WEBVTT NOTE Sentence-level transcript of https://www.youtube.com/watch?v=IddXPepIAS4 NOTE One cue per sentence. Cue ids are the line anchors on /transcripts/IddXPepIAS4.html. A cue ends where the next begins, or 2 s after its last word. s1 00:00:00.000 --> 00:00:09.360 Hey everyone, my name is Eric Hanchett, and today I want to tell you guys all about spec-driven development and how you might use it in your workflows today. s2 00:00:09.360 --> 00:00:11.160 So, there's a couple of things we want to talk about. s3 00:00:11.160 --> 00:00:14.760 First, what is spec-driven development? s4 00:00:14.760 --> 00:00:18.280 And then, how can it make you a better and faster coder? s5 00:00:18.280 --> 00:00:25.600 So, we just don't want you to code faster, but we want you to actually create higher-quality code from it. s6 00:00:25.600 --> 00:00:31.240 And I think really spec-driven development helps you with this, and I'm going to explain why and how. s7 00:00:31.240 --> 00:00:35.280 It's a structured specifications are created before any code is written. s8 00:00:35.280 --> 00:00:43.720 That's the definition of really the specs or the requirements of this spec-driven development, and it goes to the heart of what it is. s9 00:00:43.720 --> 00:00:54.520 So, what we're doing is we are writing these markdown files, we're writing the specifications and the design document before any of the code is written. s10 00:00:54.520 --> 00:00:59.640 And this actually works really, really well with large language models and our coding assistants that we use. s11 00:00:59.640 --> 00:01:02.800 And I'm going to show you exactly how that works. s12 00:01:02.800 --> 00:01:05.120 Now, a question I sometimes get is why? s13 00:01:05.120 --> 00:01:11.440 Like, why should we use this spec-driven development workflow and then how can we do it? s14 00:01:11.440 --> 00:01:22.880 So, to answer that why question, I kind of like to think about our coding assistants, our large language models that we're using every day as sort of like AI interns. s15 00:01:22.880 --> 00:01:28.720 You need to really prompt them, you really need to push them the right way, no pun intended with the prompting. s16 00:01:28.720 --> 00:01:31.240 So, you really need to guide them. s17 00:01:31.240 --> 00:01:37.400 Because if you give them just a little bit of of leeway, they will go off the rails. s18 00:01:37.400 --> 00:01:41.520 And really the spec-driven development helps guide them in the right direction. s19 00:01:41.520 --> 00:01:49.880 I remember when I was an intern in one of my first jobs, we had a VP that would walk around and he would just come up with random ideas s20 00:01:49.880 --> 00:01:50.800 off the top of his head. s21 00:01:50.800 --> 00:01:54.400 And when he told me it, I would drop everything and work on it. s22 00:01:54.400 --> 00:01:58.160 And then later my boss would be like, "Hey Eric, why did you drop everything?" s23 00:01:58.160 --> 00:02:00.520 Well, the the VP said something. s24 00:02:00.520 --> 00:02:09.240 And then I learned about, well, you got to take his information that he gives you, write it down, put it in your schedule, talk to me talk to my manager about it. s25 00:02:09.240 --> 00:02:10.759 So, I was a little too eager. s26 00:02:10.759 --> 00:02:14.880 And this is what AI interns or basically these large language models do nowadays. s27 00:02:14.880 --> 00:02:18.080 And we got to be careful about it. s28 00:02:18.840 --> 00:02:22.000 I didn't really give you too much of introduction, so I'll give it to you now. s29 00:02:22.000 --> 00:02:23.320 My name is Eric Hanchett. s30 00:02:23.320 --> 00:02:26.959 I'm a senior developer advocate at AWS. s31 00:02:26.959 --> 00:02:31.400 I have over 15-plus years of software development experience. s32 00:02:31.400 --> 00:02:34.760 And if you want to connect with me, you can check out LinkedIn. s33 00:02:34.760 --> 00:02:37.880 That's probably the best place to go. s34 00:02:37.880 --> 00:02:41.240 I just look for Eric Hanchett and yeah, talk to me, follow me. s35 00:02:41.240 --> 00:02:44.440 I'd love to talk to you more about it. s36 00:02:44.440 --> 00:02:53.400 Another question when I that people often ask me about spec-driven development is, "Can't the latest frontier models just do everything for us already? s37 00:02:53.400 --> 00:02:55.080 Why do we need this?" s38 00:02:55.080 --> 00:03:04.840 And while I do say that the models are getting better and better every year and and every release, sometimes incrementally, sometimes by leaps and bounds, it's still not perfect. s39 00:03:04.840 --> 00:03:13.480 And giving the large language model these models more context is actually really important because it does help guide it in the right direction. s40 00:03:13.480 --> 00:03:23.680 And with our software always changing and with the requirements always changing, and with new paradigms and shifts, you always need to kind of guide it where it needs to go. s41 00:03:23.680 --> 00:03:34.400 Some large language models, some harnesses are starting to think more about this and adding in a little bit more of a thinking or planning mode before it jumps into when it writes code. s42 00:03:34.400 --> 00:03:45.120 But there's nothing better than actually having those documents created and having you be in the middle before it jumps into the net that next part with it creating code. s43 00:03:46.200 --> 00:03:49.920 So, let's keep a few things in mind before we get too far. s44 00:03:49.920 --> 00:03:53.160 There is quite a lot of information out there. s45 00:03:53.160 --> 00:03:54.360 I have some opinions. s46 00:03:54.360 --> 00:03:56.160 These are my opinions. s47 00:03:56.160 --> 00:03:57.480 Uh let me know. s48 00:03:57.480 --> 00:04:06.280 Connect with me on LinkedIn if you agree or disagree, but I think you need to keep these in mind before using spec-driven development or vibe coding or anything in between. s49 00:04:06.280 --> 00:04:09.040 We have to You have to be careful with context. s50 00:04:09.040 --> 00:04:17.720 Now, I told you before that spec-driven development gives you a lot of context that gets fed into that large language model, but sometimes you have too much of a good thing. s51 00:04:17.720 --> 00:04:27.880 Uh what you need to do is when you are working with these large language models, usually you create like an agents.md or claw.md at the beginning with some information in it. s52 00:04:27.880 --> 00:04:32.080 I would be very careful not to put too much information or too little information in that. s53 00:04:32.080 --> 00:04:35.200 Kind of like that Goldilocks zone of information is what you need. s54 00:04:35.200 --> 00:04:40.919 So, you just need to give it enough rules and guidelines that it knows what to do. s55 00:04:41.680 --> 00:04:49.000 Uh we call it steering docs in Cara, which I'll mention about in a minute, but you just got to be very careful what you put in those documents. s56 00:04:49.800 --> 00:04:54.320 I also highly recommend when you're using spec-driven development to use skills. s57 00:04:54.320 --> 00:05:01.160 Skills are like instruction files that you can give to your coding agents that are ran on demand. s58 00:05:01.160 --> 00:05:10.280 So, usually they have keywords in them, and when the coding assistant sees those keywords, they'll activate the skill or you can do {slash} and the skill name and then run it. s59 00:05:10.280 --> 00:05:13.840 But, this actually really helps when you're doing the spec-driven development process. s60 00:05:13.840 --> 00:05:18.560 You can actually add skills um when it creates your design documents or whatever else. s61 00:05:18.560 --> 00:05:25.680 You can use those in parallel or with the spec-driven development or when you actually implement the tasks. s62 00:05:26.640 --> 00:05:29.760 And just be careful about giving too much trust. s63 00:05:29.760 --> 00:05:38.680 Uh remember everything in the spec-driven development flow is you, meaning that you are the human in the loop, so to speak. s64 00:05:38.680 --> 00:05:43.640 You need to be the code reviewer of all the code that's generated through this process. s65 00:05:43.640 --> 00:05:56.960 You need to be interactively looking at the design documents and the requirements documents that are created because at the end of the day, if something goes wrong, you are the person that are going to be blamed for it, not the agent. s66 00:05:58.880 --> 00:06:03.400 And uh that's not to say that we don't use tools to help do code reviews. s67 00:06:03.400 --> 00:06:07.560 Someone once told me like, are you saying you're the human in the loop, you do the code review, no one else? s68 00:06:07.560 --> 00:06:20.760 Of course, we might have multiple be doing people doing code reviews and definitely use your normal AI coding assistants and and review tools to help you review any pull request that you create. s69 00:06:20.760 --> 00:06:24.200 Or CRs. s70 00:06:25.160 --> 00:06:32.240 So, let me give you a little bit of a history lesson of where we have been, especially at AWS. s71 00:06:32.240 --> 00:06:42.480 We uh saw this need, especially from our teams and our customers, that they were vibe coding, they were using this coding assistance, but they weren't getting the outputs that they wanted. s72 00:06:42.480 --> 00:06:57.480 And what we saw is this these customers were using this bespoke pattern of having the agents, the coding assistants, create uh these full down requirements documents and these design documents and we decided to create s73 00:06:57.480 --> 00:07:02.600 a application to kind of do this all for them. s74 00:07:02.600 --> 00:07:08.720 So, we released Qiro uh early late last year in a general availability. s75 00:07:08.720 --> 00:07:11.440 It's our new AI IDE coding assistant. s76 00:07:11.440 --> 00:07:16.200 We also have a command line interface or CLI version, which is just as popular. s77 00:07:16.200 --> 00:07:18.680 I think right now it's switching. s78 00:07:18.680 --> 00:07:21.800 Most people are starting to use CLIs more often than IDEs. s79 00:07:21.800 --> 00:07:25.400 So, if you like to do that, you can use the CLI. s80 00:07:25.400 --> 00:07:32.240 But, what we really focused in on with Qiro is, you can see in these screenshots, is this vibe and spec mode. s81 00:07:32.240 --> 00:07:40.960 So, we're going to jump into spec mode in a little bit, but we really thought this really encompass what we were hearing from our customers, what people wanted more s82 00:07:40.960 --> 00:07:51.600 ways to work with larger features and more complex projects and they really needed the coding system to know exactly what their project was doing. s83 00:07:52.280 --> 00:07:55.240 We also released the website at kiro.dev. s84 00:07:55.240 --> 00:07:57.800 Feel free to after this presentation to go check it out. s85 00:07:57.800 --> 00:08:04.000 You can download our CLI, our IDE, and give it a whirl cuz you know, we're really proud of it. s86 00:08:04.000 --> 00:08:04.440 It's fun. s87 00:08:04.440 --> 00:08:07.040 When we released this, it actually went a little bit viral. s88 00:08:07.040 --> 00:08:08.560 We got tens of thousands of downloads. s89 00:08:08.560 --> 00:08:23.200 A lot of people were asking us asking us about Kiro so we were really happy to to release this and we actually when we set it up in preview, we had so many downloads we had to put a gate in front of it and then people found a around the gate, but now it is publicly available for s90 00:08:23.200 --> 00:08:27.280 everyone to try and and and give it a shot. s91 00:08:28.160 --> 00:08:32.200 But I don't want this to be a half-hour pitch for Kiro. s92 00:08:32.200 --> 00:08:37.360 You can do this process spec-driven development without Kiro and there's a few few ways. s93 00:08:37.360 --> 00:08:42.680 One is that you basically have to just you can do this manually with anything. s94 00:08:42.680 --> 00:08:47.120 You could tell the your assistant, your large language model to include the following. s95 00:08:47.120 --> 00:08:49.600 First is the user requirements. s96 00:08:49.600 --> 00:08:54.920 You would tell them please based on XYZ create a whole set of user requirements. s97 00:08:54.920 --> 00:09:07.000 Now, you can feed it your own user requirements beforehand and say, okay, use this as a basis or you can have the editor or the AI IDE go ahead and create it for you. s98 00:09:07.000 --> 00:09:09.280 Then you look back and forth and make sure it's okay. s99 00:09:09.280 --> 00:09:14.240 Then you tell the assistant to create the design document. s100 00:09:14.240 --> 00:09:18.360 Then you look back at that and then you have it create the implementation details. s101 00:09:18.360 --> 00:09:25.800 These are the the tasks list, the tasks one by one to actually implement it based on the requirements and the design document you created. s102 00:09:25.800 --> 00:09:30.320 So you could do this kind of a manual way if you like. s103 00:09:30.760 --> 00:09:32.160 I'll give a shout-out to Spec It. s104 00:09:32.160 --> 00:09:36.040 This is GitHub's open source way of doing this process. s105 00:09:36.040 --> 00:09:40.600 You can install it in a variety of different coding assistants. s106 00:09:41.720 --> 00:09:45.120 Now, in Kiro, you have this lab This is the IDE version. s107 00:09:45.120 --> 00:09:51.800 We have plan mode inside the CLI if you prefer the CLI, but in the IDE you would choose spec. s108 00:09:51.800 --> 00:09:54.040 And then you would go ahead and give it a prompt. s109 00:09:54.040 --> 00:10:01.680 So, in this case I could say like build a movie MCP server to keep track of movies I've watched or build an a movie website, whatever you want. s110 00:10:01.680 --> 00:10:07.760 Now, you're probably thinking is this spec only good for greenfield brand new projects? s111 00:10:07.760 --> 00:10:09.120 And I would say absolutely not. s112 00:10:09.120 --> 00:10:16.360 I've seen existing apps that are years old have dozens and dozens of these different spec files in them. s113 00:10:16.360 --> 00:10:22.320 And this example, it's kind of more of a greenfield, but you can definitely use it for existing legacy projects as well. s114 00:10:22.320 --> 00:10:24.960 It thinks it works very well. s115 00:10:24.960 --> 00:10:35.280 And I would just say that these are really good for when you're doing in-depth features, when you need your project needs a little more upfront planning, uh and you just want to build things in a structured way. s116 00:10:35.280 --> 00:10:38.840 For definitely works really well with complex projects. s117 00:10:38.840 --> 00:10:41.760 We actually added a spec mode for bug fixes, too. s118 00:10:41.760 --> 00:10:48.160 So, you can use it for smaller things, though you may want to just kind of vibe code those things instead. s119 00:10:49.240 --> 00:10:52.200 Here's what our requirements document looks like. s120 00:10:52.200 --> 00:11:01.320 So, in the first part it creates this uh requirements phase where it's in this ears format and it gives you the introduction, like the requirements, the user stories. s121 00:11:01.320 --> 00:11:05.000 Uh we also have a process where it asks you questions beforehand. s122 00:11:05.000 --> 00:11:08.400 And so, we added in some extra question and answer. s123 00:11:08.400 --> 00:11:14.160 So, as you put your prompt in, it'll ask you some clarifying questions before it creates this requirements phase. s124 00:11:14.160 --> 00:11:23.160 You can also do this quick plan mode we just added, which uh will go ahead and just generate all these documents quickly for you based on the questions and answers you give. s125 00:11:23.160 --> 00:11:26.320 So, there's a lot of different ways you can do this. s126 00:11:26.320 --> 00:11:29.400 Then you'll create a design document, which is a higher-level design document. s127 00:11:29.400 --> 00:11:31.640 You can also start with the design document first. s128 00:11:31.640 --> 00:11:35.040 Some people kind of take this design and requirements, combine them. s129 00:11:35.040 --> 00:11:40.160 But typically with with our spec-driven development flow, you can either start one or the other. s130 00:11:40.160 --> 00:11:44.880 And then you might have mermaid diagrams here, ASCII art with all the information. s131 00:11:44.880 --> 00:11:54.520 Now, this is at the point I would highly recommend, if you're trying this at home, to stop and go in and update it with your knowledge and expertise and taste s132 00:11:54.520 --> 00:11:56.600 to exactly what you're looking for. s133 00:11:56.600 --> 00:11:59.520 Because it's only as good as what you put in. s134 00:11:59.520 --> 00:12:07.080 But you'll find out that it even with a smaller prompt it and with the clarifying questions, it does a pretty good job. s135 00:12:07.200 --> 00:12:13.520 And course you always want to review it for markdown the markdown files for inconsistencies, hallucinations, and errors. s136 00:12:13.520 --> 00:12:15.320 And then finally you get to the implementation phase. s137 00:12:15.320 --> 00:12:18.560 This will be a list of tasks. s138 00:12:18.560 --> 00:12:26.080 Uh it also has something called property-based tests in them, which are tests that are against the requirements document and design document. s139 00:12:26.080 --> 00:12:29.120 You can additionally have those created and ran. s140 00:12:29.120 --> 00:12:35.800 I would highly recommend it, so that way it makes sure that the the tasks actually are implemented correctly. s141 00:12:35.800 --> 00:12:46.560 I usually also, uh quick tip here, once it creates this task list, I tell it, "Please take the top four tasks, put them at the top, and create an MVP for me first." s142 00:12:46.560 --> 00:12:48.560 So that way I can actually see it working. s143 00:12:48.560 --> 00:12:54.120 And then I'll implement them I'll implement the MVP version first. s144 00:12:55.720 --> 00:12:57.640 Now, I love model context protocol. s145 00:12:57.640 --> 00:13:05.480 I'm not going to go into exactly all the details of it, but it's a way essentially for your model providers to connect to different data sources. s146 00:13:05.480 --> 00:13:14.240 So, in this case, I'm using Kiro, and maybe I'm using like OpenAI or Anthropic, and we want to connect to different data sources. s147 00:13:14.240 --> 00:13:22.680 So, we have this MCP protocol in the middle, and it connects everything together for us, which is awesome. s148 00:13:23.000 --> 00:13:27.800 Now, I get this question, "Isn't MCP 6 months old and it's dead now?" s149 00:13:27.800 --> 00:13:38.120 Now, a lot of people on the internet have opinions that change every other day, and some people were saying that, well, you can do a lot of stuff with MCP just with command line tools. s150 00:13:38.120 --> 00:13:39.960 I really think MCP is still maturing. s151 00:13:39.960 --> 00:13:46.480 There's a lot uh a long road ahead for it, especially with some of the security stuff it's doing. s152 00:13:46.480 --> 00:13:56.080 I would keep an eye out for MCP, but one of the main reasons I like it, especially for the Spec Driven Development Flow, is you can have something like in Jira or Asana, s153 00:13:56.080 --> 00:14:08.280 all your tickets, all your uh large requirements docs, and then you can have it being pulled into your Spec Driven Development Flow when it creates your specs, which makes it much easier to work with. s154 00:14:08.280 --> 00:14:15.360 So, if you have a product manager that has actually written a requirements document, you can pull that in. s155 00:14:16.120 --> 00:14:17.720 And uh it just makes a little bit easier. s156 00:14:17.720 --> 00:14:21.640 And I'm I'm talking about any any kind of project management service. s157 00:14:21.640 --> 00:14:25.160 And grab technical details for it. s158 00:14:25.160 --> 00:14:35.800 You can also put additional rules in your steering or your agents.md files that tells it, so if you're using Spec Driven Development Flow, make sure you grab the information out of this MCP server. s159 00:14:35.800 --> 00:14:37.240 And that way it knows where to look. s160 00:14:37.240 --> 00:14:43.880 Or you can just specify it in the first step, like, "Hey, look at this MCP server when you create the requirements." s161 00:14:45.800 --> 00:14:47.960 So, let me show you a quick demo. s162 00:14:47.960 --> 00:14:53.640 This is the latest version of Kiro, the IDE version. s163 00:14:53.640 --> 00:14:56.000 And let me give you a a quick show of it. s164 00:14:56.000 --> 00:15:03.839 Now, just for the sake of time, I don't have time to actually go in and create a brand new project from scratch, but I used this one. s165 00:15:03.839 --> 00:15:05.320 It's a movie database. s166 00:15:05.320 --> 00:15:09.560 So, this is more of a green field, brand new project, but I think you'll still get the idea. s167 00:15:09.560 --> 00:15:10.640 I was in the spec mode here. s168 00:15:10.640 --> 00:15:14.320 I said, "Please create me a movie website." s169 00:15:14.320 --> 00:15:17.240 So, what it did is it started with this design document. s170 00:15:17.240 --> 00:15:21.520 And this case In this case, I actually told it to start with design instead of requirements. s171 00:15:21.520 --> 00:15:27.240 And then this shows the actual preview of the markdown file. s172 00:15:27.240 --> 00:15:32.400 So you can see it created a bunch of different mermaid diagrams that show the architecture of it. s173 00:15:32.400 --> 00:15:33.800 Gives you sequence diagrams. s174 00:15:33.800 --> 00:15:39.080 And so I went back and forth with this and updated and edited and it created this all for me. s175 00:15:39.080 --> 00:15:47.280 If I scroll down to the bottom of this document it created, you can see here it created a bunch of property test tests. s176 00:15:47.280 --> 00:15:48.760 And these are really special tests. s177 00:15:48.760 --> 00:15:58.920 It uses fast check in this TypeScript in the node world to do these tests which actually run dozens if not hundreds of times with different values to make sure that s178 00:15:58.920 --> 00:16:04.080 it's uh these requirements are satisfied correctly, which is really nice. s179 00:16:04.080 --> 00:16:12.320 And then it created this requirements document in once again this EARS format, so it's really uh very simple with these user stories, requirements. s180 00:16:12.320 --> 00:16:14.680 Uh let me see if I can open it up again. s181 00:16:14.680 --> 00:16:16.320 Here's it in preview. s182 00:16:16.320 --> 00:16:18.480 Uh so here's the requirements document. s183 00:16:18.480 --> 00:16:20.480 Requirements one, movie data loading. s184 00:16:20.480 --> 00:16:23.040 When the application initialize, the filter engine shall load. s185 00:16:23.040 --> 00:16:29.080 So it kind of gives you step-by-step exactly what's uh what it's doing here. s186 00:16:29.080 --> 00:16:35.080 And then when I went back and forth with this, I got to the task list and I asked it, "Hey, in the task list, can we create like an MVP?" s187 00:16:35.080 --> 00:16:45.280 So it went ahead and updated the task list and even says right here, "Task one through four deliver a working MVP, browsable movie grid with search, genre filtering, sorting, s188 00:16:45.280 --> 00:16:47.400 and the full 80s synthwave theme." s189 00:16:47.400 --> 00:16:56.920 So it actually redid all the requirements and to get this uh basically MVP up and running out of the door, which I really enjoyed. s190 00:16:56.920 --> 00:17:01.040 I went ahead and just implemented all of them after that, but you could see how it works. s191 00:17:01.040 --> 00:17:08.400 You can also see it created this property paste based tests, which are nice which I can go back in and take a look at the execution. s192 00:17:08.400 --> 00:17:11.720 You can also, let me see if I scroll to the bottom here. s193 00:17:11.720 --> 00:17:12.480 Here it is. s194 00:17:12.480 --> 00:17:18.439 So if I just kind of hover, it shows like a little bit of information about the property-based test. s195 00:17:18.439 --> 00:17:26.199 Like with property test test for genre filtering for any movie data set and if the extracted genre list must contain exactly a sorted set of unique genre. s196 00:17:26.199 --> 00:17:28.280 So, it tells me exactly what it did at this point. s197 00:17:28.280 --> 00:17:34.360 So, if you want to learn more, you can go to kiro.dev or you can join our Discord at discord.gg/kiro.dev. s198 00:17:34.360 --> 00:17:40.640 I highly recommend everyone watching check it out and make sure you connect with me on LinkedIn at Eric Hanchett. s199 00:17:40.640 --> 00:17:47.760 Love to just connect with as many people as I can and also on Blue Sky at Eric C H or X. Thanks.