TRANSCRIPTION 🔗
Craig Sturgis: it's really easy to step out of your your comfort zone. Like here's one sort of like big picture agent or AI thing I believe, which is if you're an expert at something and you know like to a deep level what's good or bad, you can work with an AI to make something really good. The challenge is I mean, this is not my idea, but like when you step out of your like zone of expertise, you can get something back that looks really good to you, but very quickly like other people will recognize as hot garbage.
Hunter Harris: Hey everybody and welcome to Build & Break Through. With me today is Craig Sturgis. He has 20 years of product tech experience. He's a four-deployed tech executive and he is the founder of Vibecto. Craig Sturgis. Craig, do you want to introduce yourself a little bit?
Craig Sturgis: Yeah, so like you said, I'm I'm Craig. I've been doing this a long time. I've been really kind of AI and agent forward for really the last three years, but explicitly in the last 18 months, the the story of my company is it did start. I repositioned as Vibe CTO thinking I was gonna be spending a lot of time fixing people's vibe-coded problems and with with using my experience, and I think I was a little early on that. More recently and over the course of that time, people have asked me when I've had a name tag on, what's Vibecto? And so I've kind of leaned into that as I've been doing more kind of non-technical agent work. That's just it's become my brand now. And and I I do I focus on I am an agency that helps people deploy agents effectively, whether that's in software or in the rest of the business. And so that people can kind of grow and and double down on their growth without necessarily having to hire at all or as fast.
Hunter Harris: I dig it. you've segued for me a little bit here. I was really interested in our pre-call about learning a bit about what those agents look like outside of product and technology. So especially right now, I feel like I see a lot of people talking about AI agents in go to market and sales and marketing. So what sort of agents have you been at work building, deploying in The areas outside of our specialty.
Craig Sturgis: Sure. So I know there are lots of products on the market and people bringing their own forward-deployed engineers and they're all primarily enterprise targeted. and the in my experience, like I tend to work with founder-led companies, kind of smaller, more like at most like a hundred peop a hundred or a couple hundred people, typically in like 50 or less, 20 person, like they're they're well, like they're growing companies, they've got a good business. and they're looking to kind of leverage agents to help them grow because it's they've got key people that it's hard to hire another version of. If they could clone
Hunter Harris: Mm-hmm.
Craig Sturgis: like these key people in their operation, they would be they would be able to grow even more, but it's really hard to hire and train another version of that person. And so
Hunter Harris: Mm-hmm.
Craig Sturgis: income agents to help those people be more effective. And agents As a term is it somewhat of a confusing thing for people outside of software. to me, an agent is like now ChatGPT, since they unified Codex with it, is an agent. Claude, the Claude Desktop app, they've been they're unifying co-work with chat. And so whether people like it or not, by opening ChatGPT or Claude, they're working with an agent. And I'm not sure everyone like fully internalizes that, but
Hunter Harris: Mm-hmm.
Craig Sturgis: That's what's happening. And so a lot of my energy and attention is not on helping people buy and implement some of these bigger, higher ticket price agents. It's how do I take my Claude my Claude team, Claude Enterprise subscription, or my Chat GPT team enterprise subscription, and empower my folks across the organization to use that effectively. Looks like a whole bunch of different things, but it's just like how do we find the workflows? How do we find the parts of what's going to help a marketing person effectively leverage
Hunter Harris: Mm-hmm.
Craig Sturgis: it, but also use their creative zone of genius to make something good and not
Hunter Harris: Mm-hmm.
Craig Sturgis: just basically slap canon everything all over the internet and not stand out at all? and so, like for a a content person or a or a marketer, it might be helping them do The right type of internal and external research against what they sell in a way that they're any one person can't keep the whole business in their head. How can we help them find the right things to talk about in a marketing message or in a sales conversation? How to mine the conversations they're having out with prospects and customers to find the language that those people are using to more effectively market to them? To find what's working in the sales process, to turn that into something that you can scale throughout your team and really distill, really mine all the signals and information that are happening, talking to customers, talking to prospects inside the business, and kind of turn that into the thing I'm really big on right now is just like the concept of an agent knowledge base and really distilling all these facts and information and useful things into a common place. that agents and people can use.
Hunter Harris: Yeah, I think that approach is really, really interesting. And I mean, I have a follow-up question, and I think you've per probably alluded to that a little bit. So in terms of getting into these organizations and doing this sort of like enablement and implementation work, how do you go about starting for the most part? Like what is your in your opinion, like the critical lever that you use when you're initially getting in there to identify some of these opportunities, right? Because there is We'll we'll get to further questions around this here, right? But when you're first getting started, how do you grok that
Load full transcript
landscape and figure out like, I need to be over here versus over here, when agents can be, you know, very theoretically applicable to almost everything.
Craig Sturgis: Sure. So the way I like to think about it is I am the product leader, like I'm the head of product applying or a really good product manager or designer. They're gonna be doing workshops with people, like with customers to try to figure out what's the right way to build a solution for them. It's another reason why I think people who have worked in product roles and design roles inside tech companies are really well suited to this type of work because it's the same work. I'm going in and doing
Hunter Harris: Mm-hmm.
Craig Sturgis: a workshop first at the both at the executive level to talk big picture strategy of what is it that AI and agents change about the landscape and how does that apply to your business? Let's do the hundred thousand-foot version of this and kind of understand where are the big picture opportunities. And then doing workshops at the ground level with the actual team members to in domain-driven design, they call it an event storming. And in
Hunter Harris: Mm-hmm.
Craig Sturgis: like UX, they call it a service blueprint. You're finding all the actions that are happening. And you're you're doing some kind of like we're making a bet with the executive team, which is like, where do you think the biggest
Hunter Harris: Course.
Craig Sturgis: opportunity is? And then we go dig in there. And I'm literally building like opportunity solution trees, like hat tip to Teresa Torres, one of my favorites, like continuous discovery habits is an amazing. type of way of looking at this, in my opinion.
Hunter Harris: Mm-hmm.
Craig Sturgis: But it's basically applying that to an organization with the idea that you're gonna apply find ways to apply agents and agent native software to those workflows, to those problems. And that's how I build it out and prioritize it and bring that back to the team to say where do we need to dig in on that?
Hunter Harris: Okay. So I guess well going a little bit further on that too, given that you're running these workshops, you start assembling visibility into some of these potential bets, some of these potential workflows. you mentioned, for example, like building these solutions so that they play to the individual's strengths and don't just turn them into slop cannons. What is your framework for deciding like, okay, we got this set of tasks, these tasks tasks are going to be better done by a human, these tasks look more like they're better delegated to an agent.
Craig Sturgis: Mm-hmm. The the short answer is is trial and error. I've I've gotten
Hunter Harris: Mm-hmm.
Craig Sturgis: a lot of it's it's reps with the right person. And there and everyone's context is a little bit different. What might work in one company for a marketing workflow or a or a sales or an operations or a finance workflow may not work there. it's it's a lot of pattern matching, but I've I've had to come at it with a lot of humility because I I think
Hunter Harris: Mm-hmm.
Craig Sturgis: there's there's a an industry wide problem in tech where it we people think that job must be easy. The reality is is there's onion peelings,
Hunter Harris: It's not
Craig Sturgis: like layers upon layers in everyone's job. And it's it's coming at it with an open mind, like just a discovery mindset of like what's happening here and what's difficult. And where where I've had some success is like, okay, we're we're gonna try this. And I and I like to talk about it as like I came With this marketer and I I built this workflow end to end that did so much of the job. And it kind of stomped on the like the creative part of it, where
Hunter Harris: Mm-hmm.
Craig Sturgis: as opposed to finding like, here's the spot where it's gonna, here's I'm gonna give you the tools instead of building a an autopilot workflow. It's like, here's the thing that helps you do this type of research, here's the thing that makes it easy for you to review against what your boss is looking for so that you can kind of proactively do that before it goes to actual review. It's gonna learn from how things have gone from one side of the process to the other in the past. So that you can kind of cut out the amount of times it has to go back and forth to get to the end state of something good. And so I on sort of just the highest level it is I look at it as like handing them a toolbox. That's really awesome, that has awesome tools in it, as opposed to a conveyor belt that does their job for them. Like where I get my software development brain going and it's like, man, this is like how we would want a CI pipeline or a data pipeline to work. It's like, it's not like that. You you want to find the how to enable the most human parts of the process so that they can be good. Because if you just throw agents at it, it's gonna sound like everybody else. I I think people are learning that everyone is empowered with the same tools if you just are being lazy about it. If you think you've got a shortcut and you're gonna like you're very quickly gonna sound exactly like everyone else.
Hunter Harris: Definitely agree there. So one thing that kind of stands out to me when you describe that sort of workflow is it's almost bringing almost like a more product delivery lifecycle approach to something like marketing or something like sales, which is a very, very interesting. I wonder, are you seeing some of the same trends that we see in the software delivery lifecycle or product delivery lifecycle? in some of these agentic flows where for example, everyone's talking about quality verification as a bottleneck, for example, right? i is that same thing happening in these sorts of agentic spaces?
Craig Sturgis: Not well, in a way, right? I think the SDLC is, I think, ahead. it's where the most experimentation is happening. elsewhere, I think it's there are pieces of it. You can kind of see
how, we've learned this over here. Maybe it might be applicable here. I don't think I have any broad statements to make about what could apply across the board in a way that I'm pretty opinionated about what's happening in software. I d may just don't have enough reps yet. I do think there are concepts.
Hunter Harris: Mm-hmm.
Craig Sturgis: I mean, thinking about it systematically. I I think a good product development lifecycle is a systematic approach to how do we make something better that customers want and how do we bring the inputs and how do we then put that and learn from it from contact with the marketplace, with human beings. and I think that sort of mindset definitely applies, but you have to be careful about like The output of a product development lifecycle is a product. It is something that
Hunter Harris: Mm.
Craig Sturgis: people can can test. The output of like sales and marketing or even internal operations is like you're interacting with other humans and there's just sort of a different set of expectations that people have. I think it's why you're seeing these cultural norms develop around like rejecting AI content and AI writing
Hunter Harris: Mm-hmm.
Craig Sturgis: or AI messaging, because it it communicates to someone you didn't put any thought or intention to this. If I wanted the output of Claude, I could go prompt Claude myself. And so Right.
Hunter Harris: Yeah, just give me the prompt and cut out the middleman, right?
Craig Sturgis: Yeah. And so if it communicates to me as a receiver of that that you're putting zero effort into this. And so why should I put any effort into reading, internalizing or responding to it?
Hunter Harris: Yeah, I I guess I want to push back on one piece there, which is the product being testable, because I do know a lot of people in my network and GTM spaces are releasing cont I mean, obv obviously a product and something like a GTM cycle is different things. But in the same way, they are shipping content and they are evaluating the co quality of the content. They just can't evaluate the quality of the content until they ship it. the quality evaluation cycle happens in post when it's in front of people when you get your impressions, your click-through rate, et cetera.
Craig Sturgis: Thank you.
Hunter Harris: So they're doing a lot of workflows where they're doing a wide variety of formats and structures, et cetera, and then seeing what plays well and then essentially the same cycle, but slightly looser because it's not like you can lint on your sales copy effectively or have an automated test suite on your marketing messaging, right? I guess you could, but in my experience that probably doesn't work very well.
Craig Sturgis: Sure. And and it's all like it it's not gonna be necessarily effective. And and even on the product side of things, like you can validate whether something is linted correctly or whether it satisfies the spec, but you don't know whether that actually works until s it touches
human beings in the marketplace and
Hunter Harris: Exactly.
Craig Sturgis: they choose whether to continue using it or to buy it in an ongoing way. Like there is still that thing where even in product we don't have that. You need to have it make contact with the world. And I think at a certain to go back to your your statement, which is like, okay, well, some people are testing this systematically, I think those are still leading indicators. Like you might get a lot of views on certain content, but there are lots of people where like they may get views in aggregate. Like, are you optimizing for the right thing? I have many like my content approach is I'm not trying to maximize for views. I'm trying to maximize for things that lead people to seek me out for closed business. And oftentimes
Hunter Harris: Mm-hmm.
Craig Sturgis: the stuff that people tell me about that got them thinking about working with me flopped on LinkedIn. Or it's
Hunter Harris: Mm-hmm.
Craig Sturgis: it is not the thing that I post about to get the most views are the things that LinkedIn cares about or but it's driving zero
Hunter Harris: That's true.
Craig Sturgis: business for me. And honestly, I'll speak for myself. If I see somebody who is like hyper optimized their content, I flip the trust bit off for them because they're they're
Hunter Harris: Mm-hmm.
Craig Sturgis: optimizing for something that I don't care about. I don't care how how many views your stuff gets. It's like,
Hunter Harris: Right.
Craig Sturgis: is this useful to me? Then I if it's not, if I can tell you are hyper optimizing your content, it's actually less valuable to me and I'm probably gonna stop paying attention.
Hunter Harris: That is a very good point. I I I guess this is kind of like a interesting rabbit trail to go on a little bit. So what are you seeing in terms of the market with something like using I AI agents for optimization of content, right? Like are a lot of these people out here trying to hyper optimize their content with the assistance of LinkedIn agents? Like what are you seeing with your particular viewpoint on the market here?
Craig Sturgis: I think, and so I've been following a lot more kind of people who live in this space. I don't. I'm I'm a newcomer. I'm trying to bring my approach to it and figure
Hunter Harris: Mm-hmm.
Craig Sturgis: out how I can learn from people who've been doing it a long time. And I think we are in this moment that I think very much rhymes with early in SEO, where there were some people who
Hunter Harris: Mm-hmm. Definitely agree.
Craig Sturgis: are out in front and seeing a lot of results from what. very quickly became labeled as black hat tools or black hat approaches to things that then got
Hunter Harris: Mm-hmm.
Craig Sturgis: punished in a big way by Google. And and I think we are, if I had to bet, I'm guessing we're headed towards that where people are seeing affect like they're building in the same way that people brag about building a software factory that's helping them produce a bunch more software that's effective. People are building content factories that they're seeing results. I I think that's going to be a short term gain and in a way that won't pay off over the long term. I personally will not let AI write a word for me. I've experimented
Hunter Harris: Mm-hmm.
Craig Sturgis: with it and in the end, every time I use AI to write for me or even like I'll use an example I launched a newsletter and I wanted to reach out to 200 people that I know to invite them to the newsletter. So I'm not just gonna sign them up for it. And I wanted to okay, I'm gonna go through and I'm gonna like Go to my CRM and and my LinkedIn and find the relevant context for me to send them a message that is grounded in the last time we talked. And I would
Hunter Harris: Mm-hmm.
Craig Sturgis: do that one by one. And I ended up with a message that I sent, and I sent 200 of them. And I I did personally like I literally wrote the main part of each one. And then as they would respond days later, it would it would come back and I would read what I wrote and I'm like, I wouldn't write that if I was just like sitting down to do that.
Hunter Harris: Right.
Craig Sturgis: And so like even in those situations where I was by hand writing things, but part of it was brought together, like it I've regretted those pieces of it where I'm just not like writing from scratch, looking at the context where because it's it then
Hunter Harris: Mm-hmm.
Craig Sturgis: produced something that was less human. And these people will forgive me for that, but at the same time. Trying to find, okay, how am I if I was gonna do that again, how could I leverage the tools to gather all the stuff together for me and then switch my
Hunter Harris: Mm-hmm.
Craig Sturgis: brain into a mode where I'm gonna write something that's like authentically from me and that's gonna be the most effective, but then I can spend less time remembering or digging through my CRM or my LinkedIn messages to f to get to jog my memory of of what's the last conversation we had. those types of thing and I think you can apply that in all these different circumstances, which is can I give
Hunter Harris: Definitely.
Craig Sturgis: you the magic wand that gives you the information you need to do your best job?
Hunter Harris: Yeah, essentially you're playing to the strength of AI as a researcher, right? Versus I mean, what's fascinating is I definitely experienced the exact same thing, right? Like for the past several months, I've had a bunch of copywriting ex experiments and I've looked at my time utilization and I've dumped so much time into trying to build effective copywriting, which it seems like agents would be good at, right? Like based on the structure of how these things work, you would expect that to play to their strengths, but man, effective copywriting is I feel like it was almost better, let's say nine months ago than it is now. and I don't know if that's model
bias or something like that, but man, is it really bad right now? Like I I've adopted personally for a lot of my outreach that exact same format there where I'm like, All right, agents, I just can't have you writing any of this stuff because it just is not like I can have it almost nail structure or I can have it almost nail my voice, but if I try and combine structure and voice, it just fully beefs it and you get something absolutely terrible out of it. And so
Craig Sturgis: Mm-hmm.
Hunter Harris: like my recommendation to so many people building agents, for example, because there's so many people that are trying to build agents for LinkedIn DMs and LinkedIn comments. I mean You've probably seen all of the gallons and gallons and gallons of slop on LinkedIn comments
Craig Sturgis: Mm-hmm.
Hunter Harris: and LinkedIn posts as well. but man, the agents just for whatever reason are not there with that. and it's fa fascinating to see how maladapted they are to that use case when all of us think that that should be a strength of the thing, right?
Craig Sturgis: Yeah.
Hunter Harris: Do you have any hypotheses for what you think that is?
Craig Sturgis: I think it is it's really easy to step out of your your comfort zone. Like here's one sort of like big picture agent or AI thing I believe, which is if you're an expert at something and you know like to a deep level what's good or bad, you can work with an AI to make something really good. The challenge
Hunter Harris: Yes.
Craig Sturgis: is I mean, this is not my idea, but like when you step out of your like zone of expertise, you can get something back that looks really good to you, but very quickly like other people will recognize as hot garbage. And so
Hunter Harris: Mm-hmm.
Craig Sturgis: like I think some of that is like we are it's it's like a magic trick. Like we are it's it's amazing that this machine can do this thing. And on the surface it looks really good, but if you sleep on it and it's like that's actually even I can recognize that that's not up to a standard that I would have. for some of this. And I I think that's true for copywriting. I think it's true for code. Like I I think when you're you're
Hunter Harris: Absolutely, yeah.
Craig Sturgis: writing code, it's like I've been doing software engineering for twenty years. Like I know that like I'm not gonna throw a fifteen hundred line PR at somebody and expect that it's any good. Like you wanna make these things. Yeah.
Hunter Harris: Or if only it was fifteen hundred lines, right? I'm I'm exposed to a lot of a hundred and fifty thousand line VRs and whatnot.
Craig Sturgis: Yeah. There is absolutely no way that that's gonna be good. I I I
Hunter Harris: Yeah.
Craig Sturgis: come back to this line over and over again. Like on on the other side of this, like I don't care if you used AI to make something, and that could be software, it could be a book, it could be so I care if it's good. The problem is like it's
Hunter Harris: Right. Agreed.
Craig Sturgis: really hard to make something good without really being intentional about it and being able to j have the judgment to say. This is good.
Hunter Harris: And this goes back to verification, right? it is very you can definitely build a thing with AI. It is very hard to build a quality thing without doing some sort of quality verification, which means that you have to, in some capacity, maybe you don't have to pop the hood, but you at least have to give it a test drive, right? And if you have a massive piece of work, whatever it is, let's use the 150k PR, for example, right? Ain't nobody reading all that stuff, right? there's no way you sent that to somebody after reading all of that because it's just too big to actually do verification on, right? There's no way for a unit of work that big for you to read all of it. There's no way for you to test all of its extremities. There's no way that you've really test-driven all of what that could possibly be doing. It's just not possible because it's too large. And It's just a great example of utilizing the tools to do not best practices but worse practices of yesteryear just at
Craig Sturgis: Mm-hmm.
Hunter Harris: agentic speed, right? And I think that same approach applies as you were saying earlier, to non-coding use cases, right? Like if you're going to use an agent to pump out 15,000 DMs, ain't nobody reading all that stuff, right? Like there's no way that there's an effective quality bar there. And okay, maybe some of those are fine, but what about the ones that aren't, right?
Craig Sturgis: Mm-hmm. Yeah, it's I think there are there is a way. I think a true the right team could put together a system that produces something good. But I think it's a lot harder
Hunter Harris: Yes.
Craig Sturgis: than people suspect. And I think people are going out into the world with a lot of stuff that's not good and burning their credibility.
Hunter Harris: I definitely agree. I want to say my knee-jerk is doing really good quality stuff is not any easier than it has been. It's probably a little bit faster to do high quality stuff. Emphasis on a little faster, but I don't think it's any easier because there's just so many, so many bits of process, so many sharp edges, so many things we have to account for. I mean, look at all the organizations struggling with. well, both sides of verification, right? Figuring out what they need to do and why. And then figuring out if they actually built what they committed to. 'cause man, at this speed too, verification is a pernicious problem because there's just so much stuff to test out and so much volume going, right? I guess we've kind of covered parts this, but like In terms of going back a little bit to the deployments that you're doing, how what are you seeing in terms of if I am going to be implementing agents, like if I'm going to be a customer of yours, what sort of teams are getting the best successes and which ones are really struggling?
Craig Sturgis: it it depends on the organization, but like there are parts that are have faster uptake or that are easier to see the the progress versus others. Like I think the more creative the output is, the longer kind of the uptake is, but like the
Hunter Harris: Reasonable.
Craig Sturgis: more straightforward. But I do think where I have seen like I'll I'll go back to the thing I am just like most excited about, which is the concept of like a a distilled agent knowledge base. This is something that is helping in some of my clients every function. It is, okay, it's a software company, but it doesn't necessarily have to be a software company. It could be somebody who sells like medical products that have like these giant manuals and all these things that all this information that has to be distilled. Like taking a code base or that like the raw technical output and building kind of a distilled agent navigable knowledge base that acts as it is literally a wiki. It's text files with hyperlinks between them. and I use a a Git back end for that. So it's versioned and those types of things. But like that sort of intermediate artifact that you can then use your agents to navigate to say like, okay, well, this is the product we sell. And so X marketer who would never really even necessarily fully get, and they it's not that they're not smart, it's just they're not wired that way, how the product technically works, can get up-to-date, distilled knowledge about what the product is, what it's for, the problems it solves, how we might position it. And they can say, I want to send this message to this customer because I get them at at the at the marketing level about making connection with them as a as a potential customer. But I might not know the all the ins and outs and details of I want to send this message and how can I get grounded facts about how our product solves that problem? This knowledge
Hunter Harris: Mm-hmm.
Craig Sturgis: base helps them do that. on the other side, it's like, okay, well, we've got this big product. Sometimes there are workarounds that we've that have developed over time. Let's go mine the customer like support calls or training calls around like, well, there's this work,
Hunter Harris: Mm-hmm.
Craig Sturgis: the product works this way, but we use this workaround. Let's mine our transcripts to find those, put them into the knowledge base. And then as that that could both be an input into the knowledge base and an inp input into the product backlog to maybe go fix that.
Hunter Harris: Exactly.
Craig Sturgis: And as you then release a fix for that, there can be a workflow that's like updates the knowledge base, say we don't need to do that anymore. And so you're finding across support, across marketing, across sales, just the idea of, hey, This is how our customers are talking about this problem. Maybe we should be using more of that language in the way we talk to them as we're onboarding them, as we're trying to find the next customer. Like
this kind of it's got the raw material, but it's it's dist it's this action of distilling it, something which LLMs are amazing at, that then turns it into how do we kind of keep this ongoing, like taking all this information and condensing it down, compiling it into this really useful resource. for the whole organization. And I keep they I keep having my one of my customers come back to me like, we found a new use case for this. We're building product demos
Hunter Harris: Mm-hmm.
Craig Sturgis: and video, like we're building automated videos of the stuff we're doing. And they were terrible. But all of a sudden we hooked up your knowledge like it we hooked up the knowledge base and it started getting better. And we don't know why, but it is better. Like so there's just all these different things. Once it exists, people are finding more and more use cases for it.
Hunter Harris: you're completely topping the choir here, 'cause the big thing that I've been doing for my own business is I call it a content engine, but it's really exactly
Craig Sturgis: Yeah.
Hunter Harris: that where I record my meetings, my podcasts, I throw on my blog, I throw on my LinkedIn posts, everything that I'm generating, all the conversations that I'm having go into a single monorepo that I can then look for insight in, look for different ideas, look for connections, look for like It's basic stuff like being able to have a conversation with somebody, get an idea about a change to my business processes, and then just somewhat automatically, it's not perfectly automatic, but somewhat automatically roll out changes to my business processes just based on hypotheses coming out of calls, just like this one, for example, right?
Craig Sturgis: Mm-hmm.
Hunter Harris: Like it's incredibly powerful when organizations really can start. I mean, the first step is Gathering the data to do this sort of stuff, right? if you don't have an effective support inflow, it's hard to get that visibility. If you're not recording your calls, it's hard to get that visibility if you don't have LinkedIn hooked up appropriately, et cetera, to pull that information. It's hard to get that visibility. But once you have the visibility from all the things to your point exactly, the agents are fantastic at distilling and identifying patterns and trends and surfacing insights that you didn't know that you had. buried inside of there, right?
Craig Sturgis: Mm-hmm. Yeah, and I think a lot of people have a similar like they have the same idea, but they try to solve it by like just dumping everything in like Notion or Google Drive or SharePoint and those types of things. And so far, what I'm saying, like those systems are not necessarily built to have the same kind of information density and ease of navigation for agents that a Git repository does, which is why I think that it Call it any kind of version control, but just so it is text and hyperlinks versus okay, well, did Notion do a good enough job at their MCP to
effectively search? And and sometimes it can overwhelm the agent. They even outside of a million
Hunter Harris: Mm-hmm.
Craig Sturgis: token context window, it is like you have to like be the librarian of this information or make sure that you're
Hunter Harris: Right.
Craig Sturgis: working with your agents to be the librarian and make sure it's you've got your card catalog. set up th in the right way so that it you can find
Hunter Harris: Mm-hmm.
Craig Sturgis: shortest path to the right information.
Hunter Harris: Right, get your agentic Dewey de Decimal system up and running, right? Cause I
Craig Sturgis: Mm-hmm.
Hunter Harris: I definitely I have experienced Google Drive and Notion versus GitHub and man, ooh, the people using Google Drive for that are man, I think it's gotten better recently, but
Craig Sturgis: Mm-hmm.
Hunter Harris: historically Google Drive and agents have not played very well together at all whatsoever. Notion, I mean, I would rather someone use Notion than nothing at all, but I agree very strongly. Git is just really, really well suited for this sort of use case, right? And especially to with something like Google Drive, you have to manage the files themselves. But with GitHub, it's Git, it's version controlled. You have the entire history at your fingertips, right? So if you
Craig Sturgis: Mm-hmm.
Hunter Harris: need to go back in time and reference a particular file or something like that, even if it was deleted, right? It's still potentially available to you, which is incredibly helpful in many, many different ways. especially because you can do things like, hey, we've had this series of files. Hey, agent, go and look through the history and let's look at historical trends. It's just interesting stuff like that is just intrinsically built in when you use these tools like Git, which that it always i excites me to see non-technical people adopting stuff like Git because of that sort of power.
Craig Sturgis: experience with there there are downsides to it because Git and GitHub especially I think that is the most popular way to do this. It's what I've built some of these things on top of. But Then you have to get everybody access to GitHub and they can manage that. And from a permission
Hunter Harris: That is true.
Craig Sturgis: standpoint across multiple people, there's a lot of potential issues there around like who's got permission to change what. And so I've custom built in an increasingly
Hunter Harris: Mm-hmm.
Craig Sturgis: building a standardized way to build like a server layer in front of a Git repository, whether it's hosted on GitHub because GitHub keeps going down all the time or not, that helps
Hunter Harris: Saving it.
Craig Sturgis: sort of put put those permissions in place so that people can own
Hunter Harris: Mm-hmm.
Craig Sturgis: different parts of the repositories. There can be multiple Git repositories so that you can maybe put your finance information in one and your company
Hunter Harris: Mm-hmm.
Craig Sturgis: wide information in another and decide who has access to those governed
Hunter Harris: Right.
Craig Sturgis: at a level that is not at like worried about in GitHub, which is not really like GitHub was built for code bases. And but just that t Git style version control is very useful for this type of information, but there's this whole other area of opportunity to make it more accessible to people who are not software engineers and even people who are, like make the use case fit this a little bit better. I see a big opportunity there. I'm I am dabbling in it in my own to to make my own work easier to execute on. Whether I will sell it as a product, who knows? I I feel like services is is the right mode to go right now. versus trying to to sell a product in this in this space, but so much opportunity there.
Hunter Harris: I certainly agree. Well, Craig, I wanna be respectful of your time. Do you wanna plug your pluggables?
Craig Sturgis: Sure. so Vibecto, it's vibecto.ai. I do a monthly newsletter on what's practical or what's top of mind for me in a way that is accessible to everybody. just the it's a full time job to keep up with this. I don't think even if it was your full time job, you could. And so I try to distill down what's catching my eye and what working with people from super AI pilled software engineers down to like people who think AI is evil like and who are being forced to i i interact with it. I try to distill that down to a like a five minute video and and I post a bunch of stuff on LinkedIn and I I curate it there. But that's that's my big big thing. I'm trying to get more people to sign up for the newsletter just and and try to be helpful to them once a month. But I post on LinkedIn a lot. find me there, Craig Sturgis.
Hunter Harris: I dig it. I'll put a link to the new newsletter in the show notes here so that everybody that watches has access to it. But Craig, thank you so much for joining us and everybody, thank you very much for watching. This has been a fantastic conversation. I've learned a ton. love talking about agents, love learning about agents, and it sounds like there's a lot of really interesting stuff that I am already currently participating in and even more stuff that I have to build and learn and do. so I hope all of y'all have enjoyed this episode and I hope you all have a fantastic day. Peace.
Craig Sturgis: Yeah, thanks for having me.