you everything you got, like a summary of what 835 00:43:57.280 --> 00:43:59.920 you got, plus the original. 836 00:43:59.519 --> 00:44:04.039 Contact context, plus the test results you failed in this 837 00:44:04.239 --> 00:44:06.719 that and the yeah and try it again. 838 00:44:07.159 --> 00:44:09.960 Yeah, yeah, yeah, And the idea is that you know, 839 00:44:10.000 --> 00:44:14.440 you if you get like the context window too big. 840 00:44:14.599 --> 00:44:16.800 I mean I'm starting to see like people empirically are 841 00:44:16.800

--> 00:44:19.679 starting to talk about like fifty percent, you know, sixty 842 00:44:19.719 --> 00:44:23.159 percent above that kind of range. The AI like the LM, 843 00:44:23.440 --> 00:44:26.320 it gets in every time you make a request, it 844 00:44:26.320 --> 00:44:29.679 sends an entire block back over to let's drop. 845 00:44:29.440 --> 00:44:32.559 Our context windows for a minute, because there's a critical 846 00:44:32.639 --> 00:44:33.280 aspect here. 847 00:44:33.519 --> 00:44:36.480 It really is managing your context window, and that that's 848 00:44:36.519 --> 00:44:38.000 what the Ralph FLU is trying to do is make 849 00:44:38.039 --> 00:44:41.840 sure that like it starts basically fresh on every single 850 00:44:41.960 --> 00:44:43.079 loop through. 851 00:44:43.000 --> 00:44:46.920 It takes back whatever it needs, erases everything else, and 852 00:44:47.000 --> 00:44:49.960 this way keeps the context manageable. 853 00:44:50.280 --> 00:44:52.840 Yeah yeah, because we will kind of empirically found like 854 00:44:52.920 --> 00:44:55.239 you know, as I say, like you know, fifty six percent, 855 00:44:55.360 --> 00:44:57.760 and like they they after that, the LM starts to 856 00:44:57.840 --> 00:45:01.159 quote unquote get dumb, and it starts like like only 857 00:45:01.280 --> 00:45:03.800 look at the beginning and the end of the context. 858 00:45:04.360 --> 00:45:06.760 So so basically the easiest way to see it is 859 00:45:06.800 --> 00:45:10.039 if you're working in an environment cloud code or cursor 860 00:45:10.159 --> 00:45:13.519 or whatever, and you're you're you're trying to get the 861 00:45:13.960 --> 00:45:18.039 and you basically have this chat that goes through your 862 00:45:18.199 --> 00:45:21.199 entire work day. And let's say you're working on the 863 00:45:21.239 --> 00:45:24.960 same problem the whole day. You would notice that towards 864 00:45:25.039 --> 00:45:28.280 the end of the day it ends up getting smarter 865 00:45:29.039 --> 00:45:32.920 and better understanding of what you're you're trying to achieve 866 00:45:32.960 --> 00:45:36.400 given everything that you've researched and investigated over the day. 867 00:45:36.719 --> 00:45:40.840 The other lem is actually getting dumber and producing incorrect 868 00:45:40.840 --> 00:45:42.599 results rather than better results. 869 00:45:42.840 --> 00:45:45.639 It sounds like getting tired after this exactly, it needs 870 00:45:45.639 --> 00:45:49.440 a coffee, right, Well, in this case, the coffee is 871 00:45:49.440 --> 00:45:51.119 is I don't know what that is. 872 00:45:51.880 --> 00:45:55.599 Oh geez, I've tried those manly get hard palpitations entirely 873 00:45:55.639 --> 00:45:58.840 different topics. But yeah, the idea being like they're they're 874 00:45:58.960 --> 00:46:01.519 vision of Rockstar is that you basically go back to 875 00:46:01.679 --> 00:46:04.920 the original context and say, like, okay, you know, starting 876 00:46:05.079 --> 00:46:09.079 is if we haven't done well starting with where we are, now, 877 00:46:09.559 --> 00:46:12.159 let's try and hit this, hit this again, you know. 878 00:46:12.480 --> 00:46:13.519 And so you don't have that. 879 00:46:13.440 --> 00:46:17.800 So basically, so basically the context window is the input 880 00:46:17.920 --> 00:46:22.639 that you send to the LM effectively right with that. 881 00:46:22.639 --> 00:46:25.480 Yeah, it's all the messages, all the system prompt and 882 00:46:25.760 --> 00:46:28.599 that includes like all the tool calls and everything else. 883 00:46:28.679 --> 00:46:33.159 So yeah, and it has limited size. So when you're 884 00:46:33.239 --> 00:46:38.360 having a really lengthy conversation with your LLLM effected, at 885 00:46:38.400 --> 00:46:43.159 one point or another, you'll exceed your context window size, 886 00:46:43.159 --> 00:46:46.440 which can vary between lms. I think it's like how 887 00:46:46.440 --> 00:46:47.519 many cases. 888 00:46:47.800 --> 00:46:51.480 It varies wildly between elms, but it'll do most of 889 00:46:51.519 --> 00:46:54.320 these clients will do like a compaction phase. We'll basically 890 00:46:54.400 --> 00:46:57.920 do a summarization and try to do like a Wigham loop. 891 00:46:58.199 --> 00:47:01.440 The trick is that you don't really know what the 892 00:47:01.440 --> 00:47:05.800 summarization did or said, so it could have gotten it wrong. 893 00:47:06.199 --> 00:47:08.599 So what I tend to do is my workflow is 894 00:47:08.599 --> 00:47:11.039 that I'll get like pretty deep into a conversation. I'll 895 00:47:11.039 --> 00:47:13.599 get to a point where it's like it's working, and 896 00:47:13.639 --> 00:47:15.920 then I'll have it build a markdown file that's like 897 00:47:15.960 --> 00:47:18.679 this is where we got to, and then I'll kill 898 00:47:18.719 --> 00:47:21.880 that conversation by you don't don't feel any attachment to 899 00:47:21.960 --> 00:47:23.079 that that version. 900 00:47:23.000 --> 00:47:27.400 So you close, if you close, if it's cursor, you 901 00:47:27.480 --> 00:47:30.000 close that chat window and open a new chat. 902 00:47:29.840 --> 00:47:32.480 With new chat window and say, here's your context, buddy, 903 00:47:32.840 --> 00:47:35.199 this is where we got to. Now let's start from 904 00:47:35.199 --> 00:47:38.559 there and yeah, as you say, like it keeps them 905 00:47:38.559 --> 00:47:41.400 smart because they like they like being in a small 906 00:47:41.440 --> 00:47:42.159 context window. 907 00:47:42.519 --> 00:47:46.239 As a results, I have a similar but somewhat different approach. 908 00:47:46.519 --> 00:47:49.239 I've gotten kind of enamored. We've spoken about it with 909 00:47:49.400 --> 00:47:53.559 rule files. So what I do is I constantly iteratively 910 00:47:53.679 --> 00:47:56.599 try to improve the rule file, and then at a 911 00:47:56.639 --> 00:48:00.519 certain point in time, I'll say, Okay, the rule file 912 00:48:00.519

--> 00:48:03.480 has reached a certain level, I'll close the conversation and 913 00:48:03.639 --> 00:48:06.960 basically just kill it. Instead of telling it summarize what 914 00:48:07.000 --> 00:48:10.840 we've had so far, I'll continuously look at my rule 915 00:48:10.880 --> 00:48:13.800 file as an up to date summarization of the chat 916 00:48:13.840 --> 00:48:17.119 and whatever preceded it to that point. And so I'll 917 00:48:17.159 --> 00:48:20.519 close the chat and I'm starting fresh just with that 918 00:48:20.920 --> 00:48:24.800 updated rule file. And I take a really iterative approach 919 00:48:24.840 --> 00:48:27.320 to updating the rule file all the time. I also, 920 00:48:27.679 --> 00:48:30.400 every once in a while tell it please compact the 921 00:48:30.480 --> 00:48:33.280 rule file because it starts to accumulate a lot of gunk. 922 00:48:33.639 --> 00:48:36.199 Yeah, a lot off. With a big file like that, 923 00:48:36.440 --> 00:48:39.719 you're going to blow the context window right off the bat, 924 00:48:39.840 --> 00:48:41.679 you know, And exactly That's one of things that folks 925 00:48:41.719 --> 00:48:44.480 are learning is like, Okay, so we've got our next 926 00:48:44.599 --> 00:48:46.800 JS or our big tan sax start app, and it's 927 00:48:46.800 --> 00:48:49.800 got some back end code, it's got some front end code. 928 00:48:50.239 --> 00:48:52.039 And what you want to do is kind of break 929 00:48:52.119 --> 00:48:55.760 out those rule files or architecture files such that like 930 00:48:55.800 --> 00:48:59.079 the React stuff is in its own file, so that 931 00:48:59.159 --> 00:49:01.880 the l im can basically say, hey, cool, all I'm 932 00:49:01.880 --> 00:49:04.119 working on is react to right now, So I'm going 933 00:49:04.159 --> 00:49:07.519 to go and use that as my my north star 934 00:49:07.599 --> 00:49:09.760 of architecture, and that's going to deal with hooks and 935 00:49:09.760 --> 00:49:11.159 all that sort of stuff, and I'm not going to 936 00:49:11.199 --> 00:49:13.079 deal with anything on the back end, so I can 937 00:49:13.159 --> 00:49:16.599 leave all of that out of the context and really 938 00:49:16.679 --> 00:49:19.159 really really managing those contexts. 939 00:49:19.199 --> 00:49:21.599 And the only thing I got just as a as 940 00:49:21.599 --> 00:49:24.679 an aside, and just I totally agree, and just to 941 00:49:24.800 --> 00:49:28.360 mention that you in most of the environments that I've seen, 942 00:49:28.360 --> 00:49:31.360 you can actually configure the rule file to pertain to 943 00:49:31.440 --> 00:49:36.360 a specific folder and all folders. Yeah, so that's or. 944 00:49:36.639 --> 00:49:38.800 You can tell rex you know whatever. 945 00:49:38.840 --> 00:49:41.400 Oh reg x, or or you can tell it just 946 00:49:41.519 --> 00:49:43.719 use it according to the context. But then it becomes 947 00:49:43.800 --> 00:49:44.440 kind of risky. 948 00:49:44.559 --> 00:49:44.760 Yeah. 949

00:49:45.159 --> 00:49:48.159 Yeah, jeez, I'm sorry. 950 00:49:48.320 --> 00:49:51.400 No, No, it's fine. Oh man. Yeah, so I actually 951 00:49:51.639 --> 00:49:54.320 I kind of bifurcate. I have like an architecture document, 952 00:49:54.760 --> 00:49:57.960 and then I've got a status document. And the status 953 00:49:57.960 --> 00:50:01.679 documents are generally transient. I can get you know, create them, 954 00:50:01.760 --> 00:50:03.639 delete them as part of a branch or whatever like that. 955 00:50:03.679 --> 00:50:07.079 But the target extra document is what sort of remains 956 00:50:07.519 --> 00:50:11.119 over time and that it's like the structure of the feature. 957 00:50:11.159 --> 00:50:14.880 And so your your mandatory, your or the one that 958 00:50:15.239 --> 00:50:20.079 is always applied is the basic architectural document, and the 959 00:50:20.119 --> 00:50:24.400 rest are either used transiitly or used for specific parts 960 00:50:24.400 --> 00:50:25.159 of the application. 961 00:50:25.440 --> 00:50:27.840 Yeah, yeah, I remember what I was going to talk about. 962 00:50:27.920 --> 00:50:33.199 So tool calls. So if you watch your your claude 963 00:50:33.239 --> 00:50:35.840 whatever you know, clod code or whatever, you know, the 964 00:50:36.039 --> 00:50:40.400 cursor as it's executing things like look, I'm doing a 965 00:50:40.400 --> 00:50:43.000 pm BX you know, blah blah blah blah blah all 966 00:50:43.039 --> 00:50:45.760 that stuff. Every one of those is a tool call, 967 00:50:45.920 --> 00:50:51.320 which means that it's actually, uh, the AI out in 968 00:50:51.320 --> 00:50:55.679 the cloud is basically telling your client is terminating your 969 00:50:55.679 --> 00:50:58.360 connection to the client with a thing that says, hey, 970 00:50:58.559 --> 00:51:00.360 right now, I want you to go in to keep 971 00:51:00.400 --> 00:51:04.119 this thing locally and then send back whatever the results 972 00:51:04.119 --> 00:51:06.400 of it were, and then it does that, and then 973 00:51:06.440 --> 00:51:09.960 it sends it back along with all of the contexts 974 00:51:10.239 --> 00:51:13.559 in another call back to the LM. So those round 975 00:51:13.559 --> 00:51:17.000 trips one are huge, and two, like every one of 976 00:51:17.039 --> 00:51:20.840 those little invocations is just chewing, chewing, chewing, chewing, chewing 977 00:51:20.920 --> 00:51:23.119 up more and more and more context. So even just 978 00:51:23.199 --> 00:51:25.840 like oh, I'm having itery through all the files and 979 00:51:25.880 --> 00:51:29.079 blah blah blah blah blah, like that's that's just non 980 00:51:29.199 --> 00:51:30.800 nom naming your context. 981 00:51:31.039 --> 00:51:33.199 So it's not just about the fact, you know, I 982 00:51:33.280 --> 00:51:35.599 kind of learned this the hard way again, going back 983 00:51:35.639 --> 00:51:39.000 to that example that they gave that initially, I would 984 00:51:39.039 --> 00:51:43.199 literally have the LLLM run all the tests to verify 985 00:51:43.320 --> 00:51:48.519 that whatever I updated didn't break anything. But I saw 986 00:51:48.559 --> 00:51:51.440 that not only was it taking a whole lot of time, 987 00:51:51.840 --> 00:51:54.559 it was also eating up the context. So instead I 988 00:51:55.039 --> 00:51:58.880 become I became much more strategic about which tests I 989 00:51:58.920 --> 00:52:00.440 would actually be run. 990 00:52:01.360 --> 00:52:03.840 And in fact, actually I think you could even be 991 00:52:03.880 --> 00:52:06.000 more for strategic. And I haven't done this myself, but 992 00:52:06.079 --> 00:52:08.480 like you know, and we all have these like great 993 00:52:08.519 --> 00:52:10.400 test runners that have all kinds of you know. 994 00:52:10.360 --> 00:52:13.280 Like beautiful output and all these oh blah blah blah 995 00:52:13.280 --> 00:52:16.320 blahlah thing, you know, and look you've done great. That's 996 00:52:16.360 --> 00:52:19.159 a lot of text, you know, it's coming out of 997 00:52:19.199 --> 00:52:22.440 those things, and that text gets added into the context. 998 00:52:22.559 --> 00:52:26.480 So maybe having your test runner basically just output these 999 00:52:26.480 --> 00:52:27.440 are the tests that failed. 1000 00:52:27.480 --> 00:52:33.000 Otherwise don't yeah, do not do anything for bos is 1001 00:52:33.039 --> 00:52:34.920 not great. And if you've got something where you've got 1002 00:52:34.920 --> 00:52:38.280 a bad setup, like I have a one PARTYGO shell 1003 00:52:38.360 --> 00:52:41.239 that's not particularly well set up, and sometimes like Claude 1004 00:52:41.239 --> 00:52:43.880 will have to like try three different ways to run 1005 00:52:44.199 --> 00:52:47.440 the command and get like a little gnarly about that. 1006 00:52:47.719 --> 00:52:52.119 Fix that, fix that, because that again, that's just round tripping, 1007 00:52:52.400 --> 00:52:54.480 and that's that's taking a lot of cost. And so 1008 00:52:54.840 --> 00:52:57.960 it's it's it's compound interest, right. So if you know, 1009 00:52:58.000 --> 00:52:59.920 when you think about all the tokens that you spent, 1010 00:53:00.039 --> 00:53:03.679 and every time you make a request, you're basically all 1011 00:53:03.760 --> 00:53:05.960 of the tokens you've just spent in the last request, 1012 00:53:06.079 --> 00:53:08.440 plus all the tokens that you're spending on this request. 1013 00:53:08.679 --> 00:53:11.719 That's what you're spending. So it's a lot, it's a lot. 1014 00:53:11.760 --> 00:53:14.400 That's how you get like those you know, how did 1015 00:53:14.440 --> 00:53:18.280 I spend a million a million tokens on this request? Oh? 1016 00:53:18.480 --> 00:53:21.079 Is because it's compounding interest interesting. 1017 00:53:22.000 --> 00:53:23.719 Yeah, it's definitely something to think about. 1018 00:53:24.800 --> 00:53:26.760 Yeah, there's no state, it's totally stayless. 1019 00:53:26.880 -->

00:53:31.079 I have to say that this year has been an 1020 00:53:31.119 --> 00:53:33.719 amazing learning experience year. 1021 00:53:33.719 --> 00:53:37.119 For me, right. Yeah, anybody who stayed in the field 1022 00:53:37.199 --> 00:53:39.599 is like, whoa dude, this is crazy. 1023 00:53:39.800 --> 00:53:44.760 Yeah. And the great thing is everybody's talking about how 1024 00:53:44.840 --> 00:53:47.480 AI is eating our jobs, and to an extent it 1025 00:53:47.639 --> 00:53:51.559 is and doing a certain extent it is. But if 1026 00:53:51.599 --> 00:53:54.719 you're if you're the one who knows how to properly 1027 00:53:54.880 --> 00:53:58.840 use the tools, you've ten xt yourself and you were 1028 00:53:59.079 --> 00:54:00.119 going to be into. 1029 00:54:00.519 --> 00:54:04.039 Yeah, that's what I'm saying. You know, Tanner ping me on, 1030 00:54:04.159 --> 00:54:07.239 like I think it was like Friday night and I'm 1031 00:54:07.519 --> 00:54:11.440 doubted like Blade Runner, you know, the symphony symphonic version 1032 00:54:11.480 --> 00:54:13.480 with my wife, and I'm getting this thing from him 1033 00:54:13.519 --> 00:54:15.800 saying like, hey, you know, I think I'm gonna look 1034 00:54:15.840 --> 00:54:18.800 into like builder making the builder on the website, and 1035 00:54:19.360 --> 00:54:22.519 you know, it was middle of the night and I'm like, sure, 1036 00:54:22.519 --> 00:54:24.159 why not go for it? And then like by the 1037 00:54:24.199 --> 00:54:27.119 next morning it was he was done, like he was 1038 00:54:27.119 --> 00:54:30.079 done with the initial release, like no problem, there you go. 1039 00:54:30.159 --> 00:54:31.920 And then he's adding features left and right, and it's 1040 00:54:31.960 --> 00:54:34.599 like he's just on a tear, like you know, this 1041 00:54:34.679 --> 00:54:37.760 is what LM's and we just were talking about. This 1042 00:54:37.880 --> 00:54:40.239 is like all the stuff that you thought you could do, 1043 00:54:40.559 --> 00:54:42.840 and then you're like, oh, maybe I'll get there eventually, 1044 00:54:42.880 --> 00:54:44.480 blah blah blah blah blah. It's I was like, it 1045 00:54:44.679 --> 00:54:49.480 just bang done, and it's it's really cool. So saying 1046 00:54:49.679 --> 00:54:51.960 the ironically funny thing is when you give it like 1047 00:54:51.960 --> 00:54:54.880 a really big request, like hey, let's arguitect this new 1048 00:54:54.880 --> 00:54:57.440 feature or whatever. I don't know what it is, but 1049 00:54:57.480 --> 00:54:59.000 Claude always comes back like, oh, this is going to 1050 00:54:59.039 --> 00:55:01.440 take it like four weeks this part in phase one 1051 00:55:01.519 --> 00:55:03.119 and two weeks for phase two, and it's like, no, 1052 00:55:03.239 --> 00:55:05.280 this could be like this is gonna be like two hours, 1053 00:55:05.480 --> 00:55:09.199 like we're really done. Yeah, crazy exciting times. 1054 00:55:10.159 --> 00:55:14.719 Or alternatively, I I'm working on a certain legacy application. 1055 00:55:15.280 --> 00:55:17.119 I was trying to figure out what's going on. I 1056 00:55:17.199 --> 00:55:19.519 was asking the other developers on the team. None of 1057 00:55:19.519 --> 00:55:22.400 them seem to recall exactly how the functionality I was 1058 00:55:22.440 --> 00:55:26.119 asking about actually works. So I switched to ask mode 1059 00:55:26.199 --> 00:55:30.800 and I told h Cursor, tell me explain this flow 1060 00:55:31.079 --> 00:55:33.599 for me to me please, And and you know it 1061 00:55:33.719 --> 00:55:36.679 just worked and it saved so much time. 1062 00:55:37.000 --> 00:55:39.199 Yeah, yeah, and what you really should do is have 1063 00:55:39.280 --> 00:55:42.239 it then build out an MD file that explained like 1064 00:55:42.679 --> 00:55:46.760 that documents that architecture. So the next time you so 1065 00:55:46.840 --> 00:55:49.119 what I actually for it to go on animal? 1066 00:55:49.239 --> 00:55:52.000 So what I actually did, what I actually did. And 1067 00:55:52.039 --> 00:55:55.719 that's another thing that I wanted to mention is again 1068 00:55:55.920 --> 00:55:58.440 maybe people are talking about it, maybe they are not, 1069 00:55:58.960 --> 00:56:03.760 but it's the rebirth of wikis. If you've got wiki 1070 00:56:03.760 --> 00:56:07.679 in your organization, you can upload the stuff into your 1071 00:56:07.719 --> 00:56:12.239 wiki and then in future calls you can tell the 1072 00:56:12.920 --> 00:56:15.880 agent here's the u r L for the relevant document 1073 00:56:17.840 --> 00:56:21.119 or alternatively copy paste if if you want to, you know, 1074 00:56:21.280 --> 00:56:27.400 avoid the roundtrip. But so we're actually for reasons, we 1075 00:56:27.480 --> 00:56:34.159 are into as at Lasian. So they also have this 1076 00:56:34.280 --> 00:56:39.079 kind of the thing called Confluence, which is their wiki thing. Yeah, 1077 00:56:39.119 --> 00:56:43.119 but it's got an m CP server, so I can 1078 00:56:43.199 --> 00:56:46.320 put I can actually tell it, you know, do put 1079 00:56:46.679 --> 00:56:49.639 that stuff that you just mentioned. Take that empty, because 1080 00:56:49.639 --> 00:56:53.480 otherwise where would I put and organize all those all 1081 00:56:53.519 --> 00:56:58.079 those empty files? So yeah, I guess, but then search 1082 00:56:59.079 --> 00:57:00.719 searching through them comes annoying. 1083 00:57:00.920 --> 00:57:03.199 Yeah, that's true. There's an interesting project you might want, 1084 00:57:03.280 --> 00:57:06.280 like called Vibe rules, and the idea is that you 1085 00:57:06.360 --> 00:57:11.800 package the markdown rules for the library with the library, 1086 00:57:12.719 --> 00:57:15.719 and then when you bring it into node modules, there's 1087 00:57:15.719 --> 00:57:18.480 a vibe rules executable that you can run that basically 1088 00:57:18.480 --> 00:57:23.880 pull extracts all those rules into the local environment so 1089

00:57:23.920 --> 00:57:26.320 that then claud or whatever your agent is can see 1090 00:57:26.360 --> 00:57:30.519 it and then you just I guess every time you 1091 00:57:30.599 --> 00:57:31.119 run it again. 1092 00:57:31.519 --> 00:57:34.800 It's also that with Wikis, I can hyperlink across, which 1093 00:57:34.840 --> 00:57:37.440 is much easier to do than it is in Get. 1094 00:57:37.719 --> 00:57:39.920 Okay, that's fair. I'm just used to like Wikis, where 1095 00:57:39.960 --> 00:57:42.559 the actual format is marked down, so it's kind of. 1096 00:57:42.519 --> 00:57:48.119 Like yeah, in Yeah, it's kind of annoying. In Confluence 1097 00:57:48.199 --> 00:57:51.679 it's marked down ish, which which is? 1098 00:57:51.679 --> 00:57:55.519 And this is why I like which is annoying Confluence. Man, 1099 00:57:55.559 --> 00:57:57.840 I've never had good experiences of Confluence. I'm hoping that 1100 00:57:57.840 --> 00:57:59.840 their search has gotten better, but man, their their search 1101 00:57:59.920 --> 00:58:02.039 was like you can use exact right term. 1102 00:58:01.880 --> 00:58:08.400 And you they've added AI capability. Surprise, surprise. Yeah, it's improved, 1103 00:58:08.519 --> 00:58:13.760 but it seemed kind of useless to me. And now 1104 00:58:13.800 --> 00:58:17.239 all of a sudden, I'm getting this extra value from it. Okay, 1105 00:58:17.679 --> 00:58:20.840 because I can tell, like I said, I can basically 1106 00:58:21.159 --> 00:58:26.159 use their MCP server to say, use this reference architecture 1107 00:58:26.280 --> 00:58:30.440 document when you're implementing a new project or something like that, 1108 00:58:31.199 --> 00:58:34.519 and it makes life easier. But you're right, it would 1109 00:58:34.559 --> 00:58:38.199 be easier if it were if it were pure empty files, 1110 00:58:38.239 --> 00:58:40.000 it would be a whole lot easier. 1111 00:58:40.119 --> 00:58:43.239 And the other thing is doing that with like older 1112 00:58:43.280 --> 00:58:44.960 projects is great because you're not going to version them 1113 00:58:45.000 --> 00:58:48.199 that often. But if you're doing like early days work 1114 00:58:48.280 --> 00:58:50.639 and you're going to be changing around the APIs and 1115 00:58:50.679 --> 00:58:54.000 you've got like a branch where you've changed the APIs again, 1116 00:58:54.840 --> 00:58:56.679 and you know you want to bring that in and 1117 00:58:56.719 --> 00:58:58.920 see how it works, it's like, oh, now I've got 1118 00:58:59.039 --> 00:59:04.599 you know, the canonical document is on the old API 1119 00:59:05.039 --> 00:59:06.920 or it's on the new API. And it's like, you 1120 00:59:06.960 --> 00:59:10.440 want to have the document with the code and with 1121 00:59:10.480 --> 00:59:11.079 that version. 1122 00:59:11.480 --> 00:59:14.960 Oh, I agree, and and rule files definite for example, 1123 00:59:15.079 --> 00:59:18.199 definitely go with the code I've switched from. 1124 00:59:18.280 --> 00:59:18.480 Cool. 1125 00:59:19.880 --> 00:59:23.320 You know again people are thinking, you know, people often 1126 00:59:23.400 --> 00:59:28.519 don't put the dot cursor folder in. They ignore it 1127 00:59:28.599 --> 00:59:32.639 and get and or the dot vs code folder. And 1128 00:59:32.679 --> 00:59:36.199 I and I always take the opposite approach I understand 1129 00:59:36.239 --> 00:59:39.000 that some people like to have their own independent setups, 1130 00:59:39.199 --> 00:59:42.079 but I won't use a curse word, but you know, efit, 1131 00:59:42.480 --> 00:59:44.800 you know, I'm going to give you mine because I 1132 00:59:44.880 --> 00:59:48.159 created this project and now with the rule and now 1133 00:59:48.159 --> 00:59:51.400 with the rule files, it's doubly So yeah. 1134 00:59:52.199 --> 00:59:54.519 Yeah, there's there's a contributor on tansak who was just 1135 00:59:54.519 --> 00:59:57.199 asking me just recently if she had a good idea 1136 00:59:57.840 --> 01:00:04.360 that around a like an id slash agent agnostic way 1137 01:00:04.440 --> 01:00:08.519 of specifying rules and all that, and then just having 1138 01:00:08.519 --> 01:00:10.639 a compiler or like a little command line thing that 1139 01:00:10.679 --> 01:00:12.880 basically take that agnostic version and just like, oh, I'm 1140 01:00:12.920 --> 01:00:16.280 mind curser, like, just run that. Now, give me the files. 1141 01:00:16.559 --> 01:00:18.519 And I think it's a great idea. And so I 1142 01:00:18.559 --> 01:00:20.719 had Claude go and do some competitive research and turns 1143 01:00:20.760 --> 01:00:23.639 out there's nothing like that out there, so hopefully she's 1144 01:00:23.679 --> 01:00:25.920 gonna get into it and try it out. 1145 01:00:26.079 --> 01:00:28.440 It doesn't seem like a heavy lift. There are no 1146 01:00:28.559 --> 01:00:31.639 pretty similar they're more or less all marked down, more 1147 01:00:31.719 --> 01:00:34.079 or less maybe with a little bit of a header 1148 01:00:34.119 --> 01:00:36.480 that kind of explains when it should be used, but 1149 01:00:36.519 --> 01:00:38.599 other than that, they're essentially all just marked out. 1150 01:00:38.679 --> 01:00:42.519 Yeah, and having like Claude do competitive research for you, 1151 01:00:42.679 --> 01:00:46.639 underrated man, underrated. Go for it like it's will save 1152 01:00:46.679 --> 01:00:48.679 you so much time. Like, oh, I don't know, I 1153 01:00:48.679 --> 01:00:51.920 can't find this. You know library that does this and that? Cool? Great, 1154 01:00:52.159 --> 01:00:54.840 go ask Claude. Maybe it'll find it for you, and 1155 01:00:54.880 --> 01:00:57.239 maybe it'll say cool, this is an area where nobody 1156 01:00:57.599 --> 01:01:00.400 is in and you should build it. But don't go 1157 01:01:00.440 --> 01:01:03.320 into the just randomly say well, I have this great 1158 01:01:03.320 --> 01:01:06.519 idea and then build the fifteenth version of a you know, 1159

01:01:06.679 --> 01:01:08.920 I don't know some library that does yeah, I don't 1160 01:01:08.920 --> 01:01:10.880 know table virtual tables. 1161 01:01:11.239 --> 01:01:14.400 Actually the bigger the bigger thing is how ais are 1162 01:01:14.440 --> 01:01:22.360 now both the democrat is a demo democratized, democratizing the 1163 01:01:22.480 --> 01:01:24.719 SaaS market, but also kind of killing it. 1164 01:01:24.880 --> 01:01:29.199 Yeah yeah, yeah. And you know, like I remember just 1165 01:01:29.239 --> 01:01:34.559 recently Germo said, I don't know, I don't like this app. 1166 01:01:35.400 --> 01:01:37.719 You know, it's not being maintained. I think I'm right, 1167 01:01:37.760 --> 01:01:39.400 and I'm just gonna have it. I'm gonna have Claude 1168 01:01:39.440 --> 01:01:41.960 run a new version. It's like, dude, you know, people 1169 01:01:42.000 --> 01:01:44.519 could do the same thing with Next days app. You know, 1170 01:01:45.079 --> 01:01:47.800 let's be honest, like, especially if you're on pages Router, 1171 01:01:48.159 --> 01:01:50.480 like I've actually got a markdown document that basically says, 1172 01:01:50.480 --> 01:01:52.480 here's here's how you go and take mark down the 1173 01:01:52.519 --> 01:01:55.039 pages writer version, which most people are still on because 1174 01:01:55.039 --> 01:01:57.599 the app writer really didn't make all that good inroads 1175 01:01:57.639 --> 01:01:59.760 and a lot of people have problems with it. A 1176 01:01:59.760 --> 01:02:03.920 lot of pages riders can yeah, just be easily, yeah, but. 1177 01:02:04.280 --> 01:02:07.840 Just be careful with that. So recently and we'll finish 1178 01:02:07.920 --> 01:02:10.679 that with that and probably switch over to picks. But 1179 01:02:11.360 --> 01:02:15.400 I recently saw tweet by the CEO of one of 1180 01:02:15.440 --> 01:02:19.639 those AI building tools, Base forty four, saying. 1181 01:02:19.679 --> 01:02:21.599 That I haven't heard that one. That's a new one. 1182 01:02:21.599 --> 01:02:25.239 I mean, they're actually they were bought by Wix, and 1183 01:02:25.320 --> 01:02:28.440 Wix is pushing them to users, so they're actually like 1184 01:02:28.559 --> 01:02:31.119 one of the they're they compete with Lovable and stuff 1185 01:02:31.159 --> 01:02:33.840 like that. Really getting a lot of traction, it seems, 1186 01:02:33.960 --> 01:02:39.000 I don't know. But anyway, so he said that he's 1187 01:02:39.280 --> 01:02:46.800 seen customers replacing the Salesforce CRM with something they whip 1188 01:02:46.920 --> 01:02:51.599 up with an AI tool okay, and just not pay 1189 01:02:51.679 --> 01:02:55.840 the one hundred k dollars a year to Salesforce and 1190 01:02:55.880 --> 01:03:03.159 instead basically get it for free. Uh and and somebody responded, well, 1191 01:03:03.239 --> 01:03:05.760 you know if you can avoid paying and get eighty 1192 01:03:05.760 --> 01:03:08.920 percent of functionality, it might be very worthwhile for you. 1193 01:03:09.000 --> 01:03:12.159 And I said sometimes yes, sometimes. 1194 01:03:12.719 --> 01:03:14.480 Like yeah, you don't you don't want. 1195 01:03:14.280 --> 01:03:16.719 Any percent of a cure for cancer? 1196 01:03:17.000 --> 01:03:20.360 Yeah, exactly, And you don't want to, Like if you're 1197 01:03:20.800 --> 01:03:23.639 that's not the intellectual property of your company, then don't. 1198 01:03:24.159 --> 01:03:26.079 It's going to be TechEd at, like you can't just 1199 01:03:26.239 --> 01:03:29.559 you know, oh, you know we're going to order lunch. Okay, 1200 01:03:29.639 --> 01:03:32.239 well let's write a new version of uh, you know, 1201 01:03:32.400 --> 01:03:36.119 door dash. No, dude, you just use rignor dash like, 1202 01:03:36.280 --> 01:03:38.039 you know, don't don't do that. 1203 01:03:38.239 --> 01:03:41.239 Well door dash. It's not the app, it's it's the. 1204 01:03:41.719 --> 01:03:42.159 Yeah, I know. 1205 01:03:42.199 --> 01:03:45.840 I'm just saying, like, yeah, but a different example, A 1206 01:03:45.840 --> 01:03:48.920 different example. I think an actual legitimate example would be 1207 01:03:49.400 --> 01:03:52.239 you probably don't want to rewrite turbo tax because you 1208 01:03:52.280 --> 01:03:58.480 don't want the AI to hallucinate tax exemption laws. You 1209 01:03:58.519 --> 01:04:01.559 could imagine you could end up in prison like that. 1210 01:04:02.119 --> 01:04:05.760 Yeah, little little downsides, little downside, all right. 1211 01:04:05.800 --> 01:04:08.920 So with that hallucination or avoiding hallucination, we're going to 1212 01:04:08.960 --> 01:04:10.320 wrap it up. I got a hard stop here in 1213 01:04:10.320 --> 01:04:11.280 about nine minutes. 1214 01:04:11.880 --> 01:04:12.159 Cool. 1215 01:04:12.199 --> 01:04:15.239 So with that, first of all, thank you Jack for 1216 01:04:15.320 --> 01:04:15.800 coming on. 1217 01:04:16.239 --> 01:04:17.239 Oh always pressure. 1218 01:04:17.360 --> 01:04:18.599 Always nice to meet a neighbor. 1219 01:04:18.800 --> 01:04:20.840 Yeah, let's get together, yes for sure. 1220 01:04:21.599 --> 01:04:23.440 With that, will move to picks. Picks is the part 1221 01:04:23.440 --> 01:04:25.679 of the show where we get to talk about whatever 1222 01:04:25.679 --> 01:04:28.639 we want to, whether it's tech, non tech, books, movies, 1223 01:04:29.159 --> 01:04:31.119 and my dad jokes. So I will start off with 1224 01:04:31.320 --> 01:04:33.480 what is, in my opinion, the high point of every episode, 1225 01:04:33.519 --> 01:04:36.639 which is the dad jokes of the week, and then 1226 01:04:36.960 --> 01:04:37.760 we'll save the guests. 1227 01:04:37.800 --> 01:04:40.360 I haven't heard of them, so I'm looking forward to it. 1228 01:04:40.480 --> 01:04:42.360 Yeah, yeah, well hopefully you're looking forward to them. 1229 01:04:42.360 --> 01:04:43.239 We'll see. So. 1230 01:04:45.239 --> 01:04:47.199 First of all, so when I become a lawyer, which 1231 01:04:47.320

--> 01:04:49.159 eventually i'll do it, maybe I'll use AI to study 1232 01:04:49.280 --> 01:04:52.800 become a lawyer. I want to defend a penguin, so 1233 01:04:52.880 --> 01:04:55.199 that way I can say the words your honor. Clearly, 1234 01:04:55.239 --> 01:04:56.840 my client is not a flight risk. 1235 01:04:57.880 --> 01:05:04.079 Right. Just for that, I'll tell you a quick lawyer joke. 1236 01:05:04.239 --> 01:05:08.000 So I guess this young lawyer is sitting open the 1237 01:05:08.280 --> 01:05:12.079 new office. He's like in the rundown part of town, 1238 01:05:12.360 --> 01:05:15.840 zero clients, no traction on whatsoever, is sitting alone in 1239 01:05:15.840 --> 01:05:18.320 his office hoping for improvement, and all of a sudden 1240 01:05:18.400 --> 01:05:21.480 there's this puff of smoke, and the devil appears oh 1241 01:05:21.559 --> 01:05:26.119 and says, I will make you the most successful, rich, 1242 01:05:26.320 --> 01:05:30.599 famous lawyer in the entire country. But in return, your wife, 1243 01:05:30.679 --> 01:05:33.840 your kids, your entire family, or all your friends will 1244 01:05:33.880 --> 01:05:37.400 spend eternity in hell. The lawyer looks at them and goes, 1245 01:05:37.719 --> 01:05:38.760 what's the catch. 1246 01:05:38.559 --> 01:05:39.599 Anyway, man? 1247 01:05:40.960 --> 01:05:43.760 Two more quick ones. Cars these days have way too 1248 01:05:43.760 --> 01:05:46.440 many gadgets. For example, I tried to reverse the other 1249 01:05:46.519 --> 01:05:48.440 day and I played a video of somebody getting run 1250 01:05:48.480 --> 01:05:51.440 over by a car, and then finally, homa in a 1251 01:05:51.480 --> 01:05:54.320 sentence can make a huge difference. For instance, let's eat 1252 01:05:54.519 --> 01:05:58.079 comment Frank has a completely different meaning from Frank is 1253 01:05:58.119 --> 01:05:58.719 in a coma? 1254 01:05:58.800 --> 01:06:00.840 Right, Yes, that is true. 1255 01:06:02.559 --> 01:06:05.320 You know the classic one is let's eat Frank versus 1256 01:06:05.440 --> 01:06:06.440 let's eat Frank. 1257 01:06:07.679 --> 01:06:11.960 Yeah, but your is a good one. I liked it, Okay, 1258 01:06:12.000 --> 01:06:15.719 all right, Dan, what do you got? Okay? So I 1259 01:06:15.760 --> 01:06:18.639 thought about talking and having a tech pic, but it 1260 01:06:18.679 --> 01:06:20.920 was Ralph flops and we ended up talking about it. 1261 01:06:21.000 --> 01:06:25.719 I just mentioned that Matt Pocock did an excellent video 1262 01:06:25.760 --> 01:06:29.360 about it. So if you're interested, just search Ralph Wigham 1263 01:06:29.639 --> 01:06:34.159 uh and find uh Matt's video and watch it. Because 1264 01:06:34.159 --> 01:06:38.400 he gives a very thorough and interesting explanation, is very engaging. 1265 01:06:38.679 --> 01:06:41.239 Should probably have him on the show again. So that's 1266 01:06:41.280 --> 01:06:45.119 that would actually be a pick. Uh. And another pick 1267 01:06:45.280 --> 01:06:52.199 is I'm reading a series of books called the Dresden Files. Uh. 1268 01:06:52.280 --> 01:06:55.039 It's it has nothing to do with the city Dressden. 1269 01:06:55.199 --> 01:07:00.159 It's with a person called Dresden, and it's kind kind 1270 01:07:00.199 --> 01:07:06.559 of a gum shoe novel meets a fantasy. He's simultaneously 1271 01:07:06.599 --> 01:07:11.360 a private investigator and a wizard. It happens, it takes place, 1272 01:07:11.440 --> 01:07:14.079 it takes place in modern Yeah, it takes place in 1273 01:07:14.119 --> 01:07:21.519 modern day Chicago with wizardry, with wizardry. Yes, And it's 1274 01:07:21.559 --> 01:07:25.480 it's all told in first person. It's it's it's like reading. 1275 01:07:25.519 --> 01:07:28.920 It's very engaging. It's it's there's a lot of action. 1276 01:07:29.440 --> 01:07:32.719 So if you like wizards duking it out with all 1277 01:07:32.760 --> 01:07:35.480 sorts of with other wizards or monsters, it's full of 1278 01:07:35.480 --> 01:07:40.119 that stuff. And they're like a ton of books. There 1279 01:07:40.519 --> 01:07:43.519 are how many books in this series, I think like 1280 01:07:44.079 --> 01:07:48.840 seventeen books in the series, so, and each one is 1281 01:07:48.920 --> 01:07:51.960 kind of self contained, so you can read them independently, 1282 01:07:52.360 --> 01:07:55.840 but I would recommend read just reading them in order 1283 01:07:56.519 --> 01:08:01.960 and thereby Jim Butcher. Sure, they actually apparently tried to 1284 01:08:02.000 --> 01:08:04.599 make a TV series out of this, like four or five, 1285 01:08:04.719 --> 01:08:07.280 six years ago, and it only ran for one season, 1286 01:08:11.000 --> 01:08:13.159 The Dress and File. The Butcher is the name of 1287 01:08:13.159 --> 01:08:19.920 the author anyway, So yeah, that would be one pick. 1288 01:08:21.039 --> 01:08:24.479 The other pick is a really guilty pleasure. It's a 1289 01:08:24.600 --> 01:08:30.800 channel on YouTube by this guy, Arvin Haddad. He what 1290 01:08:30.880 --> 01:08:35.880 he does is he he apparently helps the ultra rich 1291 01:08:36.079 --> 01:08:37.119 purchase real estate. 1292 01:08:37.560 --> 01:08:40.479 Oh that sounds great and. 1293 01:08:40.239 --> 01:08:44.680 So but he has this really amusing channel where it 1294 01:08:44.720 --> 01:08:47.880 turns out that whenever, like this one hundred million dollar 1295 01:08:48.159 --> 01:08:51.640 villa whatever goes on the market, they create this sort 1296 01:08:51.640 --> 01:08:53.159 of a marketing video. 1297 01:08:52.920 --> 01:08:55.159 For well, of course it's one hundred million dollars. 1298 01:08:55.439 --> 01:08:58.399 Yeah. So it's like a house tour, virtual house tour 1299 01:08:58.439 --> 01:09:01.279 where this guy walks. So so what he does is 1300 01:09:01.319 --> 01:09:06.079 he takes his videos and analyzes them and makes and 1301 01:09:06.279 --> 01:09:09.479 just shows how ridiculous some of these houses are. So 1302

01:09:09.560 --> 01:09:13.319 he really jokes about it the whole the whole time. So, 1303 01:09:13.680 --> 01:09:16.319 like they show you, look at this amazing kitchen, and 1304 01:09:16.359 --> 01:09:18.279 he said, but it's an awful kitchen. 1305 01:09:18.720 --> 01:09:22.359 Like or totally not ergonomic, you know. 1306 01:09:22.680 --> 01:09:24.960 Yeah, no, but it's got to the point where his 1307 01:09:25.079 --> 01:09:27.840 channel is so popular even though he pokes fun at 1308 01:09:27.880 --> 01:09:30.960 all their videos, all the content creators want him to 1309 01:09:31.000 --> 01:09:35.000 critique their videos because it gets them more engagement exactly. 1310 01:09:35.439 --> 01:09:39.239 But you see like these one hundred million dollar houses, 1311 01:09:39.279 --> 01:09:41.600 which would say, I would not be caught dead living 1312 01:09:41.680 --> 01:09:43.880 in a house like that, even if I could afford it, 1313 01:09:45.039 --> 01:09:48.359 because it's like you're like living in an airport terminal. 1314 01:09:48.520 --> 01:09:49.880 It's like it's horrible. 1315 01:09:50.319 --> 01:09:53.920 Yeah, it's too generic, doesn't have any life of vitality 1316 01:09:53.960 --> 01:09:54.199 to it. 1317 01:09:54.520 --> 01:09:54.720 Yeah. 1318 01:09:54.920 --> 01:09:57.760 Yeah, it's like we have a thirty two undred square 1319 01:09:57.760 --> 01:09:59.720 foot house and it's like there are parts of the 1320 01:09:59.720 --> 01:10:02.720 house almost rarely ever in, you know, and it's like 1321 01:10:02.760 --> 01:10:05.199 that I can only imagine how many square feet million 1322 01:10:05.239 --> 01:10:06.600 one hundred million dollar houses. 1323 01:10:07.159 --> 01:10:10.520 Yeah, and again, So it's it's really it's really amusing. 1324 01:10:10.760 --> 01:10:13.840 Enjoy watching his video videos is really engaging. He has 1325 01:10:13.880 --> 01:10:16.439 a great sense of humor. It's it's and it's fun 1326 01:10:16.479 --> 01:10:18.720 to poke. It's fun to poke fun at the rich 1327 01:10:18.760 --> 01:10:23.159 and famous. So yeah, that would be my uh my 1328 01:10:23.239 --> 01:10:25.640 third pick, I guess, And those are my picks for today. 1329 01:10:25.680 --> 01:10:29.239 All right, I think I'm only at one, so I 1330 01:10:29.479 --> 01:10:33.399 have gotten My family hooked on open shocks bone conducting headphones. 1331 01:10:33.600 --> 01:10:36.840 So these are they keep your ears like open and 1332 01:10:36.880 --> 01:10:39.359 they rest on the back of your ear and the 1333 01:10:39.720 --> 01:10:43.760 like right right before your ear, and they are phenomenal. 1334 01:10:43.840 --> 01:10:46.600 There's like a headset version that you can use for meetings. 1335 01:10:46.680 --> 01:10:49.079 There's a running and there's a kind of running version. 1336 01:10:49.199 --> 01:10:50.119 That's the one I use. 1337 01:10:50.560 --> 01:10:52.239 I think there is a version for swimming. 1338 01:10:52.439 --> 01:10:54.359 Yeah, And I was just gonna say, my wife has 1339 01:10:54.399 --> 01:10:56.640 gotten back into swimming in a big way. She's doing 1340 01:10:56.960 --> 01:10:59.520 butterfly laps and all this stuff. And when she finally 1341 01:10:59.600 --> 01:11:02.239 kind of felt good back in the pool again, she 1342 01:11:02.479 --> 01:11:05.640 tried the headphones on. And they have they store MB 1343 01:11:05.720 --> 01:11:09.159 three files like old school like MB three files, which 1344 01:11:09.199 --> 01:11:11.760 I had to go source that. That was kind of fun. 1345 01:11:11.760 --> 01:11:14.720 But still and also bluetooth. I guess we're gonna do that. 1346 01:11:15.359 --> 01:11:21.000 But apparently she's saying like underwater. They sound phenomenal like underwater, 1347 01:11:21.239 --> 01:11:23.920 which is crazy. So that's that's pretty pretty cool. 1348 01:11:24.000 --> 01:11:29.640 Kind of uses your body is kind of this uh yeah, vibrate. 1349 01:11:29.840 --> 01:11:31.800 Right phone conduction, Yeah, exactly. 1350 01:11:31.880 --> 01:11:35.840 It's it's really interesting. They're very lightweight, but I've not 1351 01:11:35.880 --> 01:11:37.960 I've never actually tried to use them. 1352 01:11:38.439 --> 01:11:40.319 I would try it out. They're not like at least 1353 01:11:40.319 --> 01:11:43.840 in the US, they're not super expensive, and they're very lightweight. 1354 01:11:44.000 --> 01:11:46.880 And also they kind of are one piece, you know, 1355 01:11:46.960 --> 01:11:49.199 the issue of things falling all out all over the place. 1356 01:11:49.640 --> 01:11:52.199 And then the other thing is after I get back 1357 01:11:52.199 --> 01:11:54.079 from a run, I just put one of the faucet 1358 01:11:54.279 --> 01:11:58.600 and it's like, don't there you go? Clean love run 1359 01:11:59.159 --> 01:12:01.880 there you go? Yeah, gotta do something, all right? 1360 01:12:01.920 --> 01:12:03.560 Yeah? Those are those are cool. A buddy of mine 1361 01:12:03.560 --> 01:12:05.359 as a firefighter out on the west side of town 1362 01:12:05.399 --> 01:12:07.359 here and they had some of their masks that had 1363 01:12:07.399 --> 01:12:10.399 those built in and I tried them on. They're crazy crazy, 1364 01:12:10.439 --> 01:12:12.640 How good they sound. Yeah, it's really weird thinking it's 1365 01:12:12.640 --> 01:12:16.479 not coming through your ears, but yeah, yeah, yeah cool. 1366 01:12:16.560 --> 01:12:18.600 All right, Well I got a jump, so okay, I've 1367 01:12:18.640 --> 01:12:22.279 been talking for quite a while. Anyways, Thanks everybody for listening, 1368 01:12:22.279 --> 01:12:24.800 Thank you Jack for coming, and we'll talk at y'all 1369 01:12:24.840 --> 01:12:26.079 next time on JavaScript Jabber 1370 01:12:26.279 --> 01:12:27.720 Bye, see you next time.