Build your first AI agent in 90 minutes

1 00:00:00,020 --> 00:00:01,760 Speaker 1: All right, we are bringing on Guil. 2 00:00:02,300 --> 00:00:06,540 Speaker 1: Guil has relatively recently released the first Mastra video course. 3 00:00:06,700 --> 00:00:09,660 Speaker 1: If you have been paying attention to Mastra, you know that a long time ago, 4 00:00:10,140 --> 00:00:11,800 Speaker 1: we built kind of this Mastra MCP course. 5 00:00:11,860 --> 00:00:16,219 Speaker 1: It was kind of innovative at the time, but it always lacked that appeal of the traditional course 6 00:00:16,240 --> 00:00:19,680 Speaker 1: where what's the best way to get started, and I want you to kind of walk me through it, 7 00:00:19,980 --> 00:00:23,199 Speaker 1: teach me the basics, hold my hand a little bit so I know that 8 00:00:23,200 --> 00:00:24,840 Speaker 1: I'm building that base of knowledge. 9 00:00:25,040 --> 00:00:28,180 Speaker 1: I think that's something that I always look for in a really good course, and I think we 10 00:00:28,300 --> 00:00:28,940 Speaker 1: really nailed it. 11 00:00:29,360 --> 00:00:30,780 Speaker 1: We're really excited to work with Guil on this. 12 00:00:30,900 --> 00:00:34,680 Speaker 1: Guil, do you want to give a quick introduction, and then we can definitely talk a bit more 13 00:00:34,820 --> 00:00:38,640 Speaker 1: about the Mastra course, what it covers, and how you went about building it? 14 00:00:38,780 --> 00:00:38,980 Speaker 2: Absolutely. 15 00:00:39,540 --> 00:00:39,840 Speaker 2: I'm Guil. 16 00:00:40,200 --> 00:00:44,900 Speaker 2: I've been working with the Mastra team for a few months now on the course course. 17 00:00:44,901 --> 00:00:49,660 Speaker 2: But yeah, I'm down in South Florida. I've been a developer and educator. I'd say most of my work 18 00:00:49,680 --> 00:00:56,739 Speaker 2: sits in that overlap. I've spent most of my career in developer education, working across many tech 19 00:00:56,940 --> 00:01:02,660 Speaker 2: companies in the past. Back in the day at Treehouse, Scrimba, done some content for LinkedIn Learning 20 00:01:02,980 --> 00:01:08,520 Speaker 2: and built courses, workshops, learning programs across front-end, full-stack, back-end, 21 00:01:08,920 --> 00:01:12,800 Speaker 2: Core CS and more recently AI and agent systems. 22 00:01:13,620 --> 00:01:19,200 Speaker 2: Overall, I spent a lot of my time building and teaching developers, beginners and curious learners how to use them. 23 00:01:20,500 --> 00:01:24,780 Speaker 1: I'm pretty sure I saw a Treehouse course from you a long time ago. 24 00:01:25,020 --> 00:01:27,060 Speaker 2: Probably a JavaScript course or React course. 25 00:01:27,480 --> 00:01:31,480 Speaker 2: It's funny, the first day I joined the Mastra's Slack, 26 00:01:32,080 --> 00:01:34,980 Speaker 2: Abhi was like, hey, thanks for teaching me JavaScript. 27 00:01:35,490 --> 00:01:38,860 Speaker 2: So I was like, whoa, the Abhi actually watched one of my courses back then? 28 00:01:38,860 --> 00:01:41,880 Speaker 2: I learned JavaScript from you, dude, as well as CSS basics. 29 00:01:42,110 --> 00:01:43,300 Speaker 2: I remember that one, too. 30 00:01:43,550 --> 00:01:45,640 Speaker 2: Yeah, happy to have you teach the next generation. 31 00:01:45,880 --> 00:01:47,680 Speaker 2: So tell us a little bit about the Mastra course. 32 00:01:48,240 --> 00:01:49,120 Speaker 2: How did you break it down? 33 00:01:49,380 --> 00:01:50,420 Speaker 2: Yeah, so you can check it out. 34 00:01:50,580 --> 00:01:53,320 Speaker 2: It's live at mastra.ai slash learn. 35 00:01:53,560 --> 00:01:57,800 Speaker 2: Essentially, it's a hands-on course where you build an agent from... 36 00:01:57,800 --> 00:01:59,680 Speaker 2: scratch all the way to deployment with 37 00:01:59,780 --> 00:02:01,480 Speaker 2: Mastra and we really wanted to focus 38 00:02:02,200 --> 00:02:03,120 Speaker 2: on how these things actually 39 00:02:03,960 --> 00:02:05,660 Speaker 2: work when you wire them up 40 00:02:05,840 --> 00:02:07,640 Speaker 2: clear up any ambiguity around agents 41 00:02:07,880 --> 00:02:09,580 Speaker 2: I know there's a lot of people out there hearing about 42 00:02:09,720 --> 00:02:11,380 Speaker 2: what agents are, what they do and 43 00:02:11,760 --> 00:02:13,360 Speaker 2: they just don't know where to get started or they 44 00:02:13,739 --> 00:02:17,720 Speaker 2: feel like the fear of missing out, right? Like they're not catching up. So this was, of course, 45 00:02:17,879 --> 00:02:23,819 Speaker 2: to provide a smooth on ramp onto that. So we start right with a basic agent. I'm from Orlando, 46 00:02:23,950 --> 00:02:28,640 Speaker 2: so I grew up around theme parks and I thought it'd just be fun and kind of simple to create 47 00:02:29,019 --> 00:02:32,420 Speaker 2: like a theme park agent. And we'll get into that a little bit more. But 48 00:02:32,940 --> 00:02:37,680 Speaker 2: Essentially, we define what an agent is, we create a simple agent, we start adding tools, 49 00:02:37,930 --> 00:02:40,860 Speaker 2: we connect it to MCP servers, right? 50 00:02:41,120 --> 00:02:44,820 Speaker 2: We make the MCP servers tools available alongside existing local tools. 51 00:02:45,520 --> 00:02:48,400 Speaker 2: We get into the basics of multi-step workflows. 52 00:02:49,040 --> 00:02:53,800 Speaker 2: We also introduce building with Mastra Skills along with the MCP doc server, but we do 53 00:02:53,980 --> 00:02:59,620 Speaker 2: We teach learners what Mastra Skills are, how to activate them and consume those in your project. 54 00:03:00,280 --> 00:03:05,460 Speaker 2: And then we get into memory, message history, working memory, the new observational memory. 55 00:03:06,220 --> 00:03:10,840 Speaker 2: And you see how the agent behavior changes as you go forward. 56 00:03:10,841 --> 00:03:15,000 Speaker 2: And for this course in particular, we run everything in Mastra Studio, right? 57 00:03:15,010 --> 00:03:19,340 Speaker 2: So you can get that immediate feedback loop and you can see it work. 58 00:03:19,450 --> 00:03:20,780 Speaker 2: You can inspect what happened. 59 00:03:21,230 --> 00:03:25,760 Speaker 2: You can use traces and observability and you can see all the tool calls, everything in 60 00:03:25,819 --> 00:03:26,340 Speaker 2: one place, right? 61 00:03:26,340 --> 00:03:27,000 Speaker 2: You don't have to work. 62 00:03:27,001 --> 00:03:32,700 Speaker 2: worry about adding a UI to it just yet. That might be another course. We use a very simple API, 63 00:03:33,140 --> 00:03:38,560 Speaker 2: the Queue-Times API, but it can work with like the Thrill Data and other more complex APIs. But 64 00:03:38,840 --> 00:03:43,040 Speaker 2: essentially just it looks up park data, right? You might say, you know, I'm planning to go to Epic 65 00:03:43,239 --> 00:03:48,299 Speaker 2: Universe today. What are the park hours? And then it extracts some data using MCP for the park hours, 66 00:03:48,480 --> 00:03:50,099 Speaker 2: and then it looks up current wait times. 67 00:03:50,420 --> 00:03:55,980 Speaker 2: It pulls live data and recommends, okay, what should I ride next? 68 00:03:56,049 --> 00:04:00,640 Speaker 2: Put an itinerary together based on the weather, based on the wait times, the park hours. 69 00:04:00,980 --> 00:04:08,180 Speaker 2: Also, it uses fire crawl to maybe go out and extract historical crowd data, like attendance, 70 00:04:08,659 --> 00:04:10,959 Speaker 2: or what are the typical wait times on a Tuesday? 71 00:04:11,489 --> 00:04:14,780 Speaker 2: Or what do you recommend that I ride in the morning? 72 00:04:14,829 --> 00:04:16,400 Speaker 2: I have kids, for example. 73 00:04:16,510 --> 00:04:20,239 Speaker 2: I don't like roller coasters, so we use memory to kind of keep track of those things. 74 00:04:20,240 --> 00:04:22,660 Speaker 2: those things and make suggestions based on that. 75 00:04:23,160 --> 00:04:25,560 Speaker 2: And we also tie in a workflow, right? 76 00:04:25,840 --> 00:04:30,880 Speaker 2: So we build a simple workflow that simulates like a mock ticket purchase. 77 00:04:31,000 --> 00:04:32,680 Speaker 2: And it walks you through the whole step of like, 78 00:04:32,940 --> 00:04:33,740 Speaker 2: is it the right park? 79 00:04:33,960 --> 00:04:37,860 Speaker 2: And how much does a ticket to like Animal Kingdom cost? 80 00:04:38,660 --> 00:04:44,340 Speaker 2: you know, and for what date and then it prints out an itinerary and some suggested things 81 00:04:44,520 --> 00:04:46,699 Speaker 2: to do that day after it confirms the purchase. 82 00:04:47,880 --> 00:04:50,400 Speaker 2: And then we teach deployment, right? 83 00:04:50,401 --> 00:04:50,501 Speaker 2: And, 84 00:04:50,699 --> 00:04:56,180 Speaker 2: So the idea is to teach how you can actually call that behavior and how Mastra can expose 85 00:04:56,500 --> 00:04:59,500 Speaker 2: all those via endpoints and you can consume in an app later. 86 00:04:59,860 --> 00:05:02,900 Speaker 2: So yeah, essentially how Mastra can become a HTTP server. 87 00:05:03,680 --> 00:05:10,460 Speaker 2: We're going to teach a one-click enablement of a Slack bot integration, so you can actually interact with it on the go via Slack, for example. 88 00:05:10,670 --> 00:05:18,660 Speaker 2: So we created it because agents, as I said, they're starting to show up everywhere, but it's not obvious what they are, how they actually work when you build one. 89 00:05:19,440 --> 00:05:29,599 Speaker 2: I think the goal from the start was like, make that concrete and provide that smooth on-ramp for those wanting to learn about and build agents and show what actually happens when an agent runs. 90 00:05:29,740 --> 00:05:42,099 Speaker 2: People to walk away with a clear mental model of what agentic behavior is, what it looks like in practice versus like a chatbot, for example, how an agent can decide what to do via tools, when it uses those tools, when to hand off to workflow. 91 00:05:42,100 --> 00:05:42,200 Speaker 2: Yeah. 92 00:05:42,820 --> 00:05:44,700 Speaker 2: and a sense of how to structure one themselves 93 00:05:44,980 --> 00:05:47,640 Speaker 2: and then just build on top of it using TypeScript, 94 00:05:47,840 --> 00:05:49,400 Speaker 2: the stack we all love. 95 00:05:49,640 --> 00:05:52,360 Speaker 1: The stack that we have learned to love over the years. 96 00:05:52,900 --> 00:05:53,700 Speaker 3: I was curious, Guil, 97 00:05:54,220 --> 00:05:57,780 Speaker 3: how does education in these concepts differ than 98 00:05:58,000 --> 00:06:00,280 Speaker 3: maybe what you were teaching before with React 99 00:06:00,740 --> 00:06:03,760 Speaker 3: or even lower level things like languages and stuff. 100 00:06:04,280 --> 00:06:07,000 Speaker 3: Is it a lot harder to teach or is it easier? 101 00:06:07,320 --> 00:06:08,420 Speaker 3: It's a little bit of both, it depends. 102 00:06:08,720 --> 00:06:11,360 Speaker 2: I also work at an ed tech company out in Austin 103 00:06:11,610 --> 00:06:13,900 Speaker 2: and there I started by leading a collaboration 104 00:06:14,180 --> 00:06:16,200 Speaker 2: on building an entire AI native community. 105 00:06:16,300 --> 00:06:18,600 Speaker 2: programming curriculum, right, that's used at universities. 106 00:06:19,080 --> 00:06:23,840 Speaker 2: And I've also been building AI grading tools and some some feedback mechanisms 107 00:06:24,660 --> 00:06:26,020 Speaker 2: to help review student projects. 108 00:06:26,240 --> 00:06:30,360 Speaker 2: But what's become clear after working with AI in education 109 00:06:30,760 --> 00:06:34,240 Speaker 2: is that the mechanics of it were never like the hardest part, right? 110 00:06:34,340 --> 00:06:37,860 Speaker 2: AI sure can handle a lot of that, the syntax, the boilerplate, 111 00:06:38,260 --> 00:06:39,760 Speaker 2: we're wiring everything together. 112 00:06:40,180 --> 00:06:42,420 Speaker 2: But the part that still matters is how you think through a problem, 113 00:06:42,720 --> 00:06:44,260 Speaker 2: how you might structure something. 114 00:06:44,740 --> 00:06:48,360 Speaker 2: how you test it, and most importantly, how you decide if it's actually good. 115 00:06:49,460 --> 00:06:54,460 Speaker 2: So my focus has been teaching how to use AI well. There's a lot of misuse right now. And 116 00:06:54,820 --> 00:06:59,360 Speaker 2: you know, when you're learning and experimenting, that's all good. But I think like good education 117 00:06:59,620 --> 00:07:04,140 Speaker 2: can really make AI an equalizer. And that also includes helping people adapt 118 00:07:04,319 --> 00:07:07,080 Speaker 2: without losing confidence in the sense of what they bring to the table. 119 00:07:07,460 --> 00:07:09,560 Speaker 2: And yes, I've seen students in the past, 120 00:07:09,580 --> 00:07:13,460 Speaker 2: it would take nine weeks to teach a student to build like a portfolio, right? 121 00:07:13,560 --> 00:07:18,620 Speaker 2: That maybe has like some JavaScript and carousels and bootstrap or tailwind. 122 00:07:18,740 --> 00:07:22,300 Speaker 2: But now like our capstone project is like the student builds an agent 123 00:07:22,539 --> 00:07:26,139 Speaker 2: that calls APIs and does a whole lot more than what they would have done 124 00:07:26,180 --> 00:07:27,919 Speaker 2: that whole semester in building out a portfolio. 125 00:07:28,099 --> 00:07:29,860 Speaker 2: So definitely expedites their learning. 126 00:07:29,919 --> 00:07:31,199 Speaker 2: And I think you can 127 00:07:31,320 --> 00:07:36,820 Speaker 2: get something quickly working right now, but a lot of students are doing more and more of that, 128 00:07:36,850 --> 00:07:41,940 Speaker 2: and it's starting to look a lot of the same, right? It works, but it's kind of shallow. It's kind of 129 00:07:41,990 --> 00:07:46,100 Speaker 2: the same. And after a while, I think the foundation still matters. Without some foundation, it's hard 130 00:07:46,100 --> 00:07:50,740 Speaker 2: to tell, like, all right, is my code actually holding up or does it just happen to work? You know, 131 00:07:50,740 --> 00:07:55,900 Speaker 2: the better you understand something, the better questions you can ask, the faster you can spot 132 00:07:56,139 --> 00:08:01,020 Speaker 2: bad output, the more you can steer it. And like I said, I think taste matters, right? You got to 133 00:08:01,020 --> 00:08:06,419 Speaker 2: know what looks good and not stopping at the first result and having a clear definition and teaching 134 00:08:06,560 --> 00:08:09,960 Speaker 2: students, all right, what does a clear definition of done look like? And another thing is actually 135 00:08:10,199 --> 00:08:14,520 Speaker 2: knowing when to maybe step in and take that handoff, if you will, from the AI and not just 136 00:08:14,669 --> 00:08:16,820 Speaker 2: over-iterate forever and ever and fall into that. 137 00:08:18,100 --> 00:08:36,260 Speaker 1: Yeah, I mean, I think I can definitely echo that with just talking with people that are a little bit more junior on the engineering side where you maybe don't need to understand syntax anymore. The day of needing to know that there's a needs to semicolon here and how to debug obscure missing semicolon issue or something in a language is 138 00:08:36,261 --> 00:08:37,620 Speaker 1: That's mostly gone, right? 139 00:08:38,180 --> 00:08:40,840 Speaker 1: But now it comes down to clarity of thinking. 140 00:08:41,040 --> 00:08:43,400 Speaker 1: Can you logically understand what's happening 141 00:08:43,780 --> 00:08:46,060 Speaker 1: and describe it in words to get what you want? 142 00:08:46,140 --> 00:08:47,780 Speaker 1: And I think one thing I've realized is 143 00:08:48,080 --> 00:08:50,620 Speaker 1: people are starting to realize that maybe writing the code 144 00:08:51,100 --> 00:08:53,540 Speaker 1: wasn't as big of a bottleneck as what people actually thought it was. 145 00:08:53,700 --> 00:08:55,400 Speaker 1: It is a bottleneck, right? It does take time. 146 00:08:55,520 --> 00:08:57,960 Speaker 1: But it wasn't just writing the code, it's the iteration loops 147 00:08:58,080 --> 00:09:00,700 Speaker 1: and the coming up with and making sure you have creative ideas 148 00:09:00,700 --> 00:09:02,520 Speaker 1: ideas for how to solve these actual problems. 149 00:09:02,800 --> 00:09:04,640 Speaker 1: And I think that now is where people spend 150 00:09:04,780 --> 00:09:06,700 Speaker 1: more of the time. And I think that is sometimes scary for people 151 00:09:06,800 --> 00:09:08,140 Speaker 1: too, because that's actually kind of hard. 152 00:09:08,420 --> 00:09:10,620 Speaker 1: You can lose yourself into, I have to move this code around, 153 00:09:10,720 --> 00:09:12,680 Speaker 1: I have to write this function, this is going to take me 154 00:09:12,700 --> 00:09:14,300 Speaker 1: some time, I'm going to burn half a day just 155 00:09:14,740 --> 00:09:16,200 Speaker 1: reorganizing this code. But at 156 00:09:30,019 --> 00:09:30,519 Speaker 2: Yeah, absolutely. 157 00:09:30,970 --> 00:09:33,459 Speaker 2: And in learning itself, there's less friction. 158 00:09:33,920 --> 00:09:40,460 Speaker 2: You can try things out quicker, get feedback faster, but people are spending less time on long tutorials. 159 00:09:40,820 --> 00:09:44,720 Speaker 2: One of my favorite teaching approaches is the Socratic way of learning. 160 00:09:45,080 --> 00:09:49,300 Speaker 2: So one thing I like is that how I really supports that Socratic way. 161 00:09:49,301 --> 00:09:54,260 Speaker 2: approach, right? You can stay in that loop of asking questions, testing your ideas, refining 162 00:09:54,560 --> 00:09:58,340 Speaker 2: what you think you know, and being more critical in your learning. So you're actively working 163 00:09:58,510 --> 00:10:00,900 Speaker 2: through things, and AI really makes that loop faster. 164 00:10:01,140 --> 00:10:06,920 Speaker 3: When we first started Mastra, we started doing our education pieces. And people at that time, 165 00:10:07,180 --> 00:10:11,640 Speaker 3: let's just say, they were still writing code. Sonnet 3.5 was out, and people were still writing 166 00:10:11,880 --> 00:10:18,020 Speaker 3: code. Or maybe they had code assistance through the AI agent. Since then, now it's like, you know, 167 00:10:18,200 --> 00:10:22,740 Speaker 3: Maybe most things are fully through your coding agent and stuff like that. 168 00:10:22,840 --> 00:10:27,420 Speaker 3: But people kept getting tripped up with these concepts of AI engineering, 169 00:10:28,120 --> 00:10:29,940 Speaker 3: or when should I do this or that. 170 00:10:30,340 --> 00:10:33,040 Speaker 3: Have you also seen students like, oh, what is a workflow? 171 00:10:33,380 --> 00:10:34,920 Speaker 3: Why do I want to use one? 172 00:10:35,160 --> 00:10:36,100 Speaker 3: What is an eval? 173 00:10:36,360 --> 00:10:37,540 Speaker 3: Why is it called an eval? 174 00:10:37,680 --> 00:10:40,620 Speaker 3: Are there questions like that, which seem more foundational, 175 00:10:40,700 --> 00:10:42,140 Speaker 3: but the foundation keeps changing? 176 00:10:42,520 --> 00:10:43,940 Speaker 3: as the industry is progressing. 177 00:10:44,600 --> 00:10:46,880 Speaker 2: I had those same questions not too long ago. 178 00:10:47,640 --> 00:10:50,920 Speaker 2: There's a video in the actual course, the Mastra course, 179 00:10:51,100 --> 00:10:55,380 Speaker 2: that walks through the differences between agents and workflows 180 00:10:55,820 --> 00:10:59,280 Speaker 2: and when you might use one versus the other or combine them together. 181 00:10:59,650 --> 00:11:03,220 Speaker 2: But I have seen that the gap between people who just generate output 182 00:11:03,350 --> 00:11:05,340 Speaker 2: or the people who can actually shape it and work with these tools, 183 00:11:05,460 --> 00:11:06,400 Speaker 2: it's been widening. 184 00:11:06,650 --> 00:11:08,700 Speaker 2: You still need to know what you're building in the end. 185 00:11:09,040 --> 00:11:14,120 Speaker 1: Who would you say is the intended audience, right? It's probably not for someone who's 186 00:11:14,200 --> 00:11:18,000 Speaker 1: already shipped an agent into production in maybe another language and is just getting 187 00:11:18,180 --> 00:11:23,000 Speaker 1: started with Mastra, right? It's probably someone a bit more new to AI engineering in general. 188 00:11:23,200 --> 00:11:26,780 Speaker 2: It's for anyone who's curious about agents, but it's more, yeah, it's geared more towards 189 00:11:26,780 --> 00:11:31,580 Speaker 2: that audience, right? Maybe you're a TypeScript developer, yeah, you want to ship an agent, 190 00:11:31,860 --> 00:11:32,900 Speaker 2: right? Or maybe you've... 191 00:11:32,900 --> 00:11:35,120 Speaker 2: built agents using other languages 192 00:11:35,400 --> 00:11:36,920 Speaker 2: and want to learn how to get up to speed 193 00:11:37,060 --> 00:11:39,180 Speaker 2: with building one in Mastra in TypeScript. 194 00:11:39,420 --> 00:11:40,860 Speaker 2: So it's for everyone, really. 195 00:11:41,200 --> 00:11:43,000 Speaker 2: But the goal was to 196 00:11:43,390 --> 00:11:45,140 Speaker 2: help someone who's never built an agent 197 00:11:45,360 --> 00:11:47,200 Speaker 2: before build their first agent and ship it 198 00:11:47,600 --> 00:11:48,820 Speaker 2: and then just build on top of it. 199 00:11:48,940 --> 00:11:50,560 Speaker 1: I do feel like we naturally 200 00:11:51,080 --> 00:11:52,200 Speaker 1: gravitate towards people who 201 00:11:52,600 --> 00:11:54,360 Speaker 1: are kind of like us in some ways. 202 00:11:54,720 --> 00:11:57,880 Speaker 1: I think, Guil, with your courses on React and CSS 203 00:11:58,259 --> 00:12:00,959 Speaker 1: and web development, and Abhi and I have web development 204 00:12:01,160 --> 00:12:03,199 Speaker 1: backgrounds coming from Gatsby, 205 00:12:03,380 --> 00:12:06,360 Speaker 1: where I think if you're that kind of person, 206 00:12:06,459 --> 00:12:09,399 Speaker 1: this will resonate really well if you're just getting into AI engineering. 207 00:12:09,579 --> 00:12:12,220 Speaker 1: And of course, if you're just interested in agents in general, 208 00:12:12,220 --> 00:12:14,000 Speaker 1: I think you'll find value of it, but 209 00:12:14,120 --> 00:12:24,180 Speaker 1: Specifically, if you have built websites with JavaScript or TypeScript before, this is probably a really good on-ramp for getting started with what is an agent and why you might want to build one. 210 00:12:24,380 --> 00:12:31,400 Speaker 2: Yeah, you bet. It's been fun trying to keep up with all the changes and the things you're quickly shipping in Mastra, but... 211 00:12:31,640 --> 00:12:37,280 Speaker 2: Yeah, even that, I'm making sure that it's holding up and that there's nothing really out of date yet. 212 00:12:37,560 --> 00:12:43,440 Speaker 1: There is something inherently valuable, in my opinion, around being able to relate to a person that's teaching you something, right? 213 00:12:43,640 --> 00:12:45,720 Speaker 1: You have the actual teacher that is there. 214 00:12:45,720 --> 00:12:46,600 Speaker 1: I can see the person. 215 00:12:47,040 --> 00:12:47,740 Speaker 1: They can teach it. 216 00:12:47,780 --> 00:12:48,920 Speaker 1: It feels like they're talking to me. 217 00:12:49,260 --> 00:12:51,860 Speaker 1: But video is a little harder to keep up with. 218 00:12:51,980 --> 00:12:54,080 Speaker 1: And AI makes things better. 219 00:12:54,081 --> 00:13:00,519 Speaker 1: It's easier to build features and code, so the number of features we're shipping is probably higher now than it would be two or three years ago. 220 00:13:00,610 --> 00:13:03,680 Speaker 1: And it was always kind of hard to keep video courses up to date. 221 00:13:03,790 --> 00:13:06,180 Speaker 1: What have you seen others doing? Do you have ideas? 222 00:13:06,600 --> 00:13:09,319 Speaker 2: No, it's definitely been challenging, so I try to think about it more. 223 00:13:09,620 --> 00:13:14,079 Speaker 2: In discrete, sort of composable pieces, not having too many videos dependent on the other. 224 00:13:14,740 --> 00:13:23,380 Speaker 2: But I think what's really helped with this Mastra course is once we introduce Mastra Skills, I encourage the learner to lean on that, right? 225 00:13:23,560 --> 00:13:26,680 Speaker 2: Maybe something that you're doing that I'm doing doesn't quite line up. 226 00:13:26,860 --> 00:13:34,760 Speaker 2: Well, you know, ask Mastra Skills like, all right, give me the latest information about like this workflow or why this tool isn't working. 227 00:13:34,870 --> 00:13:42,380 Speaker 2: So once you have that knowledge, I think maybe with some, all right, you can provide a little context in the notes or whatever. 228 00:13:42,381 --> 00:13:43,000 Speaker 2: in the code itself. 229 00:13:43,420 --> 00:13:45,500 Speaker 2: But I think once you know that, 230 00:13:46,080 --> 00:13:48,180 Speaker 2: I don't have to be backed into this corner. 231 00:13:48,180 --> 00:13:50,060 Speaker 2: I can take full advantage of Mastra Skills 232 00:13:50,240 --> 00:13:53,040 Speaker 2: or MCP docs server and immediately get up to speed 233 00:13:53,160 --> 00:13:53,760 Speaker 2: with the latest. 234 00:13:54,060 --> 00:13:56,060 Speaker 2: But yeah, in traditional content creation, 235 00:13:56,300 --> 00:13:57,800 Speaker 2: it's a little bit more challenging. 236 00:13:58,080 --> 00:14:01,160 Speaker 3: We'll just have to make a Guil agent, just teach all the classes. 237 00:14:01,450 --> 00:14:02,780 Speaker 3: We can give him a voice. 238 00:14:03,870 --> 00:14:05,740 Speaker 3: Instead of a real person, we'll make him a cartoon. 239 00:14:08,360 --> 00:14:13,900 Speaker 1: Yeah, the Guil agent cartoon trainer that keeps up to date. 240 00:14:14,120 --> 00:14:14,940 Speaker 1: That's the future. 241 00:14:15,320 --> 00:14:16,380 Speaker 1: We will see. 242 00:14:18,040 --> 00:14:18,620 Speaker 1: All right, Guil. 243 00:14:19,100 --> 00:14:20,180 Speaker 1: It was great having you on. 244 00:14:20,300 --> 00:14:21,440 Speaker 1: We appreciate you coming in, 245 00:14:22,120 --> 00:14:22,860 Speaker 1: talking about the course, 246 00:14:23,140 --> 00:14:24,580 Speaker 1: mastra.ai/learn, 247 00:14:24,920 --> 00:14:26,200 Speaker 1: talking a little bit about just AI 248 00:14:26,440 --> 00:14:27,600 Speaker 1: and education in general. 249 00:14:27,760 --> 00:14:30,279 Speaker 1: I have a kind of education ed tech background as well. 250 00:14:30,320 --> 00:14:31,480 Speaker 1: It's kind of fascinating just how 251 00:14:31,820 --> 00:14:34,380 Speaker 1: how it's changing, but it's still the same, right? 252 00:14:34,600 --> 00:14:35,440 Speaker 1: People need to learn things. 253 00:14:35,660 --> 00:14:37,320 Speaker 1: People like to learn things from people. 254 00:14:37,680 --> 00:14:39,060 Speaker 1: And I think it's just a matter of like, 255 00:14:39,170 --> 00:14:41,000 Speaker 1: we now have different tools maybe to help 256 00:14:41,640 --> 00:14:42,839 Speaker 1: make that process even faster. 257 00:14:43,260 --> 00:14:44,360 Speaker 1: So appreciate you coming on. 258 00:14:44,620 --> 00:14:46,160 Speaker 1: We're excited to have you back on, you know, 259 00:14:46,220 --> 00:14:49,000 Speaker 4: for the Mastra course version two or updates along the way. 260 00:14:49,279 --> 00:14:50,660 Speaker 4: And we'll see you next time. 261 00:14:51,160 --> 00:14:51,980 Speaker 4: Appreciate it, Shane. 262 00:14:52,589 --> 00:14:52,880 Speaker 5: Take care. 263 00:14:53,040 --> 00:14:53,540 Speaker 4: Yeah, thanks, Guil. 264 00:14:54,220 --> 00:14:54,839 Speaker 5: That guy's a legend. 265 00:14:55,190 --> 00:14:56,399 Speaker 5: Just want to throw that out there.