WEBVTT 1 00:00:05.240 --> 00:00:09.039 Hey everybody, welcome back to another episode of JavaScript Jabber. 2 00:00:09.400 --> 00:00:12.800 This week on our panel, we have Steve Edwards. 3 00:00:12.720 --> 00:00:16.079 Yo yo yo to imitate AJ coming at you from 4 00:00:16.079 --> 00:00:18.480 a cloudy bit warming up Portland area. 5 00:00:20.199 --> 00:00:23.719 I'm Charles max Wood from Top End Devs and this 6 00:00:23.800 --> 00:00:29.199 week we're talking to Matthew henij Now, Matthew, you lived 7 00:00:29.239 --> 00:00:32.439 near me as far as I remember. We were introduced 8 00:00:32.439 --> 00:00:35.399 by my neighbor and we had a good talk about 9 00:00:35.439 --> 00:00:37.280 AI and I thought, hey, let's let's have you on 10 00:00:37.320 --> 00:00:40.200 the show and dive into this stuff. So you want 11 00:00:40.200 --> 00:00:42.439 to let people know what else they ought to know 12 00:00:42.479 --> 00:00:46.679 about you and then we can roll from there. Sure. 13 00:00:46.799 --> 00:00:50.799 Yeah, it's great to be on the show. So my 14 00:00:50.880 --> 00:00:56.079 name is Matthew Henich. I'm full stack developer professional last 15 00:00:56.079 --> 00:00:57.520 twenty years, but doing answers. 16 00:00:57.520 --> 00:00:59.679 I was a kid so about thirty years ago. 17 00:01:01.000 --> 00:01:05.680 Always love jumping into new technologies and so it's working 18 00:01:05.719 --> 00:01:10.120 on a project that started working on called wows dot 19 00:01:10.159 --> 00:01:16.159 AI about three years ago, and and then chat JPT 20 00:01:16.359 --> 00:01:19.000 three point five came out about two years ago and 21 00:01:19.159 --> 00:01:22.159 it just blew my mind. Saw a lot of huge 22 00:01:22.439 --> 00:01:28.200 utility from it and got super involved with starting using programming, 23 00:01:28.560 --> 00:01:34.319 using AI. So, yeah, I'm also from I'm from Lehigh, Utah. 24 00:01:34.519 --> 00:01:38.280 So yeah, it's great to be on the show. Thanks 25 00:01:38.319 --> 00:01:38.799 for having me. 26 00:01:39.799 --> 00:01:42.719 Yeah, thanks for coming. Yeah. So when he says Lehigh, Utah, 27 00:01:42.719 --> 00:01:49.040 that's also where I'm at, I'm same town. So anyway, so, yeah, 28 00:01:49.200 --> 00:01:52.879 you were showing me when we talked before about your system. 29 00:01:53.359 --> 00:01:55.200 What is it w AOS? I don't even know what 30 00:01:55.359 --> 00:01:56.040 that stands for. 31 00:01:56.480 --> 00:02:00.879 Yeah, stands for a Web app operating system. It's a 32 00:02:01.000 --> 00:02:05.040 way if using low code tools to build web apps 33 00:02:05.239 --> 00:02:09.400 and then you can create AI workflows that basically control 34 00:02:09.520 --> 00:02:11.000 the web app in real time. 35 00:02:12.039 --> 00:02:14.039 Yeah. It kind of reminded me a little bit of 36 00:02:15.280 --> 00:02:19.280 what's it called zappier, except put prompts in instead of 37 00:02:20.000 --> 00:02:24.319 connecting to APIs for different products. That's a good way 38 00:02:24.360 --> 00:02:28.599 to go. Yeah. Yeah, So I guess just to dive 39 00:02:28.639 --> 00:02:32.159 in because at all the coding meetups that I go 40 00:02:32.240 --> 00:02:36.439 to anymore, everybody's talking about AI. They're excited about it, 41 00:02:36.479 --> 00:02:39.199 They're excited about what you can do with it. A 42 00:02:39.280 --> 00:02:42.360 lot of people are, you know, diving into different aspects, 43 00:02:42.360 --> 00:02:45.400 you know, whether it's generating text or images or videos 44 00:02:45.520 --> 00:02:49.439 or anything like that. I'm a little curious to just 45 00:02:50.159 --> 00:02:51.719 I guess kind of get the state of the art 46 00:02:51.759 --> 00:02:55.039 as far as you see it of AI and how 47 00:02:55.080 --> 00:02:57.360 people might or might not be using it. 48 00:02:59.240 --> 00:03:03.960 Sure, Yeah, Yeah. So it's kind of interesting because AI 49 00:03:04.680 --> 00:03:07.039 is concept has been around for a very long time, 50 00:03:09.000 --> 00:03:10.759 and kind of what a lot of people are now 51 00:03:10.800 --> 00:03:13.759 referring to with AI is more of like generative AI, 52 00:03:14.520 --> 00:03:19.400 so using different types of architectures like auto regression, which 53 00:03:19.439 --> 00:03:24.800 is kind of what l MS use, like CHATCHBT or claude, 54 00:03:26.120 --> 00:03:28.599 and then you have other kind of gener of AI 55 00:03:28.840 --> 00:03:32.960 like diffusion models, which you might see something like mid 56 00:03:33.039 --> 00:03:38.840 journey type kind of use there, and other kinds like 57 00:03:39.039 --> 00:03:39.960 gams and. 58 00:03:39.919 --> 00:03:40.960 Type so. 59 00:03:42.439 --> 00:03:44.240 Kind of those are kind of some of the more 60 00:03:44.439 --> 00:03:48.240 popular ones. And so it's kind of interesting because when 61 00:03:48.280 --> 00:03:52.680 I first started getting into AI with llms, with like 62 00:03:52.759 --> 00:03:57.039 CHATBT three point five, it's one of the things I 63 00:03:57.080 --> 00:03:58.879 kind of thought about using it for is kind of 64 00:03:58.879 --> 00:04:03.560 like a a universal API where you can ask it 65 00:04:04.039 --> 00:04:06.800 to do anything and then have it basically come up 66 00:04:06.840 --> 00:04:10.080 with the answer. Kind of find out pretty quickly that 67 00:04:10.080 --> 00:04:14.280 that's not really a general little use case. It can't 68 00:04:14.319 --> 00:04:19.399 do everything. It has advantages and disadvantages. It's kind of 69 00:04:19.439 --> 00:04:21.319 interesting A lot of people. There are a lot of 70 00:04:21.560 --> 00:04:24.639 businesses that are kind of just a wrapper over over 71 00:04:24.720 --> 00:04:28.920 top of something open AI and so which a lot 72 00:04:28.959 --> 00:04:33.920 of times you can just use a chat like app 73 00:04:34.439 --> 00:04:36.879 and really get a lot of the use out of 74 00:04:36.920 --> 00:04:39.720 it that way. Some of the advantages of some of 75 00:04:39.720 --> 00:04:42.399 those kind of applications are they put a lot of 76 00:04:42.560 --> 00:04:47.639 thought and effort into the prompt engineering that's basically a 77 00:04:47.680 --> 00:04:52.639 way to explain to an AI like how it should 78 00:04:52.680 --> 00:04:56.560 behave and what rules it should follow, and so that 79 00:04:56.600 --> 00:05:00.319 there's a lot of utility out of that. But I 80 00:05:00.319 --> 00:05:04.839 think one of the big things that we're seeing a 81 00:05:04.879 --> 00:05:09.240 lot of move towards is AI agentic workflows. And there's 82 00:05:09.279 --> 00:05:11.240 different kind of names with that. You might hear like 83 00:05:11.519 --> 00:05:17.199 AI swarms, like agent swarms or agent teams, and where 84 00:05:17.240 --> 00:05:21.759 the huge benefit from that is instead of using like 85 00:05:21.879 --> 00:05:25.759 one agent, like if you have an AI agent. There's 86 00:05:25.759 --> 00:05:28.040 a lot of different definitions for that, but one way I 87 00:05:28.079 --> 00:05:32.360 like to look at it is is an agent is 88 00:05:32.399 --> 00:05:34.519 something that you can give a prompt to and you 89 00:05:34.560 --> 00:05:37.720 can get some kind of expected response from. And so 90 00:05:39.120 --> 00:05:41.399 if you're using something like CHATBT, you can give it 91 00:05:41.439 --> 00:05:44.680 like a task or asking a question and it will 92 00:05:44.680 --> 00:05:49.560 give you a response that's like one agent and it's 93 00:05:49.600 --> 00:05:54.319 more could be more general general. An agent swarm or 94 00:05:54.439 --> 00:05:57.279 team or like a workflow is a way to be 95 00:05:57.279 --> 00:06:02.040 able to have multiple different agents work together to solve 96 00:06:02.759 --> 00:06:07.439 a problem or perform some kind of task. And so 97 00:06:08.160 --> 00:06:10.160 there's a lot of advantage of that, just like having 98 00:06:10.439 --> 00:06:14.399 maybe one person in it for company that has to 99 00:06:14.439 --> 00:06:17.120 wear all the different hats, versus having a team of 100 00:06:17.120 --> 00:06:20.319 people working in a company to perform some kind of task. 101 00:06:20.720 --> 00:06:24.040 So you have different people that are specialized at different 102 00:06:24.120 --> 00:06:28.160 kind of roles, and each one of them does really 103 00:06:28.199 --> 00:06:30.720 well and what they do, and it comes together as 104 00:06:30.720 --> 00:06:34.639 a collective to provide the most value. You can do 105 00:06:34.680 --> 00:06:38.279 the same thing with a AI agent swarms or teams 106 00:06:38.360 --> 00:06:42.319 or workflows, whatever you want to call that. An example 107 00:06:42.360 --> 00:06:45.839 of that would be a project I used with WOWS 108 00:06:46.240 --> 00:06:50.399 created completely within WOWS is something called speak Magic, and 109 00:06:50.439 --> 00:06:55.000 you can check out the examples of that Speakmagic dot AI. 110 00:06:55.839 --> 00:06:58.639 And the idea is you can give one prompt and 111 00:06:58.680 --> 00:07:03.240 then have an agent swarm of like forty two different 112 00:07:03.279 --> 00:07:08.000 agents that basically take a story input, like you just 113 00:07:08.040 --> 00:07:10.720 give an idea. Maybe you have ideas for a character 114 00:07:10.839 --> 00:07:13.319 or something like that, and it'll create up to a 115 00:07:13.319 --> 00:07:18.959 two minute video of these different agents working together to 116 00:07:19.079 --> 00:07:24.000 create like a scene. And so it'll take the story prompt, 117 00:07:24.319 --> 00:07:28.040 it turns that story prompt into scenes. It's turns of 118 00:07:28.120 --> 00:07:32.600 scenes into scripts. It turns those scripts into shots, and 119 00:07:32.639 --> 00:07:35.240 then you have like different characters that are basically acting 120 00:07:35.240 --> 00:07:38.079 out the scene, and animates the characters like speaking to 121 00:07:38.120 --> 00:07:41.160 each other, and they could add like sound effects and 122 00:07:41.279 --> 00:07:44.360 music and different things like that to create the create it. 123 00:07:45.160 --> 00:07:50.040 So there's a lot of advantages of that. And so 124 00:07:50.199 --> 00:07:52.079 it's kind of like where I see things are going 125 00:07:53.120 --> 00:07:53.800 kind of the state of. 126 00:07:53.879 --> 00:07:58.560 Yeah, yeah, it's it's interesting because I talk to different 127 00:07:58.560 --> 00:08:01.680 people and it seems like everybody doing something different with AI. 128 00:08:01.839 --> 00:08:05.000 So you've got kind of like you were saying, where 129 00:08:05.000 --> 00:08:08.839 you've got people who are you know, they're trying to 130 00:08:08.839 --> 00:08:10.759 make a story and then they're maybe they're making it 131 00:08:10.800 --> 00:08:13.519 into a story with the video. Right, most of the 132 00:08:13.560 --> 00:08:16.639 people that I've been talking to and working with, they're 133 00:08:16.720 --> 00:08:20.040 using it kind of you were talking about agentic AI 134 00:08:20.240 --> 00:08:22.279 and the you know, kind of building out your team 135 00:08:22.319 --> 00:08:24.680 to do different things, and that's kind of where my 136 00:08:24.720 --> 00:08:26.800 interest has been, except it's more of a chat agent 137 00:08:26.879 --> 00:08:30.800 and less of a voice or a video agent. And 138 00:08:31.360 --> 00:08:34.120 so yeah, you give it a prompt and then you 139 00:08:34.240 --> 00:08:36.200 also give it a set of functions that allow it 140 00:08:36.200 --> 00:08:38.200 to do things, and then it can go off and 141 00:08:38.200 --> 00:08:42.639 it can do the things. But what I'm finding most 142 00:08:42.639 --> 00:08:44.639 people do with that is, like you said, they have 143 00:08:44.679 --> 00:08:47.279 a team of agents. And so you may have an 144 00:08:47.320 --> 00:08:50.559 agent that is kind of the coordinator or support agent, 145 00:08:51.279 --> 00:08:53.440 and then you know, it can go and talk to 146 00:08:53.480 --> 00:08:55.799 the scheduler agent to get stuff on the calendar, and 147 00:08:55.960 --> 00:08:59.960 go talk to the technical agent to you know, get 148 00:09:00.120 --> 00:09:02.840 more specialized technical feedback, or it can go talk to 149 00:09:03.559 --> 00:09:06.440 you know, another agent that has access to different other 150 00:09:06.480 --> 00:09:09.480 APIs to do other things, and and so that that's 151 00:09:09.559 --> 00:09:11.159 kind of the deal. And so then you get into 152 00:09:11.240 --> 00:09:15.279 using something like model fusion for JavaScript or I've been 153 00:09:15.320 --> 00:09:18.000 doing a lot with ray r ai X and Ruby 154 00:09:18.960 --> 00:09:20.879 to do a lot of this stuff. And you just 155 00:09:21.000 --> 00:09:23.360 you write a tool, and a tool is essentially that 156 00:09:23.440 --> 00:09:25.960 set of functions and and you make stuff run. And 157 00:09:26.000 --> 00:09:28.919 so yeah, one of the tools might be here's the 158 00:09:29.039 --> 00:09:33.320 video generation API, and you know it also uses a 159 00:09:33.320 --> 00:09:35.879 AI to do its work. And so anyway, it's it's 160 00:09:35.960 --> 00:09:39.360 really fascinating kind of see where it's all going. At 161 00:09:39.399 --> 00:09:43.840 the last code meet up, I went to one of 162 00:09:43.840 --> 00:09:47.759 the guys there had actually been using it to write fiction, 163 00:09:48.080 --> 00:09:50.720 and so you know, he'd use it to flesh out 164 00:09:50.759 --> 00:09:53.559 parts of the story or actually write parts of the story, 165 00:09:53.600 --> 00:09:56.120 and you know, he's like, yeah, sometimes it's really good 166 00:09:56.120 --> 00:09:59.440 and sometimes it's really not. It's been, it's been. It's 167 00:09:59.480 --> 00:10:03.879 interesting to to see how far it goes. What I 168 00:10:04.039 --> 00:10:07.440 tend to find is that if you do a lot 169 00:10:07.519 --> 00:10:10.639 of people want like the one off prompt where you 170 00:10:10.759 --> 00:10:13.399 just write the prompt and you immediately get back the 171 00:10:13.399 --> 00:10:15.840 feedback that you want. What I found is that in 172 00:10:15.879 --> 00:10:18.200 a lot of cases you have to refine it with 173 00:10:18.240 --> 00:10:19.840 the AI see what you do wind up doing as 174 00:10:19.879 --> 00:10:23.240 you wind up saying and I forgot to tell you this, 175 00:10:23.559 --> 00:10:26.080 or oh you know, I need you know, I need 176 00:10:26.080 --> 00:10:29.519 this scheduled, you know, every Thursday, and then it's okay. Well, 177 00:10:30.200 --> 00:10:32.320 when I said every Thursday, I actually meant, you know, 178 00:10:32.759 --> 00:10:35.759 except holidays and this and that and the other, you know, 179 00:10:36.000 --> 00:10:37.759 or if I'm trying to get it to write code. 180 00:10:37.799 --> 00:10:43.600 It's so one example of this I was using lately. 181 00:10:43.600 --> 00:10:45.200 I've been playing with rock but this one was on 182 00:10:45.279 --> 00:10:48.600 chat GPT and it you know, I said, hey, I 183 00:10:48.639 --> 00:10:50.679 need an audio player for my website. By the way, 184 00:10:50.759 --> 00:10:52.679 if you go to topendevs dot com, if you go 185 00:10:52.720 --> 00:10:55.360 to JavaScript jabber dot com and you the player at 186 00:10:55.360 --> 00:10:59.120 the bottom of the page is the player that Ai 187 00:10:59.159 --> 00:11:02.279 mostly wrote. And I said, hey, I need an audio 188 00:11:02.279 --> 00:11:04.919 player for the website. You know, I need this, I 189 00:11:04.960 --> 00:11:07.039 need that, I need this, right, And so it's like 190 00:11:07.080 --> 00:11:09.000 I want a volume, you know, I want to be 191 00:11:09.039 --> 00:11:11.399 able to change the volume, and I want the progress 192 00:11:11.480 --> 00:11:14.759 bar to go, and I want it to there's a 193 00:11:14.799 --> 00:11:16.480 state of the there's a state of the art thing 194 00:11:16.519 --> 00:11:20.399 for podcasts where you tell it to not load the 195 00:11:20.440 --> 00:11:23.399 audio until it's actually clicked. So until you hit play 196 00:11:23.960 --> 00:11:28.320 or download right, it doesn't it doesn't eagerload it. And 197 00:11:28.360 --> 00:11:33.879 the reason is is because the the metric DuJour for 198 00:11:33.960 --> 00:11:37.279 a long time with podcast is downloads, and so you 199 00:11:37.320 --> 00:11:40.519 can actually pad your numbers by not telling it or 200 00:11:40.559 --> 00:11:43.080 by by forgetting to tell it not to download every 201 00:11:43.080 --> 00:11:48.399 time it loads the page, right and so right, But anyway, 202 00:11:48.440 --> 00:11:50.600 so I told it, and then I actually asked it 203 00:11:50.679 --> 00:11:53.879 what other features should I put in? Right? And so anyway, 204 00:11:53.879 --> 00:11:55.440 we're seeing this kind of thing with a lot of 205 00:11:55.480 --> 00:11:57.480 different people. So some people are willing to go in 206 00:11:57.519 --> 00:12:01.679 and use the web interface on something like this or there, 207 00:12:01.799 --> 00:12:05.360 what is it? Open web. There's a there's a web 208 00:12:05.360 --> 00:12:07.480 interface that you can run on Olama on your own 209 00:12:07.519 --> 00:12:10.919 machine to do a lot of these things. And so anyway, 210 00:12:10.919 --> 00:12:12.840 there are a lot of options for this stuff, and 211 00:12:13.399 --> 00:12:15.679 some people have refined it so that it will automatically 212 00:12:15.799 --> 00:12:19.120 use tools to go and make web searches and stuff 213 00:12:19.159 --> 00:12:21.759 like that. Or if you use grock will it'll tell 214 00:12:21.799 --> 00:12:24.080 you that it's thinking, and it'll you can see that 215 00:12:24.080 --> 00:12:26.080 it's loading in different pages and you can ask it 216 00:12:26.120 --> 00:12:30.120 for its sources. But yeah, it's it's been fascinating to 217 00:12:30.200 --> 00:12:32.679 just see where all of this goes. And then of 218 00:12:32.679 --> 00:12:35.679 course they're the specialized uses like Cursor, AI and things 219 00:12:35.759 --> 00:12:39.039 like that for for programmers or you know, other AI 220 00:12:39.240 --> 00:12:42.919 systems for other folks, and so yeah, this guy's kind 221 00:12:42.919 --> 00:12:44.159 of the limit for it. I think. 222 00:12:45.480 --> 00:12:50.000 Yeah, it's such a broad range of different things that 223 00:12:50.080 --> 00:12:54.879 AI can do. We're talking about using things like Cursor. 224 00:12:55.720 --> 00:12:59.120 There's also other tools that are more like coding agents, 225 00:12:59.240 --> 00:13:01.480 so like cloud code. I don't know if you've seen 226 00:13:01.759 --> 00:13:04.960 Mannish kind of hit seen it a week or two ago. 227 00:13:06.240 --> 00:13:11.159 That one's been pretty pretty amazing what it can accomplish. 228 00:13:11.240 --> 00:13:15.320 But just using I typically most of my coding I 229 00:13:15.440 --> 00:13:18.279 use is I just open up a chat and I 230 00:13:18.399 --> 00:13:23.759 kind of explain a specific feature that I want and 231 00:13:23.799 --> 00:13:27.840 then kind of treat it like a junior dev in 232 00:13:27.879 --> 00:13:28.919 a way. I think that's kind of one of the 233 00:13:28.919 --> 00:13:32.600 best ways to treat it. Kind of give it all 234 00:13:32.639 --> 00:13:36.639 the different use cases, what you're what I expect going 235 00:13:36.679 --> 00:13:40.080 into it, and what should be coming out of the code, 236 00:13:41.519 --> 00:13:46.679 and and why and so, and then of course you 237 00:13:46.720 --> 00:13:49.720 need to check the code make sure that it's going 238 00:13:49.799 --> 00:13:52.720 to do what it wants. It's kind of interesting because 239 00:13:52.759 --> 00:13:54.799 we'll see I think we're going to see more and 240 00:13:54.799 --> 00:13:58.120 more where AI can do more and more coding. Some 241 00:13:58.159 --> 00:14:03.879 of these things like cursor are getting better, but typically 242 00:14:04.120 --> 00:14:08.360 the larger of your code base, more complicated get things, 243 00:14:08.360 --> 00:14:12.799 get it kind of runs out of context window and 244 00:14:13.200 --> 00:14:17.240 right starts kind of not performing quite as well. So 245 00:14:17.519 --> 00:14:23.120 usually smaller projects are really great. And that's changing though, 246 00:14:23.399 --> 00:14:26.320 so which is well, been all fun, you. 247 00:14:26.240 --> 00:14:28.399 Know, it's interesting. I was listening to a podcast this morning. 248 00:14:28.519 --> 00:14:31.639 They were talking about the topic of vibe coding, which 249 00:14:31.679 --> 00:14:34.639 is the idea that using stuff like yeah and so 250 00:14:35.759 --> 00:14:39.559 it was syntax FM and so what they are The 251 00:14:40.000 --> 00:14:42.919 general definition is someone who wants some simple app for 252 00:14:43.000 --> 00:14:45.879 some business purpose or the examples were given like I'm 253 00:14:46.480 --> 00:14:48.679 trying to build a little game with my kid and 254 00:14:48.960 --> 00:14:50.840 doing it the regular way takes forever in a day, 255 00:14:50.879 --> 00:14:53.360 and or a way to do a quick demo of 256 00:14:53.399 --> 00:14:55.600 something to see how something would look or things like that, 257 00:14:55.639 --> 00:14:57.440 and then you're not it's not something you're going to 258 00:14:57.480 --> 00:15:02.159 be using, you know, long term professionally, you know, deploy 259 00:15:02.240 --> 00:15:04.919 it and reuse as sort of a one off type 260 00:15:04.919 --> 00:15:08.639 of thing, and so, you know, usually with this kind 261 00:15:08.679 --> 00:15:12.919 of stuff, the code quality, from what I've understood, is 262 00:15:13.000 --> 00:15:17.120 how do we say less than desirable? But the idea 263 00:15:17.159 --> 00:15:19.159 is that you can spin up something really quick, whether 264 00:15:19.200 --> 00:15:21.279 it's just you know, to see how something would look, 265 00:15:21.879 --> 00:15:25.120 or to give you an idea of, uh, you know, 266 00:15:25.399 --> 00:15:27.159 how something could work, and then you could you know, 267 00:15:27.159 --> 00:15:29.399 tweak it from there or do something from there. But yeah, 268 00:15:29.879 --> 00:15:32.440 like I said, vibe coding is apparently one of the 269 00:15:32.840 --> 00:15:34.879 new term all the cool kids are using. 270 00:15:36.960 --> 00:15:39.000 Yeah. I'm not sure where the line is though, between 271 00:15:39.080 --> 00:15:42.559 vibe coding and actually just having the AI help you right, right, 272 00:15:42.559 --> 00:15:44.879 Because Yeah, at the end of the day, like in 273 00:15:44.960 --> 00:15:47.679 my example with the player, right, I wound up having 274 00:15:47.720 --> 00:15:50.759 to it it Probably seventy five percent of the code 275 00:15:50.759 --> 00:15:53.919 that the AI wrote I used, But the rest of it, 276 00:15:54.000 --> 00:15:56.320 I mean I had to use my own expertise to 277 00:15:56.360 --> 00:15:57.919 make it do what it do and look how I 278 00:15:57.960 --> 00:16:00.480 wanted it to, right, it didn't give me exactly what 279 00:16:00.519 --> 00:16:02.840 I wanted. So yeah, I don't know. 280 00:16:02.960 --> 00:16:05.320 So Matthew, let me ask you this question. You know, 281 00:16:05.360 --> 00:16:09.639 when it comes to this auto this uh auto generation coding. 282 00:16:10.159 --> 00:16:12.159 My question that I've always had that I've never had 283 00:16:12.159 --> 00:16:14.279 answered that I've never because I've never looked into it 284 00:16:14.320 --> 00:16:16.519 is what languages and tools are being used? I mean, 285 00:16:16.559 --> 00:16:20.200 how do you determine that? Like, you know, my focus 286 00:16:20.200 --> 00:16:22.799 of development, I tend to focus with letterbl and view 287 00:16:22.879 --> 00:16:24.799 and tailwind and inertia. 288 00:16:24.440 --> 00:16:26.000 And some other tools like that. 289 00:16:26.960 --> 00:16:30.440 And so if I'm going to tell AI to build me, 290 00:16:30.919 --> 00:16:33.240 you know, some little to do app, you know, to 291 00:16:33.279 --> 00:16:35.679 beat that one to death or even something more complex, 292 00:16:36.240 --> 00:16:37.759 do I say, Okay, I want you to do it 293 00:16:37.799 --> 00:16:39.879 with this framework and these tools in this language, or 294 00:16:39.879 --> 00:16:41.799 do you just say build me an app, and it 295 00:16:41.840 --> 00:16:44.639 builds one for you out of what it determines to 296 00:16:44.679 --> 00:16:46.679 be best. How does it determine that? I mean, what's 297 00:16:47.360 --> 00:16:50.120 what is the structure that is used to generate uh, 298 00:16:50.200 --> 00:16:52.480 these apps that AI is building for you. 299 00:16:53.480 --> 00:16:57.080 Sure, So I think there's a lot of different ways 300 00:16:57.080 --> 00:16:59.279 to look at this. One is I mean, if your 301 00:16:59.320 --> 00:17:02.320 code of yourself, I mean, you're gonna want to stick 302 00:17:02.360 --> 00:17:04.480 with the things that you know to some degree so 303 00:17:04.519 --> 00:17:08.640 that you can you can edit and and kind of 304 00:17:08.720 --> 00:17:11.079 understand why why things are going wrong when they do 305 00:17:11.160 --> 00:17:14.200 go wrong, so that you can fix things. And so 306 00:17:15.000 --> 00:17:19.200 typically a lot of things when it comes to programming 307 00:17:19.279 --> 00:17:23.319 with like using AI, like models and tools and things. 308 00:17:23.640 --> 00:17:25.119 There's a lot of things that are geared a little 309 00:17:25.160 --> 00:17:30.680 bit more towards Python. It's kind of the whole, the 310 00:17:30.720 --> 00:17:34.079 whole kind of with with things, and so you have 311 00:17:34.400 --> 00:17:36.000 a lot of tools and stuff that are kind of 312 00:17:36.000 --> 00:17:40.000 more geared towards Python. But if you're having it generate code, 313 00:17:41.400 --> 00:17:45.160 typically how I mean how lms work As you're pulling 314 00:17:45.200 --> 00:17:48.640 information from a whole bunch of data from like the Internet, 315 00:17:48.839 --> 00:17:53.039 and so the more the more data there is for 316 00:17:53.079 --> 00:17:57.160 a particular technology, the more likely it can perform a 317 00:17:57.200 --> 00:18:02.319 bit better. And so things like having like JavaScript one 318 00:18:02.359 --> 00:18:04.799 of the nice things a lot of times when I code, 319 00:18:05.079 --> 00:18:08.440 and having like AI help out with the code using 320 00:18:08.480 --> 00:18:14.480 something like Chatchyp's artifact or canvas and being able to 321 00:18:14.680 --> 00:18:16.799 have it generate code, but it can actually render that 322 00:18:16.839 --> 00:18:19.000 code for you, which I don't think it can really 323 00:18:19.039 --> 00:18:23.079 render anything else besides really JavaScript right now. And so 324 00:18:23.440 --> 00:18:26.559 that's really nice to build you something like that in 325 00:18:26.599 --> 00:18:29.279 that kind of case. But when it comes to the 326 00:18:29.279 --> 00:18:35.400 different technologies, I mean, it really comes down to I mean, 327 00:18:36.000 --> 00:18:38.440 I liked sticking with the kind of the languages that 328 00:18:38.599 --> 00:18:43.880 I use, React node in the back end, and so 329 00:18:45.119 --> 00:18:48.359 using kind of the technologies that you can help help 330 00:18:48.440 --> 00:18:50.559 steer it in the right way or connect it to 331 00:18:50.759 --> 00:18:53.079 your existing app, I think is kind of the more 332 00:18:53.119 --> 00:18:56.480 the way they go. Can I see right now, if 333 00:18:56.480 --> 00:18:59.599 you're going to go more obscure kind of languages of technologies, 334 00:19:00.400 --> 00:19:03.960 you're you're more likely to not get as much support 335 00:19:04.079 --> 00:19:10.440 from from AI helping out. One way around that is 336 00:19:10.440 --> 00:19:14.599 is getting the context for those technologies, like say if 337 00:19:14.759 --> 00:19:18.319 an ELM just doesn't really have much UH knowledge based 338 00:19:18.359 --> 00:19:21.839 on that, go into like the documentation for that and 339 00:19:21.880 --> 00:19:25.079 like giving that documentation to the relevant pieces of the 340 00:19:25.079 --> 00:19:30.720 documentation to t A I to help you with that. 341 00:19:30.920 --> 00:19:33.599 So it's kind of a way to ram as well. 342 00:19:35.039 --> 00:19:37.920 So one one thing with that is because I haven't 343 00:19:37.960 --> 00:19:40.000 really run into that. You know, most languages I use 344 00:19:40.039 --> 00:19:44.799 are fairly mainstream. I think Steve's in the same boat 345 00:19:44.839 --> 00:19:48.920 with PHP and JavaScript. But if you were using an 346 00:19:48.960 --> 00:19:53.119 excure obscure language, so maybe Elixir or I'm just trying 347 00:19:53.119 --> 00:19:55.559 to think, like, how how far obscure can you get self? 348 00:19:55.720 --> 00:20:01.480 I don't know ELM, ELM there might or might not. 349 00:20:01.839 --> 00:20:04.039 I guess it probably depends. That's one other thing that 350 00:20:04.079 --> 00:20:07.000 I found is that some models do much better with 351 00:20:07.039 --> 00:20:12.200 certain languages than others. But do you run the risk 352 00:20:12.279 --> 00:20:14.519 of so just to give people a little bit of context. 353 00:20:14.519 --> 00:20:18.559 So there are different sets of data that your LLM 354 00:20:18.640 --> 00:20:21.359 works from. So there's the latent space and that's everything 355 00:20:21.400 --> 00:20:23.640 that it got trained on, right, And you see them 356 00:20:23.640 --> 00:20:26.839 building these big data centers with all the GPUs, and 357 00:20:26.880 --> 00:20:29.119 they're trying to get as much data in as possible, right, 358 00:20:29.160 --> 00:20:30.880 So they come out with a new model, and generally 359 00:20:30.880 --> 00:20:34.400 what it means is that, hey, we've jammed more data 360 00:20:34.440 --> 00:20:36.400 into it. It's more training, and so you're going to 361 00:20:36.400 --> 00:20:39.960 get better answers from it. And then you've got the context, 362 00:20:39.960 --> 00:20:43.880 which is hey, for this particular query or set of 363 00:20:43.960 --> 00:20:47.880 queries that I make to the LM, there are prompts. 364 00:20:48.319 --> 00:20:50.400 You know a lot of times it's part of the prompt, 365 00:20:50.400 --> 00:20:53.759 but doesn't always have to be. You know, here's everything 366 00:20:53.799 --> 00:20:55.160 you need to know in order to give me a 367 00:20:55.160 --> 00:20:58.759 good answer. Right, So you're counting on its ability to process, 368 00:20:58.839 --> 00:21:00.759 plus whatever it's got in the lay and space that 369 00:21:00.839 --> 00:21:03.640 it already knows. Right. So it may already know stuff 370 00:21:03.680 --> 00:21:07.319 about programming in general, programming practices in the latent space 371 00:21:07.359 --> 00:21:09.519 because it's already been trained there. But then here are 372 00:21:09.519 --> 00:21:12.599 the specifics of the language that I'm using, right, Or 373 00:21:12.640 --> 00:21:14.519 here are the functions that are available to me, and 374 00:21:14.559 --> 00:21:17.440 here's how they work. Right. So maybe it's not the language. 375 00:21:17.480 --> 00:21:19.759 Maybe it's an obscure library that it just doesn't know 376 00:21:19.799 --> 00:21:23.000 as much about. Do you run the risk though, of 377 00:21:23.039 --> 00:21:26.319 making that or of going beyond what the context window 378 00:21:26.319 --> 00:21:30.440 will hold? Because the context window is essentially how much 379 00:21:30.480 --> 00:21:34.279 you can tell the LM right when you ask the question. 380 00:21:34.400 --> 00:21:37.000 It's it's all the data can hold. And so for 381 00:21:37.400 --> 00:21:39.839 the layperson. I'm not explaining it to Matthew. Matthew knows 382 00:21:39.839 --> 00:21:42.640 this stuff. But do you run the risk then of 383 00:21:43.880 --> 00:21:46.759 expanding beyond what the what the context will hold, because 384 00:21:47.799 --> 00:21:49.759 you know different models will allow you to give them 385 00:21:49.799 --> 00:21:52.559 different amounts of data in your context. 386 00:21:53.839 --> 00:21:56.680 Yeah, so a lot of models when you go beyond 387 00:21:57.000 --> 00:22:01.759 the context window, then they just fail and so it'll 388 00:22:01.839 --> 00:22:04.400 just give you an error and that could be very 389 00:22:04.440 --> 00:22:10.880 annoying using some different things. The other there's two other 390 00:22:11.200 --> 00:22:15.519 issues with main issues with larger context as well. One 391 00:22:15.599 --> 00:22:20.519 is the cost. So as your input context goes in 392 00:22:20.599 --> 00:22:22.599 like what you're supplying to it, which could be like 393 00:22:22.640 --> 00:22:27.119 your whole code base, then the cost for each token 394 00:22:27.400 --> 00:22:32.200 generation and it all goes up. And then the other 395 00:22:32.279 --> 00:22:36.880 problem too is as the context windows starts filling up, 396 00:22:37.200 --> 00:22:39.880 the less likely it's going to be able to recall 397 00:22:40.519 --> 00:22:43.880 and to be able to per form as well. And 398 00:22:43.920 --> 00:22:48.119 so that's where when you're making AI workflows kind of 399 00:22:48.119 --> 00:22:50.599 one of the parts is like from the whole solid concept, 400 00:22:50.680 --> 00:22:54.960 you have the single responsibility and I kind of apply 401 00:22:55.079 --> 00:22:58.640 that within each agent within it. And the idea is 402 00:22:58.880 --> 00:23:03.559 each agent that you create needs to be specialized, and 403 00:23:03.599 --> 00:23:06.200 I like to keep the context window around around four 404 00:23:06.200 --> 00:23:11.160 thousand tokens, which is like basically like four thousand words, 405 00:23:12.440 --> 00:23:17.880 but not not quite. And it seems to perform best 406 00:23:17.880 --> 00:23:20.759 in recall as well as be able to fall instructions 407 00:23:21.240 --> 00:23:24.519 with having a smaller context. And so that's one of 408 00:23:24.599 --> 00:23:27.359 the reasons why I like using just like a traditional 409 00:23:27.400 --> 00:23:30.039 just chatbot and having it generate code, is because I 410 00:23:30.079 --> 00:23:33.359 can limit the context exactly what I want, and I 411 00:23:33.440 --> 00:23:35.759 kind of pull and say, okay, I need this new 412 00:23:35.759 --> 00:23:38.400 file that does this thing. These are the inputs coming in. 413 00:23:38.480 --> 00:23:42.640 These are the outputs going out. Maybe feed it like 414 00:23:42.799 --> 00:23:46.000 a documentation for something you'll be using, and trying to 415 00:23:46.039 --> 00:23:49.200 keep that as small as possible because it'll just perform. 416 00:23:48.880 --> 00:23:49.480 There that way. 417 00:23:50.240 --> 00:23:53.000 So that's kind of how I see things. 418 00:23:54.759 --> 00:23:59.400 Yeah, it's it's it's kind of an interesting dilemma, but yeah, 419 00:23:59.440 --> 00:24:02.519 I like the I like the approach that you're advocating 420 00:24:02.559 --> 00:24:04.960 where you, yeah, you break it up and specialize it 421 00:24:05.400 --> 00:24:08.400 because then nothing has to hold too much context, right, 422 00:24:08.960 --> 00:24:11.599 all all your you know, your different things have to 423 00:24:11.640 --> 00:24:13.720 know about the things the other ones do. Is how 424 00:24:13.720 --> 00:24:16.000 to tell the other one to do it or how 425 00:24:16.000 --> 00:24:18.839 to get the data back once it's done, and so 426 00:24:19.400 --> 00:24:26.960 that that keeps your context smaller. Yeah, okay, ahead, So 427 00:24:26.960 --> 00:24:28.880 so one other thing that I'm curious about, because you're 428 00:24:28.920 --> 00:24:31.640 building a tool that allows people to build these workflows out. 429 00:24:32.240 --> 00:24:35.319 So what's that like as far as you know, not 430 00:24:35.599 --> 00:24:38.920 necessarily doing the prompts or putting the prompts together, and 431 00:24:39.079 --> 00:24:40.880 maybe you are, maybe you are doing that as part 432 00:24:40.880 --> 00:24:43.319 of your tool, But what's it like building an AI 433 00:24:43.400 --> 00:24:45.680 tool for other people to use that allows them to 434 00:24:45.720 --> 00:24:46.920 build these workflows? 435 00:24:47.680 --> 00:24:52.920 Yeah, within using WOWS. It's a it's kind of interesting, 436 00:24:53.000 --> 00:24:56.559 so in the sense that it's a new paradigm of programming, 437 00:24:56.880 --> 00:25:00.960 so kind of kind of bringing what the value of 438 00:25:01.319 --> 00:25:03.920 working with AI and AI workflows. I think we kind 439 00:25:03.920 --> 00:25:07.079 of go back a little bit kind of human evolution. So, 440 00:25:08.759 --> 00:25:11.160 I mean we started with language, which was a way 441 00:25:11.200 --> 00:25:15.880 to be able to pass on like knowledge to other people, right, 442 00:25:16.400 --> 00:25:18.880 and then the concept to be on having a written 443 00:25:18.960 --> 00:25:22.759 language where you're actually able to write down some knowledge 444 00:25:22.839 --> 00:25:27.759 that can be passed on really changed the way we 445 00:25:27.960 --> 00:25:30.400 learned and grew. Then you had like the printing press 446 00:25:30.440 --> 00:25:33.519 that allows it to be distributed really quickly, and that 447 00:25:33.759 --> 00:25:38.359 really changed humanity and that kind of sense too. Then 448 00:25:38.720 --> 00:25:41.599 kind of one of the next iterational steps I kind 449 00:25:41.599 --> 00:25:47.079 of see is is just traditional programming, where where now 450 00:25:47.240 --> 00:25:49.160 like with written language, was a way to pass on 451 00:25:49.519 --> 00:25:55.160 and to distribute knowledge, programming allowed us to capture and 452 00:25:55.200 --> 00:25:59.480 to pass on intelligence, and so so I could take 453 00:25:59.519 --> 00:26:03.119 that knowledge, I could solve a problem, create an algorithm, 454 00:26:03.559 --> 00:26:07.839 and then distribute that out through like a disc type 455 00:26:07.839 --> 00:26:09.519 of thing to other people. And now we have the 456 00:26:09.519 --> 00:26:12.480 Internet to be able to pass that on. And that 457 00:26:12.640 --> 00:26:16.240 was a way to be able to create and to 458 00:26:16.480 --> 00:26:19.799 distribute intelligence in that kind of sense that someone didn't. 459 00:26:19.960 --> 00:26:22.960 Someone could actually just reuse that algorithm to get an 460 00:26:23.000 --> 00:26:26.759 answer or solve a problem pretty quickly. I kind of 461 00:26:26.839 --> 00:26:29.920 look at working with AI workflows is kind of like 462 00:26:30.000 --> 00:26:35.480 another iteration where you're actually building and being able to 463 00:26:35.480 --> 00:26:39.200 distribute wisdom in a sense, so you can create these 464 00:26:39.240 --> 00:26:45.440 agents that can take a problem and have that context 465 00:26:45.440 --> 00:26:49.599 and understand how to apply different kind of intelligence to 466 00:26:49.720 --> 00:26:54.759 solve issues. And so one of the great things in 467 00:26:54.799 --> 00:26:57.640 that kind of sense is one of the things that 468 00:26:57.720 --> 00:27:01.599 AI is really good at is being able to understand 469 00:27:02.200 --> 00:27:06.119 like natural language and understand the intention of what someone 470 00:27:06.599 --> 00:27:10.920 is trying to accomplish and be able to then create 471 00:27:11.799 --> 00:27:14.640 many different ways of like a response. So one could 472 00:27:14.640 --> 00:27:20.920 be natural langages coming back and like specifying exactly how 473 00:27:21.519 --> 00:27:24.440 that solution could be a natural language, or could be 474 00:27:24.720 --> 00:27:29.640 calling a tool to perform some kind of tasks as well. 475 00:27:30.119 --> 00:27:33.279 So in working with AI workflows, the way I kind 476 00:27:33.279 --> 00:27:35.119 of look at it is almost like you're building like 477 00:27:36.119 --> 00:27:39.839 a team of people that would accomplish a task, and 478 00:27:39.880 --> 00:27:46.000 so you divide out the responsibilities between different agents that 479 00:27:46.079 --> 00:27:49.319 will accomplish the task. So a lot of times, like 480 00:27:50.079 --> 00:27:52.759 the first agent that you might make might be something 481 00:27:52.799 --> 00:27:56.440 called like an orchestration agent or a conductor agent, which 482 00:27:56.480 --> 00:28:01.079 basically will take it will take the prompt from the 483 00:28:01.160 --> 00:28:05.200 end user and say, okay, what team of agents should 484 00:28:05.240 --> 00:28:09.119 solve this problem. So if it's just a question that 485 00:28:09.480 --> 00:28:12.000 about the software that you're they're using, then you might 486 00:28:12.079 --> 00:28:14.599 just send it to this knowledge based agent that basically 487 00:28:15.160 --> 00:28:18.759 can give an answer really quickly. Or it might be 488 00:28:18.839 --> 00:28:23.079 like a task. So in the sense of speak Magic AI, 489 00:28:24.799 --> 00:28:27.039 you give like a story prompt and then it says, okay, 490 00:28:27.480 --> 00:28:29.160 all right, this is the very first prompt, So we 491 00:28:29.160 --> 00:28:31.440 need to create a story and we need to divide 492 00:28:31.440 --> 00:28:33.319 that up, so it sends it to this list of 493 00:28:33.359 --> 00:28:37.200 agents over here that then sequentially go through and work 494 00:28:37.240 --> 00:28:42.000 together to build a video basically. And that's one of 495 00:28:42.039 --> 00:28:45.720 the kind of different paradigm shifts as well as I 496 00:28:45.720 --> 00:28:49.880 think we'll see a lot more of is is. Even 497 00:28:49.920 --> 00:28:52.039 though a UY is very helpful and I think we'll 498 00:28:52.039 --> 00:28:55.240 always have a UI for like websites and things like that, 499 00:28:56.680 --> 00:28:59.960 there's a new kind of wave of input of natural 500 00:29:00.160 --> 00:29:02.359 language that we can start working with AI. So it's 501 00:29:02.359 --> 00:29:05.880 almost like having instead of using just a software like 502 00:29:06.000 --> 00:29:09.279 maybe like a SaaS product, it's almost like having an 503 00:29:09.319 --> 00:29:13.720 assistant that you can work with so that helps onboard 504 00:29:13.799 --> 00:29:17.160 you with that that or can even perform tasks for 505 00:29:17.240 --> 00:29:20.440 you within the software. So you say, instead of like 506 00:29:20.480 --> 00:29:22.559 trying to figure out, okay, let's say if we have 507 00:29:22.680 --> 00:29:26.279 a CRM, instead of being able to figure out, okay, 508 00:29:26.319 --> 00:29:28.519 how do I add a new customer type of thing 509 00:29:28.559 --> 00:29:33.359 and what's available here? With traditional kind of like a 510 00:29:33.480 --> 00:29:39.000 SaaS product, you have a real estate that is limited. 511 00:29:39.559 --> 00:29:42.319 So I mean, you don't want to stick a thousand 512 00:29:42.319 --> 00:29:46.440 different buttons on one page and expect someone to figure 513 00:29:46.440 --> 00:29:48.440 out how to use all these different buttons to do 514 00:29:48.480 --> 00:29:53.400 different things or forms and things like that. That would 515 00:29:54.240 --> 00:29:57.880 that's too much information for the end user. But the 516 00:29:57.880 --> 00:30:01.319 cool thing about working with something like that language is 517 00:30:01.400 --> 00:30:05.480 you could have thousands of different tools and tasks that 518 00:30:05.519 --> 00:30:08.519 it can perform for you. It can understand the whole website, 519 00:30:08.599 --> 00:30:10.279 it can navigate you to where you need to be. 520 00:30:10.680 --> 00:30:13.759 It can explain how to use it, or it can 521 00:30:13.839 --> 00:30:16.079 accomplish those tasks for you and fill in those forms 522 00:30:16.079 --> 00:30:19.519 for you, and they could once you fill those forms 523 00:30:19.720 --> 00:30:22.200 or have those kind of filled out, it can look 524 00:30:22.200 --> 00:30:25.039 at those and say, hey, I see that the way 525 00:30:25.079 --> 00:30:27.680 you fill this out is this way, but here's some 526 00:30:27.720 --> 00:30:31.359 more context on tips and tricks to make this better 527 00:30:31.480 --> 00:30:34.160 or maybe just improve improve it for them and then 528 00:30:34.319 --> 00:30:36.799 ask them to confirm, hey, this is exactly what I'm 529 00:30:36.799 --> 00:30:43.119 looking for. So working with it's definitely it's very different 530 00:30:43.200 --> 00:30:46.640 in the sense that with traditional programming, a lot of 531 00:30:46.680 --> 00:30:50.039 times we talk about like pure functions, where this one 532 00:30:50.079 --> 00:30:53.200 input is always going to give you this output. That's 533 00:30:53.279 --> 00:30:55.680 kind of one of the differences working with AI too. 534 00:30:56.000 --> 00:30:58.519 It's a little bit more of a black box where 535 00:30:58.759 --> 00:31:03.319 this input could be, uh, the answer coming out could 536 00:31:03.359 --> 00:31:06.720 be multiple different kind of outputs. So one of the 537 00:31:06.720 --> 00:31:09.799 things I like to do when making AI workflows too, 538 00:31:10.640 --> 00:31:13.680 is anytime anytime there's something that could be done with 539 00:31:13.759 --> 00:31:18.759 traditional programming, like say, if if it's parsing out information, 540 00:31:20.240 --> 00:31:26.119 it's doing math type of thing, whatever, it's typically you 541 00:31:26.160 --> 00:31:29.119 want to kind of steer away from AI, and so 542 00:31:29.640 --> 00:31:32.960 you want to parse the information out from the answers 543 00:31:33.279 --> 00:31:37.880 and then and then have kind of traditional programming actually 544 00:31:38.839 --> 00:31:42.039 a figure out what what the output should be and 545 00:31:42.079 --> 00:31:46.839 then AI use AI usually where where it performs best basically, 546 00:31:47.440 --> 00:31:51.039 and so and because one of the things that you 547 00:31:52.160 --> 00:31:57.119 dealing with AI's hallucinations, and so that's one of the 548 00:31:57.160 --> 00:32:02.279 things we're working with workflows as well, is dealing with 549 00:32:02.640 --> 00:32:05.440 hallucinations and understanding when it's important when it's not. Like 550 00:32:05.480 --> 00:32:08.519 you said, you had a friend that using AI to 551 00:32:09.000 --> 00:32:14.319 write stories, helping out that process, like that's kind of 552 00:32:14.319 --> 00:32:16.279 an interesting use case and something that kind of deal 553 00:32:16.359 --> 00:32:20.160 with with speak magic is because there's not necessarily a 554 00:32:20.240 --> 00:32:23.599 right answer or wrong answer necessarily, but there could be 555 00:32:23.839 --> 00:32:28.960 a preferred answer or a preferred response versus a non response, 556 00:32:30.759 --> 00:32:33.599 and so it could be a little bit more difficult 557 00:32:33.640 --> 00:32:37.160 on judging what is good and what's not. Except for 558 00:32:37.200 --> 00:32:39.960 having a human going and basically say hey, I like 559 00:32:40.039 --> 00:32:42.359 this and I didn't like this type of thing, and 560 00:32:42.680 --> 00:32:49.519 refining the prompts that you use with the agent versus 561 00:32:49.519 --> 00:32:54.200 something like where you might have a specific right or 562 00:32:54.440 --> 00:32:58.720 wrong answer. Then it's a lot easier to to improve 563 00:32:58.759 --> 00:33:01.000 the prompts and then just doing a whole bunch of 564 00:33:01.039 --> 00:33:04.279 iterations and tests and say, okay, this is you know 565 00:33:04.359 --> 00:33:09.119 exactly if this is a correct answer or correct response 566 00:33:09.200 --> 00:33:10.319 or incorrect response. 567 00:33:11.759 --> 00:33:14.519 It's kind of one of the ways working with it too. 568 00:33:14.599 --> 00:33:20.160 That's a little different too. So when you're writing these 569 00:33:20.160 --> 00:33:23.720 tools for other people, then are you doing some of 570 00:33:23.720 --> 00:33:27.519 that where you're actually trying to like beyond what they 571 00:33:27.559 --> 00:33:29.480 give you, you know, you add other things to the 572 00:33:29.519 --> 00:33:35.119 prompt get them better answers or yeah, so let it ride. 573 00:33:36.160 --> 00:33:41.920 Yeah, And that's one of the things ISH is trying 574 00:33:41.920 --> 00:33:46.640 to steer it down certain paths that will give you 575 00:33:46.680 --> 00:33:54.039 reliable outputs to what they're looking for. So uh and 576 00:33:54.079 --> 00:33:58.920 so one one of the ways, I mean, there's different 577 00:33:59.440 --> 00:34:01.680 kind of things to kind of look for in that 578 00:34:02.359 --> 00:34:09.519 one is when you getting like a prompt, one example 579 00:34:09.800 --> 00:34:13.320 was speak magic is is okay, we have it create 580 00:34:13.480 --> 00:34:17.239 like a story first, like a summary, and then we 581 00:34:17.320 --> 00:34:21.400 have it actually we actually have it parse out, like 582 00:34:21.440 --> 00:34:23.639 so we have within the story, we might add like 583 00:34:24.159 --> 00:34:26.840 some like H one tags that basically break out the 584 00:34:26.840 --> 00:34:29.480 story and says, okay, we're on act one or something 585 00:34:29.519 --> 00:34:33.440 like that. It actually chooses a story model first, and 586 00:34:33.440 --> 00:34:37.159 then it has these different steps within the story, and 587 00:34:37.239 --> 00:34:40.280 so we're having it kind of guiding it down a 588 00:34:40.320 --> 00:34:42.880 certain kind of framework, and then we split out that 589 00:34:42.960 --> 00:34:46.119 story by like having it add like h ones for 590 00:34:46.239 --> 00:34:49.719 the different sections, different steps of the story. And then 591 00:34:50.159 --> 00:34:53.239 and then we'll have it turn that portion of the 592 00:34:53.280 --> 00:34:59.360 story into different scenes and and then so like one 593 00:34:59.360 --> 00:35:02.800 of the things it'll do is like, okay, One thing 594 00:35:02.840 --> 00:35:05.360 I found is I would have it to do, like, say, 595 00:35:05.360 --> 00:35:07.280 if we're turning it into a scene. I was having 596 00:35:07.320 --> 00:35:10.199 it do too many different things at once. So I 597 00:35:10.239 --> 00:35:11.880 was saying, hey, I want you to write a scene 598 00:35:11.920 --> 00:35:14.320 for this portion of the story, and I want it 599 00:35:14.360 --> 00:35:17.039 to be in this very specific format that then I 600 00:35:17.039 --> 00:35:20.719 can then parse out to make shots from that scene. 601 00:35:21.360 --> 00:35:24.599 And so one of the issues I was running into 602 00:35:24.760 --> 00:35:26.599 is it was having a hard time doing both of 603 00:35:26.639 --> 00:35:31.760 those steps, and so what I ended up doing is say, okay, well, 604 00:35:32.119 --> 00:35:34.679 and all of them is actually fairly good at making 605 00:35:34.960 --> 00:35:39.360 a scene like a screenplay like a script basically because 606 00:35:39.400 --> 00:35:41.280 it has a lot of that information how to do that. 607 00:35:42.400 --> 00:35:44.639 And so I broke that out and said, okay, do 608 00:35:44.719 --> 00:35:48.239 that first, and then I have it's and it's response 609 00:35:48.400 --> 00:35:51.800 of that script to the next AI agent that says, okay, 610 00:35:51.840 --> 00:35:55.840 now let's add the formatting on here, and gave it 611 00:35:55.880 --> 00:35:58.039 all the rules of how to add that formatting to 612 00:35:58.119 --> 00:36:02.559 that script for that scene, and then that allows it 613 00:36:02.599 --> 00:36:05.400 to then parse and say, okay, this is how we 614 00:36:05.400 --> 00:36:08.320 we define all the different shots that make up that scene. 615 00:36:08.760 --> 00:36:11.239 And so then I have JavaScript basically go through and 616 00:36:11.320 --> 00:36:16.280 split all that into different pieces and into different shots, 617 00:36:16.440 --> 00:36:21.000 and then I can have the NEXTAI agent basically parse 618 00:36:21.239 --> 00:36:23.079 the shot out and figure out, Okay, what do we 619 00:36:23.159 --> 00:36:27.599 need to do to create this shot? 620 00:36:28.400 --> 00:36:31.920 Awesome? So we're using Jason mode because I know some 621 00:36:31.960 --> 00:36:34.440 of these allow you to send the data over as 622 00:36:34.559 --> 00:36:40.320 Jason as opposed to straight text or other formats. Most llms, 623 00:36:40.360 --> 00:36:42.960 I will say, are pretty good about pulling text apart 624 00:36:43.000 --> 00:36:44.960 and putting it back together and figuring out what you want. 625 00:36:45.920 --> 00:36:50.039 The Jason just gives you more accuracy, is what I found. No, 626 00:36:50.199 --> 00:36:51.159 I totally agree. 627 00:36:51.199 --> 00:36:56.079 So like with speak Magic, I'd say about eighty percent 628 00:36:56.719 --> 00:37:00.360 to ninety percent of all the agents I use actually 629 00:37:00.440 --> 00:37:05.280 using the tool use the Jayson mode and and that 630 00:37:05.440 --> 00:37:07.679 there's a lot of vantages for that. So if it's 631 00:37:07.800 --> 00:37:10.800 writing story, then I just typically I'll just have it 632 00:37:10.880 --> 00:37:15.440 just write the story using kind of just plain kind 633 00:37:15.480 --> 00:37:19.760 of texts. It's usually marked down, formatted coming out. 634 00:37:20.360 --> 00:37:20.679 And so. 635 00:37:22.320 --> 00:37:26.400 But with a Jason there's, uh one of the cool 636 00:37:26.400 --> 00:37:29.239 things that that one of the strategies I use with 637 00:37:29.280 --> 00:37:32.320 that is is with a Jason mode, you can actually 638 00:37:32.440 --> 00:37:36.960 like say okay, I want it's basically have a jasent 639 00:37:37.000 --> 00:37:39.239 object that has different properties that needs to fill in. 640 00:37:39.800 --> 00:37:41.719 And so one of the cool things about it is 641 00:37:41.719 --> 00:37:45.159 that actually has an order of how those things are 642 00:37:45.239 --> 00:37:48.360 because elms are sequential, so they do token by token 643 00:37:49.119 --> 00:37:52.960 and so it'll the top. Even though like maybe an 644 00:37:53.000 --> 00:37:57.559 object in JavaScript isn't necessarily has has an order when 645 00:37:57.559 --> 00:37:59.920 it's filling this out, it's actually has an order. So 646 00:38:00.000 --> 00:38:02.679 so one of the things I've done with it is 647 00:38:02.679 --> 00:38:06.519 is have it actually think through step by step of 648 00:38:06.599 --> 00:38:10.599 how to approach a problem and maybe several of these 649 00:38:10.639 --> 00:38:13.360 properties that's filling out. It never really ends up using. 650 00:38:13.400 --> 00:38:16.079 But it's a way to force it to think like 651 00:38:16.440 --> 00:38:19.400 lineally in a certain way that prepares it to give 652 00:38:20.239 --> 00:38:23.159 accurate answers, if that makes sense. So it's kind of 653 00:38:23.199 --> 00:38:26.880 like a form of like the thinking model that you 654 00:38:26.920 --> 00:38:31.400 see things like, oh, one model that kind of does, 655 00:38:31.480 --> 00:38:35.159 but you're you're able to then specifically guide it through 656 00:38:35.199 --> 00:38:37.880 a whole step of like, Okay, let's build our own 657 00:38:37.960 --> 00:38:41.639 context of like and have it think through different steps 658 00:38:41.679 --> 00:38:44.840 by filling in these different properties and then being able 659 00:38:44.880 --> 00:38:47.400 to use that context and it's thinking to be able 660 00:38:47.480 --> 00:38:51.719 to force it to then give the answer that or 661 00:38:51.760 --> 00:38:54.280 the properties that you're actually going to use for the 662 00:38:54.320 --> 00:38:56.480 next steps or for the end user. 663 00:38:56.559 --> 00:39:01.880 Basically, yeah, makes sense. So one other thing that I'm 664 00:39:01.880 --> 00:39:04.599 looking at, so just give a little bit of context 665 00:39:04.639 --> 00:39:06.719 for people here. I have two things going that I'm 666 00:39:06.800 --> 00:39:09.800 wanting to put together. One of them is I would 667 00:39:09.840 --> 00:39:12.599 like to have some kind of AI help agent kind 668 00:39:12.599 --> 00:39:15.320 of thing on top end depth. Right, So if people 669 00:39:15.360 --> 00:39:18.119 are looking to go through some of our courses or 670 00:39:18.159 --> 00:39:22.360 things like that, you know that there's essentially an AI 671 00:39:22.440 --> 00:39:24.480 agent that can help you figure out what to learn 672 00:39:24.519 --> 00:39:27.960 next or you know, maybe next steps for your career 673 00:39:28.039 --> 00:39:31.519 or things like that, and so I could see that 674 00:39:31.599 --> 00:39:36.599 as a kind of a coach if nothing else. And 675 00:39:36.679 --> 00:39:40.239 I'm trying to figure out how to manage the context 676 00:39:40.320 --> 00:39:42.320 right because people may ask a lot of questions or 677 00:39:42.360 --> 00:39:45.000 give it a lot of information. The other thing I'm 678 00:39:45.039 --> 00:39:49.599 running into is if somebody leaves and comes back, right, 679 00:39:49.679 --> 00:39:52.519 then do I summarize the previous context and hand it 680 00:39:52.559 --> 00:39:54.679 to a new query or can I just pick up 681 00:39:54.679 --> 00:39:57.599 where I left off? So, do you have any recommendations 682 00:39:57.599 --> 00:40:00.480 on something like that where maybe it's not one continue 683 00:40:00.480 --> 00:40:00.960 as session? 684 00:40:06.320 --> 00:40:10.840 So me, maybe can I re explain that that question again? 685 00:40:11.719 --> 00:40:15.679 Yeah, so so let's say that it's it's an ongoing tool, right. 686 00:40:15.760 --> 00:40:18.280 So the other the other idea I have that I 687 00:40:18.320 --> 00:40:22.119 want to build out is essentially I've hired virtual assistance 688 00:40:22.159 --> 00:40:23.559 in the past to help me do a lot of 689 00:40:23.840 --> 00:40:27.480 kind of routine things with the podcasts, And so with 690 00:40:27.519 --> 00:40:29.400 that one, I kind of see more of the team 691 00:40:29.480 --> 00:40:32.360 model and I can just come in and you know, 692 00:40:32.519 --> 00:40:34.360 I can just use a tool for it to go 693 00:40:34.360 --> 00:40:37.599 look up the information about my podcast, and then you know, 694 00:40:37.679 --> 00:40:39.519 I can tell it to do tasks one off at 695 00:40:39.559 --> 00:40:43.639 a time. And so for that one, I'm more thinking Okay, 696 00:40:43.639 --> 00:40:45.800 how do I just make sure that the workflow works. 697 00:40:46.199 --> 00:40:49.880 But for the other one, you know, where I wanted 698 00:40:49.880 --> 00:40:54.000 to remember things about the people who are coming and 699 00:40:54.039 --> 00:40:57.159 asking for help, and I want if they show up 700 00:40:57.159 --> 00:41:00.079 and say, okay, I just got a job interview, how 701 00:41:00.119 --> 00:41:02.880 do I prepare for this? You know, it's smart enough 702 00:41:02.920 --> 00:41:04.920 to say, okay, well, what can you tell me about 703 00:41:04.920 --> 00:41:07.599 the company? You know, maybe get a little more information, 704 00:41:08.079 --> 00:41:10.719 and then turn around and actually remember enough about them 705 00:41:11.599 --> 00:41:13.840 right to help them. Do do I have to store 706 00:41:13.880 --> 00:41:18.360 that myself to get the continuity or can I you know, 707 00:41:19.000 --> 00:41:22.400 is there some form of cashing the context window that 708 00:41:22.440 --> 00:41:25.960 the different lms do, because most of the time you're 709 00:41:26.039 --> 00:41:30.079 hitting these over APIs and so anyway, that's kind of 710 00:41:30.079 --> 00:41:32.639 what I'm wondering, guys, is allowing people to pick up 711 00:41:32.679 --> 00:41:34.519 where they left off when they're using an agent. 712 00:41:35.800 --> 00:41:41.840 Yeah, so when using all these different LM models, every 713 00:41:41.840 --> 00:41:45.800 time you send a request the API, you have to 714 00:41:46.000 --> 00:41:50.320 actually give it the full context of the history. And 715 00:41:50.360 --> 00:41:52.440 so that was one of the things that kind of 716 00:41:52.559 --> 00:41:54.639 hit me. I thought, for some reason I thought when 717 00:41:54.639 --> 00:41:56.639 I first got it involved a couple of years ago, 718 00:41:58.239 --> 00:42:00.960 is that you just remember your past calls like maybe 719 00:42:00.960 --> 00:42:03.840 you get like like a conversation idea or something like that, 720 00:42:03.960 --> 00:42:07.719 and it would handle all the state and the memory 721 00:42:08.039 --> 00:42:11.719 of past conversing with that like conversation idea or something 722 00:42:11.719 --> 00:42:16.119 like that. But it turns out like it's actually more stateless, 723 00:42:16.639 --> 00:42:19.679 and so you have to send it all the information, 724 00:42:20.280 --> 00:42:24.039 which has advantages of disvantage one and the disvantage would 725 00:42:24.079 --> 00:42:27.800 be be a lot simpler just to maybe have it 726 00:42:27.920 --> 00:42:30.239 manage it. But then you have a whole bunch of 727 00:42:30.239 --> 00:42:33.239 control over what kind of context that you're giving to it, 728 00:42:33.679 --> 00:42:36.559 and so you need to keep track of that information 729 00:42:36.760 --> 00:42:40.320 that could be I mean typically would be like in 730 00:42:40.320 --> 00:42:43.199 a database or something like that, and then you can 731 00:42:43.239 --> 00:42:46.960 pull It's also important though too. We talked about context window, 732 00:42:47.239 --> 00:42:52.679 and different models have different sizes of context window. I 733 00:42:52.719 --> 00:42:55.760 mean you're looking at like different ones. I mean some 734 00:42:55.880 --> 00:43:00.320 could be like only eight thousand output tokens. Some of 735 00:43:00.320 --> 00:43:04.400 them have like one hundred and twenty eight thousand, like 736 00:43:04.519 --> 00:43:06.639 opening eyes. A lot of them one hundred and twenty 737 00:43:06.679 --> 00:43:11.480 eight or two hundred thousand context window. And so sometimes 738 00:43:11.679 --> 00:43:15.280 what you need to do is if you want kind 739 00:43:15.280 --> 00:43:18.480 of the more full context of things, once you're you 740 00:43:18.599 --> 00:43:21.960 start using more and more up is to summarize the 741 00:43:22.039 --> 00:43:26.039 information or to pull out key pieces of information that 742 00:43:26.079 --> 00:43:30.039 are really important. And that's one of the things we do, 743 00:43:30.159 --> 00:43:34.280 like speak Magic for instance, is we have kind of 744 00:43:34.320 --> 00:43:38.320 a kind of in a sense like an adjson object 745 00:43:38.440 --> 00:43:42.199 that keeps all the important information like what kind of 746 00:43:42.199 --> 00:43:44.199 style does the video you need to be? Is it 747 00:43:44.280 --> 00:43:48.239 more animated? Is it live action like type of thing, 748 00:43:48.320 --> 00:43:51.239 because you want that consistency from shot to shot, from 749 00:43:51.239 --> 00:43:54.599 scene to scene for the video, like who are the 750 00:43:54.679 --> 00:43:58.719 characters and like character profiles of each one, so we 751 00:43:58.760 --> 00:44:01.880 have very descriptive of a lot of different things of 752 00:44:01.920 --> 00:44:04.920 that character. And that way, when when we're generating an 753 00:44:04.960 --> 00:44:09.360 image of that character, it will have all that context. 754 00:44:09.400 --> 00:44:13.599 It's okay, this character has curly black hair or something 755 00:44:13.679 --> 00:44:16.360 like that, and and different kind of things, so that 756 00:44:16.400 --> 00:44:19.280 when it generates the image before it turns into a video. 757 00:44:19.639 --> 00:44:20.360 It'll be correct. 758 00:44:20.360 --> 00:44:23.519 But and we also don't want to overload the context 759 00:44:23.559 --> 00:44:25.800 as well, so it doesn't need So it's basically, when 760 00:44:26.000 --> 00:44:28.679 when you're working with a call, you want to do 761 00:44:28.760 --> 00:44:32.920 kind of a need to know like context, and so 762 00:44:33.039 --> 00:44:36.960 you want to be dynamically being able to decide on 763 00:44:37.039 --> 00:44:40.880 what context is important. For that specific call. This is 764 00:44:40.880 --> 00:44:45.079 really important also, so and you kind of have different 765 00:44:45.119 --> 00:44:47.000 types of memory as well. So you're going to have 766 00:44:47.039 --> 00:44:50.639 memory for what is the state important for this specific 767 00:44:50.639 --> 00:44:53.800 call at the moment, there's going to be kind of 768 00:44:54.079 --> 00:44:56.320 memory where it has to do with the whole run 769 00:44:56.360 --> 00:45:00.199 of the workflow, especially when you have cycles of humans 770 00:45:00.239 --> 00:45:04.719 going through so like you get a question or you 771 00:45:05.119 --> 00:45:07.559 give a prompt, you get a response, and then you're 772 00:45:07.559 --> 00:45:11.320 continuing that conversation, and so you kind of need more 773 00:45:11.360 --> 00:45:15.639 of a more global kind of memory that matters as well. 774 00:45:18.800 --> 00:45:22.719 Gotcha, Okay, Yeah, I guess my concern was if I'm 775 00:45:22.760 --> 00:45:25.199 gathering information, you know, if I have to store information 776 00:45:25.280 --> 00:45:29.639 when people are asking more personal things like should I 777 00:45:29.719 --> 00:45:32.880 quit my job or should I you know, you know, 778 00:45:32.960 --> 00:45:34.760 they tell it I'm not happy where I'm at and 779 00:45:34.760 --> 00:45:37.760 so I'm looking for another place or things like that. 780 00:45:39.119 --> 00:45:42.679 I mean, I think the expectation is that it's confidential, right, 781 00:45:42.840 --> 00:45:45.000 I mean, if I'm given it a creative prompt and 782 00:45:45.480 --> 00:45:47.840 I'm having a create part of my story, maybe I'm 783 00:45:47.920 --> 00:45:52.320 you know, I don't feel as compromised by somebody seeing oh, 784 00:45:52.440 --> 00:45:55.239 you know, he generated a video with a you know, 785 00:45:55.360 --> 00:45:58.960 with a girl with black curly hair and a you know, 786 00:45:59.119 --> 00:46:02.039 and a you know, a villain creature thing that does 787 00:46:02.280 --> 00:46:04.360 you know, has certain powers or whatever, whatever, you know, 788 00:46:04.400 --> 00:46:09.079 whatever my story's about, right, it's different than if it's, oh, well, 789 00:46:09.199 --> 00:46:11.920 we didn't know that Chuck was not happy here and 790 00:46:12.079 --> 00:46:14.960 was going to start looking for another job, or you know, 791 00:46:15.400 --> 00:46:18.880 asked how to answer questions about things that he's not 792 00:46:19.039 --> 00:46:22.440 proud of in his career past, and so, you know, 793 00:46:23.119 --> 00:46:26.400 so keeping that confidential seems pretty important. But that's a 794 00:46:26.400 --> 00:46:28.800 common problem for a lot of other things. So you're 795 00:46:28.840 --> 00:46:30.920 telling me that, I that's something that I have to 796 00:46:30.960 --> 00:46:34.280 figure out because the LLLM isn't gonna keep track of 797 00:46:34.360 --> 00:46:37.519 it for me. One last question, because we're kind of 798 00:46:37.519 --> 00:46:42.280 getting toward the end of our scheduled time, is because 799 00:46:42.280 --> 00:46:43.840 I know a lot of people are interested in getting 800 00:46:43.880 --> 00:46:48.760 involved in AI and learning AI, and to be perfectly honest, 801 00:46:48.800 --> 00:46:52.480 I feel like if you're not if you're not getting 802 00:46:52.480 --> 00:46:54.599 a feel for how AI works and what it can 803 00:46:54.639 --> 00:46:58.079 do and how it can at least help you where 804 00:46:58.119 --> 00:47:00.880 you're at, let alone, where it can help the users 805 00:47:00.880 --> 00:47:04.320 of your applications where they're at, then in a year 806 00:47:04.400 --> 00:47:06.519 or so, you're going to be way behind. And so 807 00:47:07.039 --> 00:47:09.239 if people are sitting here and maybe you have a 808 00:47:09.239 --> 00:47:10.880 different opinion on that, and you can say that in 809 00:47:10.880 --> 00:47:14.599 a minute, But at the end of the day, if 810 00:47:14.599 --> 00:47:17.400 somebody's going I need to understand AI, what do I 811 00:47:17.480 --> 00:47:19.440 need to do in order to get there? Where do 812 00:47:19.480 --> 00:47:22.400 you recommend people start and what are the kinds of 813 00:47:22.440 --> 00:47:24.480 things that they ought to be picking up in order 814 00:47:24.559 --> 00:47:25.800 to be successful with this? 815 00:47:27.519 --> 00:47:30.880 Yeah, I think in some ways I would say kind 816 00:47:30.880 --> 00:47:34.519 of putch it like the same of like learning programming. So, 817 00:47:35.400 --> 00:47:38.679 I mean, there's so many different directions you can go 818 00:47:38.760 --> 00:47:41.960 with it, but it's usually best to pick out, like 819 00:47:42.440 --> 00:47:44.800 if you want to learn more about something like a 820 00:47:44.840 --> 00:47:48.000 programming language, is to pick out like a simple small 821 00:47:48.039 --> 00:47:52.360 project and that you're interested in or maybe that's what 822 00:47:52.400 --> 00:47:57.840 you want to provide type of things of value and 823 00:47:57.960 --> 00:48:03.199 then and then learn the parts of that to accomplish 824 00:48:03.239 --> 00:48:05.079 that kind of programming task. 825 00:48:05.719 --> 00:48:06.039 And so. 826 00:48:07.559 --> 00:48:11.320 I mean one of the great resources for that is 827 00:48:11.320 --> 00:48:16.840 is using like a AI chatbot to do the research 828 00:48:16.880 --> 00:48:21.039 for you and answer the questions for you. It's actually, 829 00:48:21.599 --> 00:48:24.719 I mean, it's a great way I've been using it 830 00:48:24.760 --> 00:48:28.960 for I've been learning Java recently, and it's a great 831 00:48:28.960 --> 00:48:32.840 way to You could have it like quiz you on 832 00:48:32.920 --> 00:48:37.039 what you know, what you don't know, or answer the questions. 833 00:48:37.079 --> 00:48:39.880 It's like a fantastic tool to be able to learn 834 00:48:39.960 --> 00:48:44.400 new technologies and things. But there are I mean, there's 835 00:48:44.800 --> 00:48:49.119 different ways of kind of approaching it, and so you 836 00:48:49.159 --> 00:48:54.960 can using JavaScript, you can go just plane, vanilla node 837 00:48:55.119 --> 00:49:00.559 type of thing setting you're accessing these different APIs UH 838 00:49:00.800 --> 00:49:04.320 work with maybe one API at first. A lot of 839 00:49:04.360 --> 00:49:06.559 them are fairly similar to each other when it comes 840 00:49:06.599 --> 00:49:11.280 like l MS, and so you can there's different tools 841 00:49:11.360 --> 00:49:14.280 you can use if you want to go beyond kind 842 00:49:14.280 --> 00:49:17.159 of figuring out that kind of thing yourself, UH which 843 00:49:17.159 --> 00:49:20.960 you can have AI help you write things different parts 844 00:49:21.119 --> 00:49:24.599 of creating these workflows and stuff. You can use tools 845 00:49:24.679 --> 00:49:26.960 though that get you a little further ahead. If you're 846 00:49:27.039 --> 00:49:31.360 using JavaScript, like something like line chain dot j s 847 00:49:31.639 --> 00:49:34.000 will give you a bit of a framework of how 848 00:49:34.079 --> 00:49:38.840 to to work with things. There are other tools out there, 849 00:49:39.760 --> 00:49:42.199 like if you're looking more for like if you're more 850 00:49:42.239 --> 00:49:46.320 interested in just AI workflows, you can use tools like 851 00:49:46.840 --> 00:49:50.400 uh in eight in which allows you to like a 852 00:49:50.480 --> 00:49:55.199 visual kind of way of putting together these workflows. Wows 853 00:49:55.280 --> 00:49:59.079 AI h is a way to do the kind of 854 00:49:59.119 --> 00:50:03.480 same thing where you're just using Localde tools, so you're 855 00:50:03.519 --> 00:50:06.559 building things out visually, and then you can use JavaScript 856 00:50:06.599 --> 00:50:12.559 expressions to basically to parse information, to decide what contexts 857 00:50:12.559 --> 00:50:16.239 to use in different places, and to kind of help 858 00:50:16.360 --> 00:50:23.079 control of how the workflow is ran. But there's I mean, 859 00:50:23.119 --> 00:50:28.159 you can use things like Cursor, which would be more 860 00:50:28.159 --> 00:50:32.679 of like an assistant that helps you within like your 861 00:50:32.719 --> 00:50:37.480 ide to build, to help you generate code quickly that 862 00:50:37.519 --> 00:50:40.079 could help you in the process of doing things. 863 00:50:40.320 --> 00:50:42.000 Or you could use like more of. 864 00:50:41.920 --> 00:50:46.920 A a coding agent like clawed code or like Manus, 865 00:50:47.039 --> 00:50:51.599 which I think Manus has a that's a not necessarily 866 00:50:51.639 --> 00:50:54.639 public for everyone right now as far as I know, 867 00:50:56.039 --> 00:50:58.880 and that will help generate larger kind of projects from 868 00:50:59.320 --> 00:51:01.880 single prompts you do more kind of like the vibe 869 00:51:02.239 --> 00:51:08.440 coding which was brought up earlier. So those are kind 870 00:51:08.440 --> 00:51:12.960 of some different tools that you could use. Another tool 871 00:51:13.880 --> 00:51:15.760 maybe look at if you want to use a lot 872 00:51:15.800 --> 00:51:21.280 of different models, would be something like open MCP, which 873 00:51:21.360 --> 00:51:25.599 is a JavaScript kind of implementation of that allows you 874 00:51:25.639 --> 00:51:32.519 to help standardize using different models and different kind of 875 00:51:32.519 --> 00:51:37.079 tools that you could use. I mean, we have something 876 00:51:37.119 --> 00:51:39.039 kind of similar that we added a long time ago, 877 00:51:39.079 --> 00:51:44.559 and to WOWS that we have like thirty I think 878 00:51:44.559 --> 00:51:47.239 it's thirty three different models that you can access, and 879 00:51:47.280 --> 00:51:49.960 they have has a very similar architecturecause you're just looking 880 00:51:50.000 --> 00:51:54.360 at bring a prompt in and then a response out basically, 881 00:51:54.719 --> 00:51:56.480 and then we handle all the API kind of stuff 882 00:51:56.480 --> 00:52:01.079 for you. So those are kind of different things. 883 00:52:01.599 --> 00:52:02.559 I kind of I. 884 00:52:02.480 --> 00:52:06.039 Would recommend for something that's learning different the tools and 885 00:52:06.079 --> 00:52:08.400 things that they can get involved with and use. 886 00:52:09.320 --> 00:52:11.400 Yeah, one thing that I'll add to that is and 887 00:52:11.679 --> 00:52:13.320 you've kind of alluded to it in the way that 888 00:52:13.360 --> 00:52:18.599 you've told people to approach stuff, is you know, before 889 00:52:18.639 --> 00:52:21.159 you're even writing code, you can just go in and 890 00:52:21.199 --> 00:52:23.480 go to like chat, gpt dot com or grock dot 891 00:52:23.519 --> 00:52:26.880 com or anthropic. I can't remember that you are all 892 00:52:26.960 --> 00:52:30.679 to use claude, you know, but just just get in 893 00:52:30.719 --> 00:52:32.960 and just start asking it questions and kind of get 894 00:52:33.039 --> 00:52:35.880 used to how it works, because ultimately what you're gonna 895 00:52:35.880 --> 00:52:40.079 be sending over is prompts a look a whole lot 896 00:52:40.119 --> 00:52:43.239 like your questions anyway, and so you can figure out, 897 00:52:43.400 --> 00:52:46.000 you know, the different tricks that work and then from there. 898 00:52:46.039 --> 00:52:48.960 I also recommend that people go pick up a course 899 00:52:49.039 --> 00:52:54.480 or a book or something that does some explanation on 900 00:52:55.079 --> 00:53:00.519 prompt engineering. The reason is is because the one is 901 00:53:00.760 --> 00:53:03.519 knowing how to access tools and access the AI and 902 00:53:03.559 --> 00:53:06.119 how to ask go questions. But the rest of it 903 00:53:06.159 --> 00:53:07.679 is just going to be down to your prompts and 904 00:53:07.679 --> 00:53:10.360 how you format them so that you're getting the best 905 00:53:10.440 --> 00:53:11.920 answers possible from your AI. 906 00:53:12.440 --> 00:53:17.440 So and there's a lot of iteration. So, yeah, something 907 00:53:17.480 --> 00:53:20.719 doesn't work with your prompt engineering, try tweaking things. Pull 908 00:53:20.719 --> 00:53:23.679 out pronouns like don't use this or that, be very 909 00:53:23.679 --> 00:53:27.320 explicit on what you mean about things. Keep the context low. 910 00:53:28.199 --> 00:53:34.159 Yeah, all right, well let's go ahead and do some pics. Now. 911 00:53:34.199 --> 00:53:36.000 I don't think you've been on the show before, so 912 00:53:36.480 --> 00:53:38.599 let me just explain what they are real quick. It's 913 00:53:38.679 --> 00:53:42.599 just us shouting out about stuff we like. So a 914 00:53:42.639 --> 00:53:46.079 lot of times people do TV shows or movies or 915 00:53:46.280 --> 00:53:50.880 technology tools or anything in between. We'll let Steve go first, 916 00:53:50.880 --> 00:53:52.400 and then I'll go, and then you can go last, 917 00:53:52.400 --> 00:53:53.440 and that way you can kind of get a feel 918 00:53:53.480 --> 00:53:54.079 for how we do it. 919 00:53:56.119 --> 00:53:58.519 Yeah, Matthew, Just one thing he didn't mention is that 920 00:53:58.599 --> 00:54:00.920 the high point of every episode of ours is my 921 00:54:01.079 --> 00:54:08.880 dad jokes, dad jokes of the week, and so anyway, 922 00:54:08.920 --> 00:54:13.840 make sure hopefully my sound effects are working properly. Okay. 923 00:54:14.039 --> 00:54:18.360 So as an example, having a conversation with my friend 924 00:54:18.360 --> 00:54:21.239 and I said I actually have a half brother and 925 00:54:21.280 --> 00:54:26.960 he said different mothers. I said, nope, shark attack, thank you, 926 00:54:27.159 --> 00:54:32.519 thank you. So last week was Saint Patrick's Day, I 927 00:54:32.519 --> 00:54:35.559 think was a week ago today, and I bought a 928 00:54:35.599 --> 00:54:38.519 diamond ring for my wife, but it turned out to 929 00:54:38.559 --> 00:54:45.159 be a fake. They gave me a sham rock. And 930 00:54:45.239 --> 00:54:48.039 then my dentist, who's actually a good friend of mine, 931 00:54:48.079 --> 00:54:48.519 I'm known for. 932 00:54:48.519 --> 00:54:49.039 A long time. 933 00:54:49.159 --> 00:54:52.840 He he got this local award where he was voted 934 00:54:52.880 --> 00:54:55.840 the dentist of the year. He didn't get a trophy though, 935 00:54:55.880 --> 00:55:03.079 he just got a little plaque. And then finally question 936 00:55:03.559 --> 00:55:06.360 in King Arthur's time, which of the Knights of the 937 00:55:06.480 --> 00:55:11.079 Roundtable collected taxes? Because you know they collected taxes, sir, 938 00:55:11.239 --> 00:55:16.760 charge those are the dad jokes. 939 00:55:16.480 --> 00:55:20.440 Of the week. All right, well how do you how 940 00:55:20.440 --> 00:55:27.360 do you follow that? Very humbly, very humbly. Right. So, yeah, 941 00:55:27.400 --> 00:55:28.960 so I don't know if I've played any new board 942 00:55:29.000 --> 00:55:33.239 games lately, so I'm just gonna throw out something that 943 00:55:33.280 --> 00:55:36.679 I've picked in the past. This is something that we've 944 00:55:36.679 --> 00:55:40.760 played before, me and the guys. I did find out though, 945 00:55:40.800 --> 00:55:42.920 that that has a different mode to it you can play, 946 00:55:42.960 --> 00:55:44.920 so I'm gonna pick it again even though I haven't 947 00:55:44.920 --> 00:55:50.239 played the mode. The campaign mode the board game is 948 00:55:50.239 --> 00:55:55.000 called Heat Pedal to the Metal. It's a racing game, 949 00:55:55.079 --> 00:55:59.800 so everyone's in race cars. You play cards in order 950 00:55:59.800 --> 00:56:02.320 to move forward. If you go around the turns too fast, 951 00:56:02.360 --> 00:56:06.599 then you take heat from your engine and put it 952 00:56:06.639 --> 00:56:11.639 into your deck. The heat cards don't do anything, and 953 00:56:11.719 --> 00:56:13.920 you have to do specific things to get rid of them. 954 00:56:14.239 --> 00:56:17.880 That will often make you less efficient moving forward, and 955 00:56:17.920 --> 00:56:21.079 so you start figuring out how to get through as 956 00:56:21.119 --> 00:56:23.920 many turns as possible as quickly as possible so you 957 00:56:23.920 --> 00:56:27.199 can get to the straightaways and take off. The campaign 958 00:56:27.280 --> 00:56:30.159 mode is you play multiple races and you collect money 959 00:56:30.159 --> 00:56:33.519 and then you can upgrade your car board game. Geek 960 00:56:33.599 --> 00:56:36.320 rate waits it at two point one nine, which is 961 00:56:36.559 --> 00:56:43.440 pretty casual gamer ish as far as that goes. It 962 00:56:43.519 --> 00:56:47.760 says ages ten plus. I think somebody a little younger 963 00:56:47.760 --> 00:56:51.480 than that could play. The strategy is not terrible, and 964 00:56:51.519 --> 00:56:53.119 if you kind of help them. With the mechanics, I 965 00:56:53.119 --> 00:56:55.440 think I think you could get like a six or 966 00:56:55.480 --> 00:56:57.280 seven year old to play and that would be fine. 967 00:56:59.440 --> 00:57:01.199 I've played it with four players that plays up to 968 00:57:01.280 --> 00:57:04.719 six and it takes about an hour to play. So anyway, 969 00:57:05.360 --> 00:57:07.079 a lot of fun. This is. This is one of 970 00:57:07.119 --> 00:57:10.719 the favorites lately. In fact, on Board Game Geek. It 971 00:57:10.800 --> 00:57:20.280 actually is number forty one overall on the games. So yeah, 972 00:57:20.440 --> 00:57:24.000 really enjoying that. So I'm going to pick that. And 973 00:57:24.039 --> 00:57:28.599 then lately I've been watching a couple of shows and 974 00:57:28.639 --> 00:57:30.400 I think I've picked them over the last few weeks, 975 00:57:30.400 --> 00:57:32.000 but I don't remember, so I'm going to pick them again. 976 00:57:32.800 --> 00:57:38.039 The first one is nineteen twenty three and it's prequel 977 00:57:38.079 --> 00:57:42.719 to Yellowstone, and I'm really enjoying that. 978 00:57:44.559 --> 00:57:44.719 Is that. 979 00:57:44.719 --> 00:57:49.440 The one is Harrison Ford. Yes, yep, yeah, I've heard that. 980 00:57:49.639 --> 00:57:52.440 Him trying to swagger down the street like a cowboys 981 00:57:53.679 --> 00:57:54.880 not the best picture, but. 982 00:57:55.599 --> 00:58:00.679 Yeah, yeah, he's definitely old, but it's funny. 983 00:58:00.760 --> 00:58:02.800 He does a new uh, he does a new commercial 984 00:58:02.920 --> 00:58:07.480 for I said Jeep or land Rover, and the very 985 00:58:07.519 --> 00:58:10.800 last thing he says is, yeah, yes, I'm doing an 986 00:58:10.840 --> 00:58:14.679 ad for them. Even though my last name is Ford. 987 00:58:15.280 --> 00:58:17.280 A sort of funny guy. 988 00:58:17.280 --> 00:58:19.000 I thought, that's just me. 989 00:58:20.639 --> 00:58:23.559 So anyway, watching that, I'm also watching Reacher. I'm enjoying that. 990 00:58:25.119 --> 00:58:28.039 And then I'm about done with the book Rhythm of War, 991 00:58:28.239 --> 00:58:32.440 which is a Brandon Sanderson book. It's the fourth book 992 00:58:32.519 --> 00:58:35.559 in the Stormlight Archives. So every time he releases a 993 00:58:35.599 --> 00:58:38.679 new book, which takes like forever to listen to on 994 00:58:38.760 --> 00:58:42.960 Audible because it's you know, I mean, I've been listening 995 00:58:43.000 --> 00:58:44.440 to the Rhythm of War, I think for almost a 996 00:58:44.480 --> 00:58:48.079 month now, just because I listened. You know, I have 997 00:58:48.079 --> 00:58:49.760 a ton of time to listen, so I listen when 998 00:58:49.760 --> 00:58:51.920 I'm like trying to go to sleep or when i'm 999 00:58:52.519 --> 00:58:56.119 you know, out in the car or something. But yeah, 1000 00:58:56.119 --> 00:58:59.679 so he released wind in Truth and I just haven't 1001 00:58:59.679 --> 00:59:03.159 gotten yet. So anyway, Rhythm of War I'm really enjoying 1002 00:59:03.199 --> 00:59:08.280 as well. And then I think that's pretty much all 1003 00:59:08.280 --> 00:59:11.039 I've got for picks this time. But Matthew, what are 1004 00:59:11.079 --> 00:59:13.719 your picks? Okay? 1005 00:59:13.800 --> 00:59:18.199 So I love doing research on AI, so I'm always 1006 00:59:18.239 --> 00:59:20.440 trying to keep up on the newest models. It seems 1007 00:59:20.480 --> 00:59:24.519 like lately a lot of new audio like texts to 1008 00:59:24.599 --> 00:59:27.159 speech audio models have been coming out, which I've been 1009 00:59:27.199 --> 00:59:30.960 waiting for I had some ideas of how to kind 1010 00:59:30.960 --> 00:59:34.280 of produce some myself, but now with these coming out, 1011 00:59:34.440 --> 00:59:37.800 like I'm pretty excited to add that to speak magic. 1012 00:59:37.920 --> 00:59:42.039 So one of them that kind of hit that was amazing. 1013 00:59:42.599 --> 00:59:46.880 It's called sessing Me, and it brought a lot of 1014 00:59:48.519 --> 00:59:52.239 has a lot of emotion and kind of context, aware 1015 00:59:52.519 --> 00:59:58.320 of of of how to provide like in real time 1016 00:59:58.960 --> 01:00:03.840 like conversations. It just sounds a lot like a human. 1017 01:00:04.800 --> 01:00:06.599 There's been a few others that have come out has 1018 01:00:06.599 --> 01:00:10.760 recently been great too. One of them from open Ai. 1019 01:00:10.960 --> 01:00:14.960 If you go to OpenAI dot fm, it has a 1020 01:00:15.000 --> 01:00:20.719 way to be able to control like how like how 1021 01:00:21.039 --> 01:00:24.719 the the speech should be generated, so you can put 1022 01:00:24.760 --> 01:00:27.760 emotion in it, you can give it different things like 1023 01:00:28.000 --> 01:00:31.719 have a kind of accents or different things too. It's 1024 01:00:31.880 --> 01:00:36.440 pretty incredible as well. So those are kind of some 1025 01:00:36.519 --> 01:00:41.199 of the new AI models that have been pretty exciting 1026 01:00:41.719 --> 01:00:46.480 coming out recently. So see and then I brought up 1027 01:00:46.480 --> 01:00:51.679 before kind of looking more into open MCP, which is 1028 01:00:51.920 --> 01:00:56.079 seems like a using JavaScript a way to be able 1029 01:00:56.119 --> 01:00:59.199 to connect to different models and different other kind of 1030 01:00:59.199 --> 01:01:06.480 services as well. Is pretty exciting. Let's see, I've watched 1031 01:01:06.559 --> 01:01:08.639 some Reacher too, So I guess they have one more 1032 01:01:08.679 --> 01:01:10.679 episode left on that's next week. 1033 01:01:11.159 --> 01:01:14.719 That's been Yeah, it's definitely getting there where it's yeah, 1034 01:01:14.840 --> 01:01:15.679 it's gonna wrap up. 1035 01:01:15.719 --> 01:01:19.599 So it sounds like it's a lot closer to like 1036 01:01:19.679 --> 01:01:23.239 the series and like the movie, like the books basically 1037 01:01:23.280 --> 01:01:28.559 than the movie. I guess it's pretty cool. Yeah, that's 1038 01:01:28.639 --> 01:01:33.239 kind of that for me. But that's uh. 1039 01:01:35.519 --> 01:01:39.119 Yeah. I meant to throw out one AI pick and 1040 01:01:39.159 --> 01:01:41.480 I forgot about it, and that is if you want 1041 01:01:41.480 --> 01:01:43.480 to play with some of the large language models, especially 1042 01:01:43.559 --> 01:01:47.440 around text. I've been using an open router. I don't 1043 01:01:47.440 --> 01:01:50.400 know if you've used them. You can also get some 1044 01:01:50.440 --> 01:01:52.079 of them to run on your own machine if you 1045 01:01:52.079 --> 01:01:54.719 get them off a hugging face. So those are the 1046 01:01:54.760 --> 01:01:58.000 two resources that I'm going to recommend. One of open Router. 1047 01:01:58.360 --> 01:02:03.440 You can use their libraries to connect to them, and 1048 01:02:03.480 --> 01:02:06.960 then they connect to all the other models, so you 1049 01:02:07.000 --> 01:02:10.559 can try out the LAMA three model, the open AI 1050 01:02:12.239 --> 01:02:15.400 GPT models, you can try out Claude, and you can 1051 01:02:15.440 --> 01:02:18.480 switch between them so you can see which ones work 1052 01:02:18.559 --> 01:02:20.360 best without having to do a whole lot of extra 1053 01:02:20.440 --> 01:02:26.639 work to program against each one. And then hugging Face 1054 01:02:27.400 --> 01:02:29.119 they have their a whole bunch of other models and 1055 01:02:29.159 --> 01:02:31.119 you can run those all locally, and then I think 1056 01:02:31.119 --> 01:02:36.199 that's huggingface. Dot co is where you get those nice 1057 01:02:36.280 --> 01:02:38.559 all right, matt If people want to find your stuff, 1058 01:02:38.559 --> 01:02:39.840 where do they find you online? 1059 01:02:41.280 --> 01:02:46.119 Yeah, check out speak Magic ai. You can see kind 1060 01:02:46.159 --> 01:02:51.280 of we need to add some newer examples on there. 1061 01:02:51.320 --> 01:02:54.119 Our quality has gone up and with some of these 1062 01:02:54.119 --> 01:02:56.559 new models that come out, they'll be even better. So 1063 01:02:58.280 --> 01:03:03.440 you can check out Wiles at Wiles dot ai and 1064 01:03:04.000 --> 01:03:08.960 uh so, kind of interesting kind of looking at transition there. 1065 01:03:10.199 --> 01:03:12.639 So I'm trying to decide on how to take things 1066 01:03:12.840 --> 01:03:17.960 further with that, possibly maybe going to open source kind 1067 01:03:17.960 --> 01:03:22.440 of route or kind of make it more community driven. 1068 01:03:23.079 --> 01:03:25.920 So if you're interested with that, you can send me 1069 01:03:25.960 --> 01:03:30.719 a message at Matthew at Wiles dot ai. Yeah, that's 1070 01:03:30.800 --> 01:03:33.599 kind of kind of two kind of places for me. 1071 01:03:35.159 --> 01:03:38.000 Awesome. All right, Well, let's go ahead and wrap it 1072 01:03:38.079 --> 01:03:40.760 up here until next time, folks. Maxxed out