← Back to search

We Built a Software Factory | Episode #068

Strictly From Nowhere: A Podcast Experiment by Cause of a Kind · 2026-10-06 · 38 min
relevance 65 7169 words Episode page ↗ Audio ↗
Show full episode description
Every customer gets their own bay in the garage. The agent lives in it, authenticated only to their repositories, and cannot get out. Mike has been building that. Justin hit record to find out what it actually means. In this episode, the two of them unpack the software factory: what the term covers, why it is loaded, and why Mike thinks the machine that built your software is as much your property as the software itself. We get into: Why Shopify was already a software factory and nobody called it that Podman pods, nightly backups, and Truman showing the agent instead of leashing it Hermes Agent, memory, and the knowledge lost every time somebody hits slash new Why domain specific skills are the ones a model will never learn The master service agreement conversation this is about to force with their attorney Why the customer experience gets more human as the machine gets more autonomous Lisp, old computer science, and programming in English as the next abstraction Advice for students: turn the AI off and learn slow while you still can People skills and taste, and why neither was required of engineers before AI enabling Agile at a fixed cost, and the body shop estimate they are chasing Software hitting done, and the agency starting to look like an IT company Strictly From Nowhere. New episodes weekly. ──────────────────────────────── LISTEN ON ALL PLATFORMS Spotify https://open.spotify.com/show/1V4SjmlnG2sntDnck8UrGW Apple Podcasts https://podcasts.apple.com/us/podcast/strictly-from-nowhere-a-podcast-experiment-by-cause/id1804380172 iHeart Radio https://www.iheart.com/podcast/269-strictly-from-nowhere-a-po-315649122/ YouTube https://www.youtube.com/channel/UCWAstEyCK6YsKVTTRsQr37w SUPPORT THE SHOW Patreon http://patreon.com/StrictlyFromNowhere CONNECT WITH US LinkedIn https://www.linkedin.com/company/strictly-from-nowhere Website https://strictlyfromnowhere.com ──────────────────────────────── FOLLOW JUSTIN ABRAMS | @cuzzinjustin LinkedIn https://www.linkedin.com/in/cuzzinjustin/ Instagram https://www.instagram.com/cuzzinjustin/ X https://x.com/cuzzinjustin TikTok https://www.tiktok.com/@cuzzinjustin Facebook https://www.facebook.com/cuzzinjustin24 Website https://www.cuzzinjustin.com ──────────────────────────────── FOLLOW MIKE RISPOLI | @mike_rispoli_cto LinkedIn https://www.linkedin.com/in/michael-rispoli-cto/ Instagram https://www.instagram.com/mike_rispoli_cto/ X https://x.com/michael_rispoli TikTok https://www.tiktok.com/@mike_rispoli Website https://michaelrispoli.com/ ──────────────────────────────── CAUSE OF A KIND | @causeofakind Website https://www.causeofakind.com LinkedIn https://www.linkedin.com/company/cause-of-a-kind Instagram https://www.instagram.com/causeofakind/ X https://x.com/causeofakind TikTok https://www.tiktok.com/@causeofakind Newsletter Signup http://eepurl.com/gcc-yj Newsletter Archive https://us16.campaign-archive.com/home/?u=847ea1526d6523a41ef1eb5a5&id=48d53e9627 ──────────────────────────────── OPEN SOURCE PROJECTS CodeDojo https://github.com/Cause-of-a-Kind/code-dojo-core OCR Ruby GEM https://activestorage-ocr-demo.fly.dev/ ──────────────────────────────── COMMUNITY INITIATIVES Long Island Technologists https://www.longislandtechnologists.com https://www.linkedin.com/company/long-island-technologists-networking https://www.instagram.com/longislandtechnologists/ ──────────────────────────────── We Build Products, Because We Want Them To Exist https://coakstudio.com ──────────────────────────────── Brought to you by Cause of a Kind A boutique software development and startup studio from Long Island, New York Stand up. Show up. Build.
✨ Episode Outline — click any point to jump to it in the episode
Problem solved
How an agency turns AI-assisted development into a persistent, per-customer 'software factory' that keeps memory, skills and security isolation across engineers and repositories.
Benefits
  • Machine works 24/7, pulling tickets from the project board autonomously
  • Shared memory and self-built skills stop context loss between engineers
  • Per-customer containers isolate agents from other clients' code and credentials
  • Customers receive the machine record, not just the software output
Use cases
  • Always-on machine pulls planned tickets and bug reports and attempts autonomous fixes
  • Hermes Agent builds memory and skills from client projects spanning many repositories
  • Each customer gets a Podman 'bay' holding their repositories and their own agent
  • Agent artifacts dumped into the FIN project management tool for shared context
  • Healthcare and fintech builds requiring extra care and security
KPIs / results
  • Machine works roughly 3x the time humans put in
  • Mark Cuban: 33 million US companies, 30 million of them solopreneurs
Tools / build
  • Hermes Agent per-customer software factory
  • Podman container bays
  • Skill Deck private skills repository
  • FIN project management (agent artifacts section)
  • Claude Code / Pi with Ralph (goal) loops
0:00 / 0:00
You know, what's up dude? What's up man? It's good to do stand up. I got a burning topic, it's burning in my eyes. Okay. Well you're about to set up FinFactory for another customer, the Software Factory for another customer. And I'm just like, well what the hell does that mean? So let's hit record. And I like gotta know what the heck you're working on. I also am going to insert here a Mark Cuban clip from a podcast that we were walking this morning. And I think it's relatively timely to like how we're innovating the business. I've been through every single technology, you know, event and evolution and this blows them all away. Now how you implement it in business is a whole different issue. Like literally when I was 24, I was walking into companies who had never seen a PC before in their lives and explaining to them the value and having these guys going, well son, I got this receptionist right there. I got that secretary. I'm never going to need that shit ever. Right. And, but then, you know, my business then was helping them figure out how to implement it to give them an advantage. There are going to be integrators that particularly young kids, like when I'm telling my kids who are 15, 18, 21 or 19, 21 and kids going to school, what should I do? What should I do? I'd be like AI research, learn all you can about AI, but learn more on how to implement them in companies. Right. Because to your point, companies don't understand how to implement all that right now to get a competitive advantage. You got the head of Microsoft saying software is dead because everything's going to be customized to your unique utilization. Right. Or use usage. Who's going to do it for them, particularly small to medium sized businesses. There are 33 million companies in this country. 30 million of them are solopreneurs, right? Single person enterprises. There are only, you know, there are millions of companies that have one, five, 10, 50, 100, 500 people that aren't going to have AI budgets. Aren't going to have AI experts. This is where kids getting hired coming out of college are really going to have a unique opportunity. If you're spending your senior year in college right now, your senior year in high school, even whatever it is, your excess time and you're learning the difference between Sora and Vio and you're learning how to do all this video. You're learning how to customize a model so that then you can then walk into a company and say, I understand your business as a shoe company selling shoes at a retail store. We're selling and selling shoes online. Let me show you how to benefit you. That is every single job that's going to be available for kids coming out of school because every single company needs that. We literally killed the old business that we used to run. And I think folks that are not prepared to do that are going to have a little bit of difficulty transitioning in this new economy, in this new era of work. And so I think like, why not put on display? You know, the reason why we're not so verbose about this and really talking about it is because this is moving so damn fast. And like the information that you knew a month ago is like irrelevant, makes you almost look like you don't know what the hell you're talking about if you speak about it today. But I think fundamentally the infrastructure of our business has changed how we engage with customers, what their sentiment is, and how we aim to build businesses. It's not even building software anymore. Like we built the software that builds your business. But what does that mean? What does business even look like these days? So dude, as you kick it off, what the heck are you working on? Yeah, yeah. So I'll start with I'm working on software factory. I know that's a loaded term. Yeah, what does that even mean, dude? Because one day you're talking to me about a freaking mechanics garage, and I'm pulling cars into the bay, and X customer is in bay one. The other day you're talking about a 24-7 factory. Think about like literally American manufacturing, like 24-7 mechanical Turk factory is humming. Machines require oil and some love, but we're just running. And so I don't even know what the heck we're really even talking about. All I know is that the machine is working 3x the amount of time that we actually humans are putting into it. What does that even mean? Yeah, so I'll start with John Crickett, our good friend John Crickett, accurately posted today that the term software factory is basically a loaded term, right? When you think about what a factory does, it outputs the same thing or things over and over and over again. We actually already have software factories. I would argue Shopify is a software factory in the sense that when I say open a new store, out comes a new store. Basically the exact store that thousands of other people launch at the exact same time, right? That's a proper factory in proper terms. I'm going to go ahead and engineers kind of hate this concept. And my response to John was, it's actually a little bit more like a garage or a kitchen where a kitchen does like custom food over and over again. Sometimes they'll output the same menu throughout that night, but right, a good chef is able to combine ingredients and whatnot and kind of put this together in a custom way, which feels adequate. Or a garage, like I said, the mechanic shop, which is, hey, you know, you customize a car, right? You put a new, you know, you're a lot of what we do as an agency is we have existing software components that we're putting together in novel ways for different businesses. We are not inventing the next great cryptographic innovation that's usually done in-house. And so sometimes when I talk to engineers, I tell, I have to explain to them where I'm coming from as a manufacturer of software. As a software development shop, you have to get out of your head that I'm doing anything really, really novel. I'm not inventing the next great programming language. I'm not inventing the next great database. Nobody hires an agency to do that. Engineers do that, right? And that type of work is great and they're doing the Lord's work out there. That's just not what I'm great at and that's not what our shop, what our business does. Our business builds software for business. It's an assembler. We assemble from existing and hardened computer components new pieces of software in novel ways with brand design. It may be an e-commerce store. It may just be a website. It may be an entire CRM or increasingly you and I are working in healthcare and fintech because those types of orgs require a level of care and security. And so why do I use the term software factory? Well, because that's just the term that the collective consciousness has decided to use for whatever this is, this AI, this clanker assisted development program that we're running. And I've done enough marketing to realize that I am not a trendsetter. So, you know, what's the famous line? Stop trying to make fetch happen, Gretchen. Right? That is so fetch. Gretchen, stop trying to make fetch happen. It's not going to happen. Like, I'm not like the software kitchen is what I'm not going to be that guy. So for those of you watching, forgive me for using the term software factory. This seems to register in the minds of the people I'm selling to. And so that's all I care about. When I say a software factory, I think people know roughly what that is in space. It's AI assisted development. They understand that there's some degree of prompting and there's some degree of machinery that we are putting together that ensures that our prompts turn into predictable custom outputs. That is not a mechanical factory. But I like the look of steampunk too. So you already know I'm playing around with terms and branding and whatever. And I'm just doing it because I like it. So that's that. So what is our, what do I think a software factory actually looks like? What I think it looks like is I don't think it looks like agent swarms. I don't think that even agent swarming is necessary for a software factory to happen. What I think that a software factory is in our sense, it's the machine that builds the machine or your machine. So I'll talk to them about this from a customer's perspective. A customer is likely building themselves a prototype at this very moment. Someone in your organization is sitting in front of Claude Code and they are prompting and seeing things come out the other side. And they likely spend hours per day doing this. Lots and lots of time doing this. And they don't really know, they don't even look at the code. They don't know what's coming out the other side. Engineers increasingly are not looking at the code either. However, we do know what's coming out the other side, right? I know that, hey, when I kind of take a peek under the hood, this is a Ruby on Rails code base. This is roughly where the modules are. This is the architecture that was chosen. I'm peeking and I'm checking and peeking and checking. Whereas the vibe kind of coder is just vibing through. But here's the other thing that was interesting in both cases. The knowledge that's getting you to the end product. The machine doesn't get better at producing your product really unless you're actively adding skills files to it or the models are getting better. Effectively, every time you start a new session, the context lives in the code or the code base that you're working on. And you'll notice that you probably have a docs directory that agents are writing lots and lots and lots of markdown to. We've done this for some time. We've done this for some time and I have a bunch of skills files that run in loops. I've gotten very into the whole Ralph loop thing, the goal loop, if you will. I don't think that vibe coders are really running that kind of stuff. So we're like the big difference between us and kind of someone using like, you know, just a tool in their computer. My computer is on 24 hours a day, seven days a week. And it is looking at our project board and it is pulling tickets from that project board, some of which I've planned, some of which are bug reports. And it's attempting to resolve them in an autonomous fashion. Using the skills and knowledge that I have given your specific machine, many of which we have developed here at Cause of a Kind or we've kind of pulled through the internet. If you've seen Skill Deck, that's the idea behind Skill Deck. We have like a private repository of skills, some of which are even domain specific to your business. This is kind of the future to me of skills, which is, you know, there are things that the AI models are never going to learn. No matter how many times they retrain them, you have proprietary stuff to your business. That kind of goes into the skills files that we put together for folks to build this. So that's been kind of one level of this. But I wasn't satisfied with that. Why wasn't I satisfied with that? Well, we share skill files. I'm working some. Some of our engineers are working on some. You still need humans to be sitting there and supervising the machine. You still need a foreman of sorts, if you want to call it. This is kind of the engineering of the future. Is constantly tuning this machine to make sure the staging environments go up. All the QA works. All the CICD is working. If there's an outage in a model, do we have backup models set up, right? Keeping this machine running is no small job. But from the client's perspective, you are outputting valuable software at a much faster rate than you've ever done it before. Right? We're really kind of keeping the machine up and running. But the machine does have this idea of memory loss just between engineers. Like if I'm using Pi and someone else is using Cloud Code and someone... Everything they do, all that little prompting that's happening in between the steps starts to get lost. You try and put that stuff in the project management solution. So our project management solution, FIN, we have a section for agent artifacts that I've put in that kind of dumps this data back there. Really not a bad place for it. But I was then introduced to a few concepts. One of which was this idea that I had, which is that code authoring almost needs to become a part of CI. In the sense that we all need to be kind of funneling up our plans and our intentions through the same machine. And not individually working in this bespoke manner on our own machines. And I was introduced to Hermes Agent by somebody at a meetup. And what I love about Hermes Agent is it keeps a memory and builds its own skills from the work being done on a client's projects, which may span many repositories. I also want that container, that box, to be something that you buy from Cause of a Kind. That as an agency, we don't now just provide you with a product. We need to provide you with the record of the machine. How to stand up the machine again that helped build that product. That is as much a part of your software as the actual valuable output. Like if you wanted to leave Cause of a Kind or hire your own developer, they should know, hey, this is the machinery. And increasingly, it's not enough to say it all lives in the repository, use whatever harness you want. Or it all lives in the repository, use whatever. And then there's the security aspect of it. This is all happening on our machines. But realistically, like my machine is authenticated across all of our GitHub, across all of our servers. And it really could reach across code bases. And I kept thinking, I don't really want a leash for my agent. Kind of want a Truman Show, the agent, for every customer. And every customer's agents or groups of agents live in a bay. And I'm using something called Podman for that. It's a container framework, similar to Docker. And so every customer gets this bay or pod with all their repositories inside of it. They get their agent. It's that agent is only authenticated to their repositories and their servers. Nothing can leak out of it. And it's backed up every single night to another hard drive, right? And I can even back it up to multiple places. So if my computer ever bricks or if I deploy this to a cloud environment, which is also possible. So that was the other thing is this could also be deployed to a cloud environment. So multiple engineers could go through each. Right now, I've been doing the testing on my machine. And increasingly, I'm thinking that's kind of the model that it'll just kind of live on local hardware because, you know, I don't want to pay for cloud machines times 100 customers. Right. I can put a bunch of these on my machine because I have a lot of resources on my machine here. All you Mac mini open claw folks. But the cool thing is everything that happens in that machine, that agent can't escape the box. It stays in there. I can have an agent from above help me push tasks in to these or they are all hooked up to our project management solution. So like I can start a work command, select which tickets I want worked on. And they also have, as Hermes does, have their own kind of cron ticker so they can check the boards for bug reports. And so every customer gets their own kind of bay in our garage, so to speak, in our shop. And that bay is permanent and that bay is backed up. It's persistent. It is self-learning. So what's really cool is that these AI tools that are building valuable pieces of your business are now yours as well. And so this is kind of the future I'm thinking of for Software Factory. Again, for better or worse, the term is, hey, do you really own the machinery that's being put into? And the cool thing about these pods is I can stand this up in your office. It could be stood up anywhere. That's kind of the idea. And it moves a lot of things back into the control of the business, the person buying the software. I don't think people are even thinking about that right now. But I'm thinking about it as a pretty big differentiator. I mean, I got to think about it. I just wrote down a note. I said, you know, we got to talk to our attorney because our master service agreement really talks about ownership rights. So who owns what within the engagement? We never even really had to talk about like machine infrastructure. We never had to talk about your agent or your historical artifacts that we use to run your business. There's all new things. Our master service agreement as a service provider always entitled folks to things like their GitHub, their domain, their hosting service that even though we provide that, they can take that wherever they want. Their GitHub repository, things that are theirs. The code belongs to a customer. There are now more things that actually belong to the customer. And we want them to have that. The infrastructure that we're building on a business, I understand what they're entitled to. I understand like the new way of building the product or providing the service that we provide. But what is the customer experience for right now? The customer experience is going to be very much human. So this is kind of the other side of it. The irony of this and where I think engineering is going and the people we try to hire. I think that the customer unequivocally, like unilaterally, I should say, does not want to talk to a machine. Now, our machines are picking up and sometimes commenting on tickets. Don't get me wrong. That's when it needs more information. And look, if you look at that, that's actually the one part of the experience I don't love is that the machine sometimes puts out these incredibly verbose comments as it's working on tickets. And I'd much rather talk with the customer and then I'll kind of respond. I really think that the future is the collaboration between the client and the product engineer. I'm going to call them a product engineer because that's really what we're increasingly seeing software engineering start turn into. And the designers too. And the incredible thing is we see UX designers building things in code now, literally designing with code as they're just shelled front ends. But it's an incredible way to do a prototype and a mock-up. So they're kind of bleeding into the code. We see engineers diving into hardware and infrastructure to build these things and to kind of think about how we're going to stand up all these different CICD processes and what changes when different pieces of the machine break. That's a totally new realm for them. But ultimately, all of us have to get great at working with the customer and providing a really human experience because no matter how you cut it, people love the phone call, man. They love getting on the video chat. They love talking about their product, talking about their business. And to be honest, some of these businesses are so complicated. And you and I have seen it like with the few customers that I have seen. The workflows are so complicated. It really requires a lot of time with the customer teasing that out, either doing like a user story mapping exercise, doing an event modeling exercise or just listening. And then you and I take that away. Yeah, and just take that away. Or we release a beta of a product and we go, you know what? This is just too complicated. We got to do something new with this. We got to change it around. But all of that, right? That those complicated business processes, if you are hiring someone and all they're doing is just like prompting, talking to you, prompting. As soon as they hit slash new, clear out the context, all that is lost, right? A lot of people are not thinking. A lot of people are building skills around programming languages and frameworks and ways to develop and do TDD. But they're not thinking about using these to build out the business logic, which is really the piece that matters the most. And one of the things that I love about Hermes is it kind of abstracts that away. It abstracts the subagent thing away from you, too. Like I spent a lot of time with this factory over the last few months trying to dictate the perfect workflows and subagent workflows. And increasingly, I'm finding that to be useless. And Uncle Bob Martin, I think, also agrees with this in one of his recent bathrobe rants, which is I'm increasingly seeing the harnesses are becoming way more capable. And that's another reason I chose Hermes is when it needs subagents, it figures out what it needs, right? All I'm telling it is I really want to make sure you're planning, you're coding, and you're reviewing using the things that we want you to review for. And you're checking against kind of the business logic that the customer has put forth as necessary. So it's kind of keeping the rails on all of that. This, to me, feels like what the concept of factory is. And that's why, really, I mean, the better term is almost more like a workshop or a kitchen or a garage because you aren't always outputting the same thing. But you need to be able to output consistent things for a given business. And that is the thing that the agency should be providing or anybody who's building, right? If you're an engineer in a company, you should be thinking about this. How do I almost build this incredible repository of knowledge that is – and it's not, oh, I have it all as markdown files and docs because that is just a terrible way to handle it. Well, this is a great segue. You know, I said at the top of the show, I wanted to talk about this interview that you and I saw. I believe the show is TBPN, the Technology Daily Show. And it was Jordi Hayes and John Coogan are the hosts of that show. And, of course, the guest was Mark Cuban. I find Mark Cuban to be super fascinating. I'm sure a lot of people are fans. It doesn't really matter of what your opinion is, good or bad. I think from the perspective of, like, what is happening at scale over the economy, over the new age of the workforce, over the 33 million-plus businesses that exist in our country, what is the new thing that businesses need to focus on? There are 33 million companies in this country. 30 million of them are solopreneurs, right? Single-person enterprises. There are only, you know, there are millions of companies that have one, five, 10, 50, 100, 500 people that aren't going to have AI budgets, aren't going to have AI experts. This is where kids getting hired coming out of college are really going to have a unique opportunity. What is the new skill that students need to focus on to make themselves valuable candidates to the next generation of the workforce? What you're talking about is where the puck is going. Like, it doesn't really serve young people today to go and learn, let's say, the fundamentals of computer science and coding. It does make a lot of sense to go really learn what product engineering looks like, what good design looks like, how to manage agent systems, how to think about transforming business requirements and adopting the new generation of tooling that's available. What does it look like to do AI implementation, AI workflow support? What does it feel like to create your own tooling rather than assemble tooling from other SaaS products that may exist within the space? This is like the new age of skill. Well, I would correct one thing. Yeah, let's hear it. And it does matter to learn computer science. For everything we're doing with AI, at its core, everything I'm building around factories, if you look at how I'm thinking about using different languages, that has all come from computer science and knowing what's available to me. If you've heard me talk a little bit in Slack, I'm very big on working with Lisp these days because the characteristics of Lisp to me enable another echelon of AI programming. That is based in old, old, old computer science principles that have been around for decades. A lot of us are seeking our inspiration from that. If all you know is how to set up AI tools, you aren't really going to be part of the wave of people that are going to kind of build on top of the giants that came before us. And so everything I do is standing on the shoulders of giants. Every new abstraction, every new programming language stands on the shoulders of everything that came before it. The idea of programming in English is simply the next level in that long chain of abstractions that has come before me. I would say if you're a student right now, by the time you graduate school, everything we're doing today is going to be irrelevant because we'll have invented new things and new paradigms, all of which you'll be able to pick up in like a week or two. None of the AI stuff, like all the stuff that I've been working on, loops and these idea of graphs. If you guys were around for the whole graphs craze, which lasted like a week, it's over. It's done. If you miss that entire thing, it's in the harness. You don't need to worry about it anymore. We've moved past it and now we're on to the next thing. The next thing is thinking about all these like language ergonomics and how can we make this developer experience better? What does the IDE look like when it's not focused on reading the code anymore? So if you're in school right now, I would say I don't know how you learn all of that without coding and reading a lot of these books and how things used to work. You are you have an exorbitant four years is an exorbitant amount of time. All of your classes are what? An hour long, an hour and a half long. Maybe if you had a lab, it was like three hours long. You have an exorbitant amount of time to spend learning slow. Turn the AI off and learn it slow because when you finally turn the AI on, you're going to have a wealth of this kind of experience because you got to bring what I'm looking for for engineers when they're working with our customers. Two big things. One is your people skills. So if you don't have that, go bartend or work in a restaurant and like break the ice on yourself because you're going to have to learn that very well. The second thing, the sales. Yeah, well, everything is sales, right? Everything is sales in life. So you'll learn that just by like getting out and interacting with people. So get out, interact with people. And the second thing is spend that time doing the hand code, reading about how these machines work, reading about different types of languages, different types of frameworks. No piece of information is useless for you to learn. You may not use it directly because you need to bring inspiration to the table. Remember, the LLM is trending everything towards kind of the like someone told this to us at CTO Lunches. He goes like the trending towards the gray. It kind of brings everything to the middle in some way. And there's all these nodes. If you think about the LLM is like all of these like parameters in space, right? You, your inspiration, your prompts can send it down in novel directions and you can come up with these really unique solutions to really difficult problems. I don't think you get there if you don't know anything. Like, you know what I mean? If you don't have your own taste, that's the taste that you got to bring to the table and you got to develop taste. So you got to develop people skills and you got to develop taste. Those are two things that really you could be a software engineer prior to this, this change and not have either of those things. You could sit in a basin and just type code to a spec that was handed down from you from people who sat in meetings all day. I don't think that world exists anymore. What an interesting evolution. You know, it's really fun to be, I'm kind of looking in. My responsibility is to figure out how do I simplify this and communicate it? How do I make it so that the average person that we speak to understands exactly the type of business that we run and the value that we deliver back to them? And so what I find interesting is I have to kind of not necessarily concern myself with the progression. I think you're opposite. Like, you're obsessed with where progression is. Like, you're always open to finding the next thing that is in service of the best possible developer experience and customer experience simultaneously. And so some things are fantastic developer experience, but terrible for a service business. And then vice versa. Some things would be great for service and just absolutely dismal for us to sit in. I concern myself with the sexiness of all of that. So, well, I'll tell you that. See, I think yes, but I'll tell you, you're going to know exactly where I'm going with this. The problem with software over the last 10 years that you and I have been selling custom software solutions is selling this idea of we don't know how much it's going to cost. We don't know how long it's going to take. And we don't know what the value is going to be. You're basically saying, cut me a blank check. And then there was somebody out there that said, cut me a check for this amount, which allowed them to kick the can on saying I need more money because we didn't launch it in time. But anybody who's been in the software, like buying software from agencies game knows. And so you and I have been very big on the agile thing. I'm still very big. I think that what's happened now is that AI has enabled agile at a speed that basically it allows for fixed cost. So the people here who are like agile, they're flipping their desk over right now at me like, no, there's no way you could do that. But the customer has always wanted this to feel like anything else they buy in life. I walk into the store or I go to a construction company and I say, hey, how much is it going to cost? They walk around, they give you an estimate. And that estimate stands. I know a lot of people have horror stories about customers. Listen, I know the contractors probably win some and they lose some under the hood. I don't know what happens in their business, but I know that when that guy hands me an estimate and tells me I'm going to put gutters on and siding on, it's going to look like this and it's going to do this. He's got it here. He's got it here to that price. You know, like that's pretty much how it is. When I got an estimate for my car at the body shop, that's that's going to be the price. You know, he might he may win some or lose some right. How he works that out under the hood. We've talked about this. There's the whole concept of shop hours versus the hours there. I'm not over here looking at looking at that. Right. This is the cost. That's what it's going to cost. Right. Buying software was never like that before. And I've been in pursuit of that experience. And I've heard of people getting it that way. I mean, I'm in pursuit of that experience because that's what the customer wants when they hire somebody to do a job is tell me how much it's going to cost. Tell me when it's going to be done. And that's it. Now, I'll tell you right now, you and I both know when it's going to be done is usually flexible for 99 percent of cases. Yeah, right. They're like, listen, if we're running the clankers overnight and like, hey, listen, we're going to be a day late. That's like going to be two weeks. That's fine. As long as it doesn't cost me more. When it gets hard is when it's going to cost me more. Some people, though, yes, there's sometimes like a hard deadline. And it's like, hey, the launch date for this NFT is on this date and you got to hit this date. Yes, yes, yes. That is like that's happened like four times in my whole career where like somebody's like got some kind of like the we're building an application for the New York State Marathon. It's got to happen by that date. Most of the time it's like, hey, they want to know time cost. But really cost is the thing that they don't want to have to pay more. And that's it. Dude, they want to know the price is the price. So I'm in pursuit of how do we that how do we provide incredible value, the whole agile workflow where we're producing software that provides value. But how do we change it so that when you come here, your idea manifests itself at the cost we agreed to at the start of the project? Imagine that. I know that's what customers want to buy. So I'm never going to stop. I'm never going to stop looking for that. That's exactly right. And now it's our responsibility to remain as fiscally as responsible as we possibly can to manage the money that it costs. But I there's no way that I could possibly say, you know, historically, when you get quotes from software development companies, you get somebody that just does like the old I don't know what this costs and I'll just three exit. Just cover whatever force majeure might happen down the line. You have the people that say, I can't estimate. I only do time and materials and I cost 200 bucks an hour. You know, we'll bill you in 100 hour increments or whatever combination of that scenario that people produce. But today, let's just say your AI versus my AI. If I take the same project specification, which we see a lot of these days, a lot of people are coming to us with with PRDs, with specs, with with design files, with architect systems, with like the whole UI kit assembled like crazy. But if I took your package that you gave me and I put it through my AI and you took yours and put it through yours, just like taking your car to a body shop, you're going to get relative timelines that are basically the same, same biddable within a degree. Instead of how it used to be, which is like whatever salesman can assemble the best possible deal on that day and take your money out of your wallet. You and I are looking for more standardization, more predictability than that. I'm also looking for right because the machine doesn't know like, you know, the machine sort of like adds hours to things. It's so interesting because then you're like, yeah, but I know that it's not going to take in like it's hard to predict hours. It's hard to know what this new metric is, what I think. And also it depends on the thing. Like some things do require like a human input. Like if we're setting up Stripe, a machine can't just get you or an app go and put you in the app store. The machine can't just do that. Or if we're doing user testing, we need to watch and collect feedback. There's a whole bunch of stuff. So we have to kind of separate these tasks into what does the clanker do and what does the human do? And which becomes most expensive in that world? And then the other concept is we get to a state of done a lot faster. So, you know, a lot of businesses come to us. So there's this world where it was like, oh, you're always taking feedback. You're always building the next new feature. That worked for the SaaS world where you were trying to capture market share and just keep capturing market share and grow at all costs. Increasingly, though, we're seeing people who are like, I need you to build this software for my business. These are the features. I don't need anything more than this. And maybe they think of two or three things during the process that they'd like to add. But it's not like, OK, the next big product, you know, what's our OKR for the next quarter? They're like, it's done. So it starts to look a little bit more like an IT company where now we just go into maintenance mode and we make sure your packages are up to date and the security stuff is handled. And so I always say, you know, I wrote something yesterday that it starts to look a lot like what happened with DevOps. And I worked at a bunch of different agencies. And for the most part, we never really hired someone for DevOps because if you had someone full time for DevOps, they'd stand up servers, they'd stand up the development environments and flows. And then there'd be nothing to do for a really long time. Until something went wrong, there was nothing to do. So it was sort of like, OK, you're better off using a contractor for that or a company where you pay kind of a monthly maintenance fee and say, hey, when I, you know, for this price, if I call you, will you drop everything and come and help me? Right. That's what your IT company does, too. Right. It's like if the email is down, they'll drop everything and help you. But you pay kind of monthly. And for the most part, they're setting up Outlook anyway. And it just kind of runs right. I think that our software agency, the software factory starts to look a little more like that, too, where after the software is done, adding new features should be pretty low cost because we've built the factory. We've built all of these skills. It knows all about your business. So every ticket that goes into that sort of requires less and less human input outside of like the initial design of it. You start to get used to that. And so do we start looking at more like an IT company where it's like, hey, we're going to handle the infrastructure. We're going to handle the, you know, your servers, make sure all your subscriptions are up to date, make sure all the security is up to date and then also handle some feature work throughout the year. But you get to that state of done a lot quicker because, again, you're building a tiny slice of a SaaS product. So they don't need the SaaS product anymore. But you don't have the pressure to keep growing it because there isn't a new feature to keep building. A world of done. A world of done. Software hits a done. Unbelievable how the sales pitchers always, you know, software is never done. Let's just figure out how you can sustain it forever. Well, some people are in a sense that they keep thinking of new things to automate. Sure. Which kind of dovetails into your Mark Cuban. Sure. Cuban thing, which was, hey, nobody thought they needed a computer either. Literally when I was 24, I was walking into companies who had never seen a PC before in their lives and explaining to them the value and having these guys going, well, son, I got this receptionist right there. I got that secretary. I'm never going to need that shit ever. And he's like, nobody thought they needed a computer. I got my secretary. I got this. I don't need this. It's the same thing. Then people start seeing what AI could do. It's like, okay, we're going to process these documents. And they go, you know what? I got more documents than I got to process. Oh, you know what? I found another thing that's really manual in my business. Can we do medical billing? Can we code the, yeah, yeah, we can build systems for that. Testing those systems, monitoring those systems. All that starts to turn into the future of what the business is. So on some sense, though, I don't know if I'd always look at it as like the bottom line goes away totally. Ideally, the efficiency should help you produce more top line revenue when you start doing things like that.