Welcome And Guest Setup

Tim 0:13

Hello and welcome back to another episode of the Cables to Clouds podcast. And uh with me as always is my co-host, Chris Miles. I am Tim McConaughey, and we have a guest with us uh from Learning at Cisco, uh Jason Belk. Jason, go ahead and introduce yourself. I don't think you've been on the podcast before.

Jason Belk 0:31

Hi. Thanks, Tim. Thanks, thanks, Chris. Yeah, I am Jason Belk. I am a senior technical advocate with Learned with Cisco. Um, the people who do the training and certification, CCNA up to CCIE and everything in between. And uh a little bit about myself. I live in Reno, Nevada, married 12 years, have two young daughters, a black lab, and uh my background is in I've been at Cisco since 2010, off and on. And I I started out at Cisco IT as an analyst, network engineer, network automation, started teaching networking automation before DevNet to my our networking team, learning it myself, and then I left and went to network to code for a few years. Um, came back to Cisco as a in developer relations in the Cisco NSO BU, which is like basically a big automation product, and then now more recently in learning with Cisco. So kind of done around the circuit inside and outside of Cisco and have some operational chops to show for it. So I kind of know the struggles of all the operators out there, uh at least for the on-premise stuff and the enterprise. I was on the campus and branch team when I was on call. Um, so yeah, and and I'm really thrilled to be here. I mean, we've had conversations with with you guys off and on for a while now, and big fan of the show and love what you guys do both for the news perspective as well as covering, you know, all the different stuff for network engineers who are trying to grapple with AI and cloud and everything else. Thanks, man.

What Spec-Driven Development Fixes

Tim 1:45

Um, so yeah, yeah, what's that? Oh, I just said I appreciate that. Oh, appreciate that. Yes, we are appreciative. Uh it's always nice to meet someone who actually listens to the show. I'm just kidding. Um anyway, so we brought actually we brought Jason on uh because he is working on something that's actually pretty cool. Uh we've been we've been talking it back and forth for a long time, actually. We've been having trying to get uh Jason on the show, so we finally made it work. Um and Jason is here to talk about something that I didn't know existed, and I guess that probably a lot of the listeners were not aware existed. Uh something called spec driven development. And uh I believe it has something to do with with building for AI, but actually, Jason, I'll you're the expert. I'll let you I'll let you take it from here. What is spec driven development?

Jason Belk 2:29

Sure. I mean, I've been using it for about six weeks. Uh and AI speak that makes me an expert. Um that's right. Uh to be fair, no, like one I actually asked Claude Coder or today how much I'd been using it just on one project, and it said 150 hours in the past three weeks. Um project, one one personal project. Um and I've and I I have like five or six personal projects going on top of some work stuff. And and so all that to say is what is spectrum development? I mean, if you think if you've been in IT or any amount of time, you're familiar with project managers. They're those friendly people who help keep you on track and get all your T's crossed and your I's dotted. And if you if you've tried to build anything of significance, you know it helps to have a plan. And so when you're talking about an AI, you you I think we've all heard of the example of AI being, you know, like an expert idiot or something. Like they have tons of knowledge, tons of ability to execute, but what I would say is very poor wisdom and judgment. And so how do how do you help somebody who is able to do whatever you say and sometimes interprets it a little too in a direction you weren't expecting? Is that that you give them a plan. And there's definitely this isn't the only way of approaching that problem. And this isn't for um, you know, I'm on chat GBT, Jim and I, Claude, you know, I'm just doing my chat bot and I'm talking back and forth. I'm not talking about that scenario. The the way that I'm I'm talking about specifically is is for a quote unquote agentic coding, um, where where you are giving prompts, instructions to your computer through a text interface, in my case, usually a terminal, and you're you're typing in just as you would on ChatGBT or Claude or Gemini, and instead of getting a response back that says, hey, I can't do that, sorry, it actually goes off and tries to do it using your computer resources. Um so you have to have permissions and trust in the place where it's running for it to execute that code. And by default, something like Claude Code, which is one I'll be referencing in the rest of this conversation just for sake of simplicity, um, will ask you permissions when it wants to create files, when it wants to do web searches, when it wants to edit edit or delete files, do bash commands, stuff like that. But all that to say is that it can do all those things for you. And not just from a coding perspective. Um, there's another podcast I really enjoy called Software Defined Talk. And they had a whole episode about cloud code I'd recommend people check out. If you haven't heard about it before, played around with it before, it gives you a good primer of what I would say somebody who's not a coder or an active coder using cloud code just for productivity stuff, like sorting your screenshots in your screenshots folder and naming them, um, or uh, you know, putting together your notes and things like that. Because it can basically once you give it access to those files, and that's of course a trust issue, it will then be able to do instructions based on renaming, reading, and all that kind of stuff. So going back to spec driven development is that okay, you you give this a local AI agent the capability to do things on your behalf. And if you start asking to do things, as I think we've people who have worked with AI for a while will experience there's a context window issue. And I'd say there's also like a scope issue. And like I meant to do this, but actually you started doing this other thing, and then that's not too bad, but actually you were kind of deviating from what I originally talked about, or you built a production grade system, and I was just trying to figure out if it worked. Um, and or most common one I still even run into, even with spec driven development sometimes, is that you're trying to pip install this on my system Python, and I want to use UV. Um, so spec driven development, not to say it solves every possible problem, but I've gone from like having to intervene and course correct my quad code multiple times per session to like maybe once a week. And I can just let it run for anywhere between 30 seconds and a minute. I've had it sometimes running up to several hours.

Agentic Coding With Claude Code

Chris 6:12

So it's funny you mentioned that episode of Software Defined Talk. I just listened to that last week and it was uh it's it's it's funny because like they that was like my motivator was like, man, I I really need to dig in with clawed code because they even call it out in the episode, but like the fact that it's called clawed code kind of limits what you think in your brain that you can do with it because you can do very vast things with it. But um, yeah, I like I like what you said about the the idea of you know spec driven development kind of being that specification that you define up front to kind of keep your AI in line. But I'm I'm curious, like uh I feel like I've I've heard of spec-driven development uh uh you know several times, obviously very recently as well. Um, was this the thing that just came along with AI almost as a not guardrails, but kind of like guiding light for AI um coding? Or was this there before?

Jason Belk 7:05

I'd say that the principles, um, and and I'm I'm we'll share some links in the show notes that people can take a look at. Um the principles I think go back to project management principles in terms of defining requirements and and and executing executing upon them, uh, you know, clarifying assumptions uh in a more formal way. Uh, but I'd say, at least from my knowledge of it, I've been following it since August of 2025. I think that's around the time it came out. Um, because I I follow a lot of different YouTube channels. One of them is like the GitHub, you know, AI folk, and and they had some Microsoft Developer Day back then where they were walking through it. Um and so the the the one that I use is the GitHub, what they call spec kit, and it's on GitHub for free. I'm I'm also sharing, um, I'll share a link to my personal spec-driven development kind of fork of it, um, with the caveat that the cut the code I have in there that is the spec driven development part is uh is not like updating with the remote fork. I haven't got that complicated on it yet. So if you want to have the latest GitHub spec kit part, you might have to swap that part out and then use the rest of my bits I have in there. All that to say is um it's it's still relatively recent, as of August. And then I I tried to get it set up myself a couple times, but I'd say the documentation out there for me as an intermediate level programmer, novice level AI person who is figuring stuff out as long as I for the past since 2023 or whatever, since GPT 3.5. Um I I I had a couple of stops and starts until this guy named Jason Gekki, who who's um I I'm connected with who's people at work pointed me to some um documentation that he does. He basically he created a presentation that is happy to share with whoever it would help. And he's using this a lot, his startup and and consulting. And so I'm gonna share the link with that as well, which I think just helps visualize the flow and how all the pieces fit together. Because if you just go to the GitHub repo, and this is where I got stuck, I'm like, okay, so what do I do with this? I I see we have a repo and I see I have clawed code, but like, where do the dots connect?

Chris 9:01

Yeah, for sure.

Jason Belk 9:02

And and the first couple of times I tried, I kind of gave up before I got through his stuff. Um, not through it in a bad way, but like before I I I went through his his the way he explained it, and then I've tried to explain it as well on Cisco U and a tutorial we'll talk we can talk about a little bit later, is just kind of like connecting that dots of like, okay, how do I actually use this? And like it's not it's really not as mystical as it sounds. It's and once you I I even went through a demo of this with one of my friends from church, and he's he's like he has some ideas on on a potential business idea, and so I I shared my screen and my laptop on the computer, and I literally did the spec spec flow with him for his idea. He was defining the requirements verbally to me. I would just type in whatever he said, and then we literally built a product in like 45 minutes.

Spec Kit Origins And Key Links

Tim 9:41

So help me out here because um I'm new to the spec driven thing. Sure. Uh the spectrum, so you know, I've been messing around with cloud code, I've done some things, I've failed at more things probably than I've succeeded at. Uh I tried to make a news agent for the show and I posted that about that on Substack. Um if you know, me being knowing as little as I do, but and and knowing, you know, being dangerous with it, I supposedly. How do I how would I, as a person who's messing with it, getting better but not great, um integrate? Or where do I even start trying to pull in spec driven development into like my own workflow to build stuff?

Setup Gotchas And Slash Commands

Jason Belk 10:19

Sure. I I would say start with a greenfield project, something that like you had an idea. It's you're not trying to boil the ocean, like I need I need to make the next Facebook app or something. But if you're an infrastructure engineer, you're like, okay, maybe I I'm not super comfortable with Ansible and I want to deploy some Terraform as well. And like, okay, uh, or I know Terraform really well and I don't know Ansible. And so it's like you have an you have a context area that you already kind of understand, but it's kind of rusty and you have to look up a lot of docs. So you're able to define the requirements reasonably well. So I'd say, like, like brainstorm, okay, this is what I kind of want to do. And you don't have to get it all right in the first shot. This is more just kind of like to make sure set you up to succeed. Now I'd say check out the links that we were we set up because it's it's something that I'll talk through, but it's almost more important for you to look through and do it yourself with your fingers on the keyboard. Um, because it it's it's really not rocket science. Like once you have on one screen the flow and the other screen your terminal, and like you have it set up. Okay. The the installation process is basically just a bunch of bash scripts. It's not like installing any new software, it can kind of run independently too, just as it's a bunch of markdown files and bash scripts that kind of like guide the quad code to do the the project management workflow. And so the the the GitHub repo that I have for for my um kind of additional context, I'll just say some additional instructions on the spec kit flow, which is optional, or the core one that you can get from the official Microsoft GitHub. Um you can basically get clone down the spec kit. I'd say the one thing I learned the first time I did that I did it incorrectly is that I initialized my clawed code at the wrong directory because I was kind of new to Claude Code, where I was like, okay, uh I'm I have my project folder, make dira project, and then I put spec kit in that in that subfolder, and then I ran clawed code at that layer above where the spec kit was installed. So it didn't look because these are technically what are called claud skills, and it doesn't it doesn't load in the skill if you're not in that directory when you initialize claude code, which it didn't tell me. And I mean like I I went all that to say is like if you're doing it, just make sure initialized when you type in claude on your terminal when you have it installed, do it in the same place, the same layer of the directory where you have the spec kit installed so it can load in those skills.

Tim 12:19

Oh, okay, I see. So so I mean you're pulling down a repo, like you said, basically, and and and in that repo is gonna be like skill file, like files essentially that that define all the skills for that you're gonna use for the for the workflow that you're gonna follow on the spec driven development. Okay, that that makes it a little bit more real to visualize. Okay, I see.

Jason Belk 12:38

And you can inspect them, and it's not like some binary that you can't ever touch. You can look at the bash scripts, you can look at the the markdown files that are a part of it. Um but basically as soon as you load up that clawed instance and it loads in those skills, you can then access on slash commands so you to to load it to use those skills. So you go, you know, like slash spec kit dot, and then the first one would be constitution. And if you do it in the wrong order, it'll kind of like gently correct you. It's like, oh, you're trying to implement, you don't have a plan yet.

Constitution Through Implement Loop

Tim 13:05

You want to have a plan it's a workflow, it's yeah, it's a whole it's like uh uh waterfall. Yeah, yeah. Okay, yeah, that makes sense.

Jason Belk 13:13

Yep. So so the the the first step, and this is one that I'd say it doesn't always enforce, but it's good to start with because it kind of gets you in the right headspace and helps it up farther down the road, is the constitution. So just like in the United States and other countries, they have a governing body that defines the principles of what should be executed. Uh so in in this case, it's between I think three and five. Generally, it suggests. And and it's not that you have to walk in with like a quill and like some parchment and and and and have it all worked out before you start. Like Hancock on there.

Chris 13:44

Yeah, get your drawing Hancock.

Jason Belk 13:46

That's right. I actually use um this is kind of like a JSON hack that I started recently, one of my friends pointed me to as Whisper Flow. This is different than spec hit development, WISPR flow. It allows you to do voice to text real easily. So you have to have a hotkey, and you can do that prompting into your terminal. So it makes my prompts so much better on Claud Code, where I'm not like getting carpal tunnel DM VPN on my on my Claud Code. I just hold down option, that's the key I set up, and I just talk for like 30 or 45 seconds, brainstorming what I want to do at the consultation at the constitution where it'd be more at first principles like, okay, yeah, I'm trying to do a in this case Ansible Terraform app. I kind of I know Ansible really well, but I don't know Terraform so well. Um my my resources are all gonna be public cloud, AWS, Azure, and like my keys are gonna be in this.m file or in my you know my my ZH S HRC and all that kind of stuff. So you just kind of like give it the lay of the land and kind of what direct what what direction you're heading. And I just kind of record 34 to 5 seconds, paste it in there, and I'm done. And and then it processes through there and uses the the LLM. And that's a one other key thing for me is not to use sonnet, but to use Opus. When I did spec driven development, like it default. I mean, I think quad code defaults to Sonnet, which is fine for like you know setting simple stuff, but you can do slash model and switch to opus. And yeah, you're gonna use up your tokens a lot faster, but it's like cry once, try once, or whatever, and like because otherwise it was just spinning in circles with Sonnet. It's like, oh sorry, I broke something. Let me fix it. Oh, I broke something else. Let me fix it. And so I was like, with Opus, it still breaks things on occasion, like once a day for me, or and what stuff like that. But I'd just say, even if you have less Opus tokens based on your pro or whatever level, just wait till it refreshes. Yeah.

Tim 15:30

Yeah. So this is interesting because when I was working on my agent, um, I was trying to save save tokens. So like I was using um I used haiku, which I was like, oh wow, this is a really bad idea. It's a really bad idea. I realized it after the fact, after I was building the code for it, I was like, wait a minute, I shouldn't use haiku to build this. But what haiku is good at is writing. So maybe from like this part of it, this constitution defining like doing the writing part of it, would that be better just to get it all in the mod? Like get it because all you're doing at this point with spec I think, and correct me, is at this point, really what you're doing is creating the files that clause refer to uh refer to for the development part of it, right?

Jason Belk 16:12

Yeah, yeah, I'd say I haven't tried that. It's a good idea. Like in terms of the flow, I'd say definitely when you get to the implementation layer, you want to switch to over. But you could probably get away with it on Sonic and Haiku on the on the initial ones. I I haven't done like A-B testing to see what the difference is, but I think for most, I think on the implementation level, that's definitely for your own.

Tim 16:29

Oh, it's definitely no doubt, no doubt. I can tell you haiku is terrible at implementation, not accident. I can tell you that for sure.

Jason Belk 16:35

That's a good idea. Um, and so yeah, for once you get your constitution. So I I do my brain dump, it responds back basically, usually like, hey, I've noticed based on what you said, or if you didn't give enough information, I'll ask them follow up questions. Um, based on what you said, here's three to five principles that I'm seeing. Like that you you want to have simplicity as a design, you want to have cloud native as a design, you want to have these you're doing this as a proof of concept rather than a production build. So like it kind of extracts out or clarifies for you what type of project you're trying to build. And this doesn't just have to be for coding. Um, this could be for for other stuff too. Like that's but all all that to say is like what then it creates from the that response, then it generates your constitution file for you. So I'm not like going in there handcrafting. You can you can go in there and edit the file, but it basically writes it for you. And then you go to specify. So once you have your constitution going, you flow through through these this loop of this workflow for each of your features. So you go from specify to plan to tasks to implement, back to specify to plan to tasks to implement for each of your features, basically. Um, and and and I've found for bigger projects, oftentimes my first spec is like an architectural one where its output is not code, but just like architectural documents, um, where I'm like, okay, let's specify what cloud we're gonna be using, what you know, what logging we're gonna be using, what um all this kind of stuff. Where uh if it's a much bigger, if you know it's gonna be a big project, that'd be my recommendation because sometimes it just it just makes decisions and does even with the spec kit flow is gonna start making some assumptions on things that you might have to go back and rework later, which isn't in the world, but like that'd be my one recommendation for a bigger project. But if you're just getting started doing a small thing like this, it doesn't matter. I would say you're probably just gonna have maybe two or three specs tops. Um so the first the first one would be, you know, like the first one that flows. So we have the constitution, and then you go to and then you do the slash command slash spec kit, and it'll have like tab complete if that skill is you know um loaded in there correctly. And then it sometimes really it'll even suggest, like at the end of its constitution, like, okay, we have the constitution already. Do you are you ready for spec kit specify and like copy and paste? Yep. Um and and and so then for spec kit specify, you're documenting your requirements. So going back to that project management example, you're sitting down with the project manager, and they're like, Okay, so what's what's the problem? Like, well, we have a new sales office and we need to get access in for salesforce.com. And they we we have an app we'll have to take down the network over the weekend in order for this this network change to take place. Um, and no, we're gonna do it on the Cisco iOS cat9 gates and Cisco Catalyst Center or something. So you're just defining your your requirements in terms of what you're trying to accomplish for this particular feature.

Tim 19:12

Yeah, that makes sense. I mean, you're setting up the guard essentially the guardrails, like what are you going to be working with? What are you gonna be doing, right?

Chris 19:19

Like I think uh I think one thing that I've definitely observed is it's amazing to me how fast Claude and Anthropic have kind of switched to the front runner in this in this conversation. Um it's it's been you know specifically over the last it seems like over the last month like it's really really picked up steam. Um but one thing I've also noticed is the the kind of you know kind of variety of tools that people are using. Like you said, you're using something specifically for voice to text, and there's integrations and there's people's building skills and using different MCPs and things like that. So maybe can we can we take a step back and let's talk a little bit about skills and because I know with with spec driven development, um it seems one of the kind of concrete things is is the idea of because you've well defined your requirements and your intent up front, this allows for kind of parallel development, right? You have this you have agents working on one piece of it and agents working on another piece of it, and they can operate autonomously, right? And they won't conflict with each other because they all are working off of the same requirement, right? Correct. Um, so can you tell me a little bit how agents and skills play into that?

Skills, Guardrails, And Model Choices

Jason Belk 20:32

Yeah, I I I'm by no means an expert on um skills, but I'll my humble explanation is that they're just a series of bash scripts and markdown files, primarily markdown files, that um help the model, usually Claude, understand what parts of its training and brain to use in that particular context. So just like if you're on ChatGBT and you say, pretend you're an expert network engineer and analyze this ACL. Okay. Um it's it's it's it's it's it's it's funny because like on the software-defined talk podcast, they actually even talked about like if you if you look at like the market share that's been destroyed by a bunch of markdown files, they're like, it's tremendous. They're like, you know, you're destroying accounting, it's destroying legal. It's like you look at what actually is destroying, it's just it's just like some markdown files on these repos. It's not like they they did anything super special, it's just like it it's such a powerful model, it just needs like little guardrails and like some to say, okay, just focus on this part and take on this identity. Um yeah.

Tim 21:29

So actually, real quick, because uh you just reminded me of something, and I I this is more of an opinion thing because like I don't think anybody knows what's gonna happen in the next six months or three months or two months with AI, these things. But isn't this this feels like uh, you know, what I'm hearing more and what I truly believe probably will be the the result, the end result here is as we seek to make these things more efficient, and as training becomes things that, you know, we have enough material uh hardware to start doing more training, specialized training. Do you think that we're gonna gravitate towards more like small, you know. Small specialized models that are going to be built kind of for this, or unless, or do you think it's going to be like, oh, well, here's an agent, we're going to put the guardrails in place, and then we're just going to have the LOM still back at all?

Jason Belk 22:11

I think it's going to, my opinion, just the way I've seen other stuff work, you're not going to know or care as a user. Sure, no, 100%. But but but like from a technical perspective, um Yeah, that's what I'm thinking. I I I I can definitely see that being in the approach. I mean, that's why I work at Cisco, like they released a security um L um special trained model. Yeah, a small language model for that last June on Hugging Face. Because uh yeah, like I think it's also just not feasible to maintain these models. I mean, I might be wrong, I'm not an expert by any means, to like try to pretend like the corpus of every possible job in existence that will exist and that will exist, could be maintained by something that or would want to be maintained, even if we could, it's like, well, we could we could use all of the power from Idaho to like power this thing, or or or we could just run it on your run on your phone.

Tim 22:59

And so that's kind of my feeling too. Like, so I I think that's probably where we're going to be headed once once we get more of the of the process of being able to truly kick crank out these small language models or specialized language models, depending on how you want to talk about it.

Jason Belk 23:14

Yeah, I think it's right now such a new thing for so many people and so disruptive to people's workflows and such an open form thing. Or previously, with like cloud, you're like, okay, I have to get like my AWS keys and like you know, go onto the console and like figure out what I want to do here, or or even from a SaaS perspective, like okay, I need to like get my credit card and like get this SaaS thing plugged into my work workflow. And that there was still like some complexity there and like some changes to people's existing paradigm. They're like on-prem versus cloud, hybrid, you know, all that kind of stuff versus like LMs just blows everything out of the water. I was like, what if you could just like talk to your computer like in Star Trek and it could like answer? Right. And it's just like I I I I mean, I'm not trying to like trivialize anything about this, but I think that's in my mind one of the foundational differences that it's so open form and free form that it's hard to train people on like what to do. You're like, oh, I you mean I can have it help me cook dinner or build out ACLs on my router. Okay. Um, but I guess we'll do that. Um, so I I I think that's my mind what the adoption curve is gonna is is is gonna eventually look like is that it's gonna get more adopted as you get more specialized bottles where people can understand their use case. They're like, oh, this is my security model, I'll ask it security questions, even though they could probably do that right now.

Tim 24:29

Um yeah, yeah, yeah. But efficiency-wise, of course, you gotta when you every time the ask and LM a question, you're lighting up like half the planet to to get the answer. Um you said something earlier also about the um oh about the uh oh man, I just lost my train of thought, of course. Darn it, I I had it for a second and then I just lost it. I give up. Uh Chris, uh I'll head it over to you. I was I was kidding.

Jason Belk 24:54

I was I was put I was put in the spot for no reason for that.

Tim 24:57

Sorry.

Context Windows And Written Plans

Jason Belk 24:58

I I can keep going through this stuff that's uh so so we we we covered constitution, we covered specify. So after specify, you have the planning phase. So you've defined your requirements, you've defined your governing principles, and then now you have your requirements, you go per feature designing your technical approach. And and and this would be just the next flash command. So after it does the specify, and that might take several minutes or longer in terms of the specifying the constitution. So you gotta be patient, go to the bathroom, take the dog for a walk, um, and then come back and then do the spec kit plan. And you can do these in multiple tabs. Like I'll sometimes have three or five different spec kits running on different projects at the same time, which just makes you feel like superhuman. Um and so, like spec kit plan per feature design technical approach, and that that's basically like, okay, now that we have our requirements, like what's gonna be our approach for the Ansible Terraform example? It's like, okay, you know, like what modules are we gonna use or for Ansible and and what um providers are we gonna use for Terraform? And this isn't something that like you have to define that out because like it based on the LM's knowledge, and if it has access to the internet, it can figure that stuff out. So, like this plan is generally, in my experience, generated from the your main input is on the specify, and then the plan it builds a plan for you. You're not like okay, I have to spend half an hour figuring out my plan. It's like, hey, based on your requirements, this is why I think a good plan would be based on your governing principles from your constitution. And you're like, that is a good plan. Let's do that one, or you can go in there and you can clarify, and this is for the voice to text input, I'd be like, I'd be like, no, actually, like I decided not to use Ansible. We're we're we're gonna use Puppet and Chef instead. And it'd be like, okay, I guess we need to go, we we need to rechange everything. And so, like, you you can still part way through change your mind and do different things, but or maybe a more realistic change to the plan would be like, okay, we're gonna have this on specific jump posts that you need to know about and and and and stuff like that, having more details on your on on what is gonna affect the plan.

Tim 26:49

Actually, you just reminded me, just reminding me what I lost my frame of thought on, which is what we were talking about, and and I think everybody who's probably now put any amount of time into AI is figuring has already figured this out, but just for setting purposes, um, the more you know about the technology you want to use, the easier it is to get the AI to to under because of course, because you can say, like, oh, I know you're gonna need this information, right? I know you're gonna need to know about this, I know you're gonna end up like if you know Ansible, like if you know enough about Ansible, you know exactly what that you gotta tell make makes it into the spec. Yeah, you you know exactly what needs to go in there.

Jason Belk 27:26

Yeah, yeah, you know what the gotchas are. And so this is where I think people who don't have a technical background use AI, do kind of get caught up or get stuck or get in rabbit holes. And so this is like exactly to your point, like your advantage as a technologist in defining these things is that you can front load your assumptions on all the things you wish, as if you were talking to yourself from a mentor version of yourself where you've forgotten everything, but you also you're like this is yeah, or a junior, a junior engineer, and so you need to clarify your assumptions. And and this is kind of my thing that I've realized over the past year, year and a half is that like your leadership skills, your soft skills as a human being, make your ability to execute prompting and planning like this uh significantly better. So I'd say brush up on some books on leadership if you have nothing to do and you're bored on the technology side, and it's gonna help you be a better AI user in the future because it gives you things like clarifying assumptions, you know, defining requirements and all these things that like project managers and people with with you know technical leadership backgrounds are all used to or consultants oftentimes too, where you have to sit down and define your workflows. Um it's uh it's all in your head somewhere where you you need to get it on paper, kind of assuming nothing.

Chris 28:40

Yeah, I think um one thing I've definitely noticed in working with AI for specifically this type of uh this type of use case is uh I'm very good at giving context, and I often give a very large amount of context, which as you guys talked about, it really uh it really messes with your you know your windowing uh availability when you're interacting with AI, so it's like burnt me a few times because I'm like I'm like no, this is important, you need to know this. So I do like this idea. Yeah, I do like this idea of being able to define that up front and that just using that as a reference source over and over again. Um because yeah, I mean, but then there's tools as well out there, like it we should we should put a link to that software defined talk episode in the show notes because it's so valuable to me. Um but there's um they referenced that there's even tools out there that you can use that actually remove all the fluff from any of your prompts that like can't will not, you know, so that you're not eating up tokens that's cost. You might be saving them. Yeah, exactly. Like you might be using them uh, you know, maybe 10% less, but that 10% can be you know that's valuable for you down the down the road, right?

Jason Belk 29:52

And I'd say if you have other documents, you can reference those as long as you're you're willing to give access to those files to this system. Right. So like for one of my projects, I had done some work in a planning phase, not even before I even knew about Spectre Development. I I had like a 20-page Microsoft Word document that I I had defined basically requirements. It was the first half of it was governing principles, and the second half of it was execution lines, not like code, but more like defining tasks and things. Yeah. Um, and I p I found PDFs to be the longer franca of of AI, and so I exported the doc. I think you can read doc X, but I always just do it as a PDF just to make sure the formatting's right. So I export as a PDF, included that in my constitution specifying that like it was really good at that, because it's like it it extracted all the important details and kind of like filled out the stuff for me on that stuff. So if you already have a project plan from like an actual project manager, and they're like, okay, okay, Chris, we need we need you to to build out this cloud deployment. Here here's your here's your project plan. You're like, okay, I'll export as a PDF, and if my company lets me use these types of tools, I could just use spectrum development to extract out and build out the plan from their plan.

Tim 30:58

Yeah, that makes sense. And I don't run into that.

Jason Belk 31:01

Oh, go ahead. I was just saying once so once you one thing is that it sometimes your plan might have ambiguities where e either um based on what you're trying to implement or the different technologies and outcomes you're trying to achieve. And I'm trying to think of a good one. Um, but an example might be if you weren't specific on a cloud. If you're like, okay, I want to do Ansible and Terraform, but you never really said if it was AWS, Azure, or something else. And it will oftentimes auto-suggest in at the end of its response, like, do you want to go to the next phase or do you want to do spec kit for clarify? And you can still run it, even if it doesn't suggest it, and it might say, Nope, there's nothing just to clarify, but uh sometimes it will it will actually go through a clarify phase where it goes through all the tasks, I mean all the plan that it's built and say, Here's my plan. There are actually some ambiguities on some of these plans, and then it will it'll do between one and five questions that are multiple choice, or you can fill in your own answer and and it and and allows you to clarify. So it's like like, for example, those questions you'll say, like, I noticed you know our plan we're we're we're we're not saying which cloud we're using. Do you and it'll kind of know because it's smart in terms of like, did you mean to use AWS A, B, Azure, C, you know, GCP, or D, something else? Fill it in. And it doesn't have to be a one-word answer. You could I can say, well, I I I I have an Azure account, but I prefer AWS, and so it'll be like, okay, no, let me update the plan.

Tim 32:20

Yeah. I like to use the uh I mean, I this is uh obviously spec driven is much more uh not just complicated, not just complex, but obviously very, I assume in the weeds with it. But I mean I've been using the planning mode of Claude quite a bit. Um for the it for the same for the same reason, right? Create me a file because I know what's gonna happen is that I'm gonna run out of context when you can compact the conversation. I'm gonna lose all the details I put in. Exactly. Go back and refer to this file again when when you've when you after you've done that, right?

Jason Belk 32:52

Exactly. And that's one thing I mentioned early on, but I forgot to mention again, is that that is the huge advantage of this or the planning mode on Claude Code is that once you get to a certain point, all its tokens in its brain get scrunched down to a little dot, and and then it you you kind of have to start over again to certain steps. But when you have this plan all written out, it just refers back to the files. It kind of knows where it is in the flow, it has like its own checklist. Like, okay, I'm on the planning phase. Now let's move on to the tasks phase.

Cisco U Training And Cisco Live

Tim 33:14

Yeah, it's a it's a big big bonus. Um I I yeah, I had to learn about it uh because I didn't realize that until I did three or four projects and I was losing all my context and it was forgetting everything. And I was like, man, there's gonna be a better way. And so yeah, that's the anyway. No, this is good. Um we've got all the so we have all of the the links, including the one for uh that goes through the entire process of spectrum development. I think that's gonna be use uh really useful. I think we just that I think that's probably the first thing where people what I'm thinking about is okay, there's two things, right? One is where do people start? We kind of covered that, which is like literally just go look at these documents and then go fork a repo or grab download the repo. That's so that's just to the listeners that are already work with AI or that want to start working with AI and want more of a framework to start from. That's the first one. Um, the second one is uh since we have to we'll we'll start uh wrapping up here in a minute, but let's talk a little bit just before we wrap about the Cisco U cores that you built. What's at like compared to the to the docs that are the the docs we were like to see? What is the what covers what's covered in the course?

Jason Belk 34:20

So so yeah, I we have a couple links here. We have one that's a tutorial, these are free. You don't need any subscription, you just need a CCO Cisco account, which doesn't even have to be a Cisco customer. You just you can create one for free with your Gmail, I think, or whatever. And and so this to Cisco U tutorial, Cisco U is our learning and certification, learning with Cisco, place where you get your certification training officially from Cisco. So they have you know CCNA up to I think CCNP and and other training in between. Um and so we also have free tutorials on a variety of topics from cloud, security, AI, networking, the gamut. Uh I think we have over 200 at this point. So not just this one tutorial, but check out the other ones. We have ones linked to different certifications you might be pursuing as well. So for for this particular one I wrote a few weeks ago, I basically took the spec driven development um gecky link that I share, I talked about earlier and kind of converted that to a network engineering context. So if you uh if you look at his great presentation link and you're like, okay, I kind of get what it is, but how do I do that as a network engineer? Look at my link on the Cisco U1 where I basically have you go through a spec driven flow for setting up the skill and then using it to like basically build like a simple Ginja 2 YAML parser thing that creates some config files. Um, so that should get you up and running, at least with a very simple use case that's all gonna be run locally. You don't need any remote devices, you can just kind of do it with your Claud code locally on your personal laptop or whatever to get started. Uh and then I'd say on top of that, so that's the free one on the paid side, and I think they might have it free right now, depending on when this gets published, of a Cisco AI technical practitioner course. And we came out with a certification last December. So, this in terms of complexity level, on those who are familiar with Cisco certifications, the CCNA Cisco certified network associate is an associate level exam. This is one step before that. So this is meant to be entry level. I don't know what I'm doing quite yet. Let me figure it out, type of course and exam. So we came out with a course and exam in December. So it's been out for a few months now, and it it covers prompt engineering, you know, AI security, AI-driven software engineering, um, advanced prompting strategies, uh, you know, what is an LLM, what is RAG, and all that kind of stuff. And I'm I'm actually doing a Cisco Live session covering this AI tech course and certification. So if you're coming to Cisco Live US, US, come check out my breakout session with with George. Nice an hour, it'd be a lot of fun. Um and then I'm also doing another um Cisco Live session on the new um CCMP automation, autocore, EN core, like the enterprise networking automation, professional level that used to be called DevNet Professional Exams. So I have another one there that people should check out. We're running Cisco Live, and then third, I have what's called Cisco U session. So we have 45-minute sessions on on the floor, and that one will be on Spec Driven development specifically. So if you're still unsure and you happen to be a Cisco Live, check out my session there or the recording will be up, you know, after Cisco Live as well. Awesome.

Chris 37:01

Yeah, we'll definitely get all those shows or all those links in the show notes as well. I think this will definitely publish before Cisco Live. So yeah, for sure. If you are if you are going to Cisco Live US, make sure you do check out the session. That sounds great. Is it our details? Do you have details of the session as well? We can put the I can get this, yeah. Yeah, let me pop. I can pop the side.

Where To Find Jason And Wrap

Tim 37:20

Yeah, the catalog is up, catalogs up now, so we should be able to we'll we'll get that. Um I'm doing uh I'm doing a lab, but I'm not here to plug. I'll plug my stuff on a different show. But uh yeah, so we'll get we'll get that. We'll get that in the show notes. Uh definitely sign up for all of Jason's stuff. Um also, Jason, where can people find you online before we uh wrap up?

Jason Belk 37:40

Sure, yeah. So I I'm primarily on LinkedIn and and X Twitter uh as Reno Belk. I have a blue sky, but I I I only have so many places I can keep track of. And so I'd say I'd say but and then I'm you know, I'm also on a number of different Discords there out there, but yeah, uh like I'm on the R of Network Exchange. People want to ping me there for that one. Um, but uh of course I'd say like the primary primary one would be LinkedIn. That's the one I check the most often for Reno Belk. Reno Belk. At Reno Belk.

Tim 38:08

Yeah.

Chris 38:08

We'll we'll make sure we have the links on the sky's dying. Um hate to say it, but you know.

Tim 38:13

Yeah, is it well? I mean a lot of people that's the problem with the that's the problem with moving uh people over, is that a certain center of gravity has to move in order for the thing to work. So anyway. Yeah. All right. Um okay, well, yeah, let's go ahead and uh wrap it there. And we'll get all the links, and there's gonna be a lot of links in this one. Uh so definitely go through all of them. And uh yeah, we will thanks for joining us, Jason, and we'll um see everybody on the next uh show. Sounds good.

Chris 38:42

Thanks for guys. See you guys.