Subscribe
Share
Share
Embed
In this episode of Pragmatic AI, host Matt Stauffer interviews Aaron Francis, a Laravel developer and co-founder of Try Hard Studios. They discuss Aaron's journey from being a software developer to an educator, the impact of AI on the development landscape, and the importance of adapting to new technologies. Aaron shares his insights on the challenges and opportunities presented by AI, the shift in developer education, and his current ventures, including faster.dev, which focuses on teaching engineers how to use AI effectively. The conversation also touches on the quality of code produced by AI and the importance of maintaining standards in software development.
- Matt Stauffer on Twitter
- Tighten Website
- Aaron Francis on Twitter
- Aaron Francis Website
- Try Hard Studios
- Faster.dev
- Mostly Technical Podcast
-----
Editing and transcription sponsored by Tighten.
Creators and Guests
Host
What is Pragmatic AI with Matt Stauffer?
Pragmatic AI is a podcast about practical AI for business leaders and developers—the stuff that actually works, not the hate or the hype. We cover using AI tools day-to-day and integrating AI into the apps you build.
Matt Stauffer:
Hey and welcome back to Pragmatic AI where we talk about using AI in the real world. What works, how to use it well, and when it causes more harm than good. Talking about practical tools and real trade-offs for builders and business leaders. And my guest today for episode one is a builder and a business leader, Aaron Francis, Laravel developer, co-founder of Tryhard Studios and the creator of upcoming faster.dev. So Aaron, who is Aaron Francis? What do you do day to day right now?
Because it's been a lot over the last few years. You've done a lot of different things. But who are you right now?
Aaron Francis:
Boy.
Well, yeah, life has changed a lot. Thanks for having me. One thing that hasn't changed is I still have four kids. I've kept them all alive, so we still have the four. So that's, four under five, two sets of twins. That's my number one priority. And that is the only constant, is that I have the same number of kids. Everything else in the world has changed. So historically, yes, very much a wife guy. She's still the boss and the best.
Matt Stauffer:
Of course.
Matt Stauffer:
Four under five.
Matt Stauffer:
And you're a wife guy. That's never stopped. I love that.
Aaron Francis:
The things that have changed is everything else, literally everything else. Historically, I am a software developer. In the past three years, probably, I've been more of a developer educator, where I've been teaching people kind of like the deep.
Matt Stauffer:
Mm-hmm.
Aaron Francis:
how to use databases as an application engineer. And that's been going really well and I enjoy that. I feel in my heart of hearts that I'm like an educator. But around Thanksgiving, a couple of new AI things came out and I feel like I saw the writing on the wall that like certain types of developer education were going away and new types of developer education were needed.
And so since Thanksgiving, I've been not leaving databases behind, because I still think that's valuable, that's pretty low in the stack, but also trying to like skip over that middle and bring developers into the AI era. that one, so that they don't get left behind, because I think things are changing. And two, it's just a lot of fun. And we're, it's like so productive. And I want to help people do it in like a thoughtful and conscientious and engineering for
word way.
Matt Stauffer:
thoughtful and conscientious, I love that. mean, we talked for a minute before and I was just saying like, it's surprisingly difficult to find people who are trying to be anything other than a doomer or just like a full on maximalist in the AI world. And so for a background, I mentioned also that I'm trying to expose this to people who've never heard of you before. You're not like an AI bro, right? Like you have not been.
Aaron Francis:
Mm-hmm
Aaron Francis:
I sure hope not. No.
Matt Stauffer:
pitching AI for the last four years. You weren't trying to get us all on web three NFTs prior to that. You were trying to push back on AI for quite a while. Even when people were saying, Aaron, you're a great educator. I think this should be the thing you educate on. I watched to be like, nah, I'm not trying to do it. So this is, and I don't mean to say you're new at this to say that doesn't mean you have something to offer because you certainly do. But I think it puts you in a unique perspective of being able to say,
Aaron Francis:
never owned an NFT.
Mm-hmm.
Aaron Francis:
Mm-hmm. Mm-hmm.
Matt Stauffer:
I'm not saying there's nothing to learn from maximalists. I hope to have some people in the podcast who've been full out AI for the last five years and learn about what they know that the rest of us don't. But I'm very curious about people whose day-to-day lives are impacted by the usage and the encroaching of AI and good and for bad. So I think the first kind of aspects of that I want to talk about for you is,
Aaron Francis:
Mm-hmm.
Matt Stauffer:
You mentioned both it felt like the know, the bell was ringing about what you used to do, which could be like a negative, but you also mentioned the excitement. So where do you feel on the meter? Like how much of this is like a inevitability of, well, I have to change what I'm doing because this is where the world is pushing me. And how much of this is, you know what, God, this is really fun. And is it a little bit of both? Like, where are you landing on?
Aaron Francis:
Mm-hmm
Aaron Francis:
Yeah, I mean it's definitely-
both. I've gone through the five stages of grief already. But yeah, I like I appreciate you saying that because I feel like I do come by it honestly. And one of the perks of having like a long running, you know, How's Your Father podcast, which I have with my friend Ian is we just talk for an hour every week out in public. And it's pretty well documented that I was like, guys, this ain't it. Like, I know that, you know, the social media wants you to believe that it can like change
Matt Stauffer:
Yeah.
Matt Stauffer:
Mm-hmm.
Matt Stauffer:
Yeah.
Aaron Francis:
your life, but I've tried it and it can't and it sucks. And so that's well documented, which in hindsight is a great thing. And then there's a very specific episode where I had the turn and that is when I was implementing a new website for one of my education platforms and I tried some of the new models and I came on the podcast and I was like, Ian, it's good now. Like it's just good now. And
Matt Stauffer:
You
Matt Stauffer:
Yeah.
Matt Stauffer:
Yeah.
Yeah.
Aaron Francis:
you know, that didn't serve my needs at the time, right? Like me saying, hey y'all, AI is good and you should be using it is actually counter to my bottom line of let me teach you how to be a developer. So hopefully that like buys me some skin in the game credibility that I'm speaking against my own book here. But.
Matt Stauffer:
Yeah. Yep.
Aaron Francis:
Yeah, I definitely went through, you know, I definitely went through some stages of grief of like, man, I've been on my own for two or three years now and trying to make it work and it's been like up and down and good. And finally, when I start to feel like I get my footing underneath me, like this, this tsunami of change comes to our industry that decimates kind of the thing I'm doing first, which is like, let me teach you how this works. And people are like, I don't need to know how it works. I'll just ask the gods.
Matt Stauffer:
Yeah, I don't need it.
Aaron Francis:
And so in the beginning it was very much like...
Matt Stauffer:
Yeah.
Aaron Francis:
I kind of have this theory that you can fight or you can win. And in the beginning I was trying to fight because to my credit I think it wasn't very good and so I was correctly pushing back on like, y'all you have to continue to be super deep in the weeds. And then I just started to see a change in myself and that was the thing that I really, I stopped watching other people's videos. I stopped looking for really deep, nuanced,
blog posts and I just was kind of like, hey, let's see what this artificial god can do. And it did pretty well. And so I noticed that change in myself and I realized, okay, well I need to adapt to the game on the field, otherwise I'm going to lose. And so I just went all in and was like, what if I don't open my editor? As an experiment, what if I don't open my editor and I see what's possible? And it turns out a lot is possible. And so that was when,
That was when my internal calculus kind of changed. And that was also when I was like, hey, if I'm not gonna be like, of course I still look at the code, but if I'm not gonna be like hand typing it, what other opportunities for building things are available to me? Like I've been limited in the past by the tech stack that I've chosen, which is a web-based tech stack. And now I'm like, well, desktop is available to me. If I wanted mobile is available to me. If I wanted a different,
Matt Stauffer:
Yeah.
Matt Stauffer:
Yeah.
Aaron Francis:
web stack is available to me. And that has just been really fun. And I have found personally that I get a lot of joy out of the building and not necessarily the writing. And I get a lot of joy out of the architecture of the underlying code, not necessarily the typing it out. And so I'm arguing with the artificial gods of like, hey, this is a really stupid pattern, you shouldn't do this. And it's like, actually, that's a good point. But then I don't have to go and change
Matt Stauffer:
Yeah. Mm-hmm. Mm-hmm.
Aaron Francis:
2,000 lines of code, it has to do that. And I'm like, haha, I'm a genius. And so it's been like, I've finally come out of that doom loop of, okay, I'm still valuable in this new world. And a lot of the toil has been taken away from my work. I still work, but there's just less toil.
Matt Stauffer:
you
Matt Stauffer:
That's it. You mentioned builders and I do think that, you know, again, in this podcast, I'd like to cover lots of folks experiences like my friend Jordan keeps telling me it's funny watching you coders deal with this because as writers, we dealt with this two years ago. So like, I'm just watching you through the same stages. But I do notice that within our world, there has been it's not a perfect line, but there's there's the the experience of the builders. And I have been mentioning a lot of people, the people you know, who are most we say AI PIL, but the most excited about AI in our world tend to be entrepreneurs.
Aaron Francis:
Mm-hmm. Mm-hmm.
Matt Stauffer:
who often have a company of their own and have historically either had to do the development, even though that's not necessarily their number one goal, or paid other people to develop, and now they can get so much of their ideas more done, more of their ideas done than they could have before. And those are the AI-pill people, and I think the most negative on AI people, not always, but often are people who are like, I live for the code. I live for every line being exactly the way I want.
Aaron Francis:
Mm-hmm.
Matt Stauffer:
And I feel like that could become a caricature, because a lot of us who are excited about it are also like people who care about delivering quality products and good code. But are you noticing, you know, like that there's a part of yourself that is giving more of your time and energy, this builder part, as a result of being less of a coder? And do you feel like that's a loss, or is it sort of like, you know what, like I'm actually happy with what I'm doing with my time right now?
Aaron Francis:
Mm-hmm
Aaron Francis:
Totally.
Aaron Francis:
Mm-hmm
Aaron Francis:
I'm actually happy with what I'm doing with my time right now. So the things that remain for me as the human, I think in hindsight are the things I always enjoyed, is coming up with good patterns and good abstractions and good architectures and good theories of how the, you kinda like close your eyes and you're like, the castle should be built this way. And historically you'd have to go in and lay the bricks too, right? But now I'm just like, man, it should be like,
Matt Stauffer:
Yeah.
Mm-hmm. Mm-hmm.
Aaron Francis:
and then this thing should, and then it should call, and then that would be so clean. And then the AI is like, that's a very good idea, sir, let me do that. And that part still gives me a lot of joy, and that's on the coding side, but on the actual product development side, that part's always been fun. I want to see something come to fruition, and I have a vision for how it should look, feel, work, delight me, help me, whatever.
Matt Stauffer:
That's brilliant!
Matt Stauffer:
Yeah.
Matt Stauffer:
Mm-hmm.
Aaron Francis:
can get to that point much faster. And so none of the product side has been lost. Some of the development side has been lost. But turns out, not the part that I care about. The part that I care about is like, what does the system feel like? What are these structures? What are they gonna be? And I still...
Matt Stauffer:
Yeah.
Aaron Francis:
Not only get to do that, in my opinion I have to do that because the AI is very like, let me write a thousand lines of code. You're like, hey, eight lines would work. You should write eight lines. And it's like, it's a good point, my friend. And so I still get a lot of joy out of that part.
Matt Stauffer:
Yeah, yeah.
Matt Stauffer:
Yeah, yeah, no, that's really good. And before we move on to the next thing, I did remember I wanted to mention that the podcast you were talking about with your friend is mostly technical. We will link it in the show notes. Best podcast out there. I'm sorry to all the other podcast hosts, but it's my favorite. And it's just the perfect balance of technical, business, banter. I don't care about Seinfeld, but you guys bring Seinfeld on G, the way I actually enjoy. It's just, it's great podcast.
Aaron Francis:
Mm-hmm.
Aaron Francis:
I appreciate that.
Aaron Francis:
Yeah.
And for clarity, the name is in fact mostly technical. That's not only a descriptor, that is the proper name.
Matt Stauffer:
Yep, actually the name of the podcast and we'll link again. Okay, so you have pivoted. So you were a paid developer educator for other people. You went out on your own with a friend, you made the studios and you guys were producing educational videos and you were making your own educational videos and for other people. And then your focus was retightened as he went to get a day job.
and you were going to do database education. You spent a whole bunch of time thinking about that and then you saw the world of education shifting and you said, I need to pivot and I need to do some AI stuff. So what's your business right now? Where are you in your launch process? And kind of like if you were to describe to somebody who's never heard of what your next venture is, what are you doing?
Aaron Francis:
Mm-hmm.
Aaron Francis:
Yeah, so I retain Database School, which is my database education platform. And I think that will be valuable for a few more years. But I do think the pure, like, let's learn underlying concepts.
And this breaks my heart a little bit, like whether by, whether in reality or in perception, I think it becomes less valuable. So whether the models get good enough to be like, you actually don't need to know how to optimize this query, or people think the models are good enough, like to me, that difference is a distinction without a difference or difference without a distinction, they don't come to my site anymore.
Matt Stauffer:
Yeah, I was gonna say they're not paying you either way, so yeah.
Aaron Francis:
Exactly. And so I still in my heart love databases and love teaching and so that continues on for now. I think in two or three years will be an open question. But the new thing that I'm teaching is at faster.dev and that is how to, as an engineer or an aspiring engineer, how to use AI in a like, in a.
a thoughtful and rigid and like scientific and tested way such that you get these truly 20 to 50 times, you know, productivity improvements without creating a bunch of slop that's unmaintainable and you run headlong into a wall. And so as a like,
to be able to actually teach something, have to do it, right? And so from Thanksgiving on, it's just been this all day, all night, burning the midnight oil and the midnight tokens to like actually build products that I will actually release to the public so that I can learn workflows that I think I can personally vet. Because I can't just like scroll Twitter and be like, hmm, that seems neat, let me make a video about that. People do that.
Matt Stauffer:
Yeah.
Aaron Francis:
and I'm not making a moral judgment, I just can't do that as like, I can't put my name on it if I haven't seen it through to fruition to say like...
Matt Stauffer:
Yeah.
Aaron Francis:
hey, I have been using this for more than one freaking day. I can't do it. It's been very difficult because that requires that I build 1,000 things and throw away 994 of them so that I can figure out these paths are all dead ends, these products all suck, this workflow keeps coming back to me as I'm building these things. There might be something here.
Matt Stauffer:
Yeah.
Matt Stauffer:
Yeah.
Aaron Francis:
that, let me distill that and then let me teach that. so faster.dev is the distillation of everything I've learned building products with AI and on the platform, I mean, you'll follow along as I like actually build and release products and hopefully get people to pay for them because that keeps like that keeps skin in the game for me. I'm not just like a hype boy who can just say, a new model changes everything. Like I actually have things I got to do. And so I'm really trying
Matt Stauffer:
Yeah, yeah.
Aaron Francis:
to put a focus on working people, not people that are just hanging out for points on the internet.
Matt Stauffer:
Okay, wow. Yeah, I talked about the maximalists and one of the things you called out here that I really appreciate is that there's a lot of people in developer education prior to AI whose job it is to create content and so they don't have the ability to stop, and discern before they create the content. So, and then you've got a lot of people learning who are like, oh, well, this person made a video about it, therefore it's vetted and I should be used and everything I need to learn is in there and I'm like.
Aaron Francis:
Yeah.
Aaron Francis:
Yeah.
Matt Stauffer:
No, the reality of doing the thing really kind of is the best teacher. One of the things you mentioned before was that you have a product that you're going to release, but by the time this comes out, will have released. And it's in languages that you're not even familiar with. Tell me a little bit about the difference of going from being like an expert in a specific tech stack, building, knowing not just like knowing how to build, but being educator level, knowing to build a thing.
Aaron Francis:
Yeah.
Aaron Francis:
Mm-hmm.
Aaron Francis:
Yeah.
Aaron Francis:
Yeah!
Matt Stauffer:
And now you're building something and potentially about to release it in a language you don't know in an environment you're not familiar with in code that you have not even read.
Aaron Francis:
Mm-hmm. Yeah, it's a wild feeling, to be honest. And a big part of me feels like unleashed. Like, I can do anything. And I think anyone is capable of doing anything if they try hard enough, but I think having, you know...
what is it, if I count from when I first picked up a HTML book, we're at like 23 years of building software, you know? Having all of that knowledge turns out is pretty transferable once you get past the hurdle of I don't know the syntax here, I don't know the library here, I don't know the language here. And so,
Matt Stauffer:
Mm-hmm.
Yeah.
Aaron Francis:
You start from first principles of, all right, I'm going to build a desktop app, which is what we're talking about. I'm going to build a desktop app.
In my mind, I can see, and every engineer understands what I'm talking about, I can see the structure of how this thing should be put together. Not the UI, like the actual, it's a little memory palace of code, and like, this part should handle these concerns, and over here. And so you just go and you start talking to, I typically start with Claude on the web, so I'll go to Claude at .ai, and I'll turn on my little voice to text thing.
Matt Stauffer:
Yeah.
Matt Stauffer:
Yeah.
Yeah.
Aaron Francis:
And I'll say, all right, here's my plan. I want to build this thing. Here's how I envision the thing we're talking about as a desktop app to handle your local development stack. And so I'll say, here's how I envision the project and processes to be managed. Here are things we certainly need to consider cleaning up these processes. We need to make sure that we have these sorts of fail safes. And I just talked to it for a long time. And you can also ask it.
Matt Stauffer:
Mm-hmm.
Aaron Francis:
ask me questions that I haven't thought about. And so it becomes this little interview style, which is really nice. And then you take all of that learning, and then what I do is I send it off on what's called a deep research task, which you can trigger from Cloud AI, Gemini on the web. Many of these tools have deep research.
and I'll say like, hey, here's what I'm planning on building. I am planning on building it in a framework called Tari, which is Rust on the back end, and I wanna use React on the front end. Here is my description of the problem and how I think it should be architected. I want you to go do deep research and come back with some more ideas, thoughts, questions, plans, et cetera. And so we just kinda go round and round on that for a while, and then you take the resulting doc
which is like a markdown file and you download it and you put it onto your desktop or onto your machine and Then you open Claude in the console and say like hey, I've written you just lie to it You say I've written I've written this amazing document I want you to look through it and start to plan out some the bones of some of this structure and one thing that I have learned is like If anyone tells you they one-shotted something and it's something that looks pretty good. They're just lying to you
Matt Stauffer:
you
Aaron Francis:
Like it's just a lie. It's just not possible. And so one thing you have to do in my opinion is you have to lie to the AI. You say, I just want you to do this part. I don't know what we're gonna do next. You know what we're gonna do next, but you want the chunks to be right because you wanna be building on a firm foundation as you go. You don't wanna slop it together in an infinite loop that runs overnight with no oversight and then you come back and you're like, I don't know, looks good to me.
Matt Stauffer:
That's not possible.
Aaron Francis:
Like you need to like it's kind of like a gardening metaphor where you let it overgrow and then you trim it back and you're like great now We're in a good spot. We had a few we went off on a few Tangents a few things got out of control. Let's chop all those off clean it up. Let's lint it Let's test it. Let's make sure that everything's tight and then let's let it go crazy again So it's kind of like this organic expand contract thing and even if you know like and then it should do this and then
Matt Stauffer:
Mm-hmm. Yeah.
Matt Stauffer:
Yeah.
Aaron Francis:
it should do this and then it should just tell it, hey, I just want you to implement this. Even if like,
Sometimes it's like I have the like the mental structure of this entire code base in my head And I'm going to tell it to set itself up for success in the future without telling it We're gonna work on that thing in the future But you're like I want you to set this up in such a way that we can do X Y and Z and it's like sure I'll do that because you know later you're gonna do X Y and Z and so this whole like This is where I think some of the some of the hype gets a little out of turn is
Matt Stauffer:
Yeah, yeah.
Aaron Francis:
hey, just write a giant document and then send 100 agents to work on it overnight. And I'm like, I don't know anybody that has done anything that is valuable with that yet. And I could very well be wrong. The way that I work is like, have one Claude do it, check it, have one Claude do it, check it. And so like, am I moving 20 to 50 times faster? Yes. Am I moving a thousand times faster? No.
Matt Stauffer:
Yeah. Yeah. Yeah.
Aaron Francis:
That's okay, I'm fine with that. And so that's kind of the process that I worked through it. And I'm watching the code and I'm like, hey, this part feels brittle.
What if we introduce a new data structure here that unifies these code paths in a language that I don't know, but that doesn't matter, because I know what code paths are, I know what data structures are. And you know, like after you chase a bug in a circle twice, you're like, there's something fundamentally wrong with how we're doing this. This should not be so brittle. Let's step back and assess, like, what is our underlying structure here and why is that wrong? And that's something only an engineer can do for now.
Matt Stauffer:
Yeah.
Right. Yeah.
Matt Stauffer:
Yeah. You just led my brain so many directions. was trying to think. So one of the first things, my next question I had queued up for you is a little different now, but I'm going to ask it anyway. What I wanted to ask was, you mentioned a lot of these outcomes can end up being AI slop, right? That's what we talk about in the code world, where we're basically just saying, you just end up with really bad code that technically does what it's supposed to do, but it's brittle, it's insecure, it leads to people's...
Aaron Francis:
Mm-hmm.
Matt Stauffer:
driver's licenses ended up on a public thing, just like that iOS app happened, right? Like it's just, it's not actually good code. And so what I was gonna ask is having written something through AI where you don't write the code, you don't add the code, it's in language you don't know, how do you avoid making it slop? I think you kind of answered that, but I don't know if you wanted to add onto that anymore. I mean, it sounds like you're very actively involved. Is there anything else you want to say about that?
Aaron Francis:
Mm-hmm.
Aaron Francis:
Mm-hmm.
Aaron Francis:
Yeah.
Yeah, I'll say for sake of clarity, I've built an entire desktop application without opening my IDE one time. And so, like, I'm fully in it. And the desktop application manages running processes and gathers output and manages ports and manages security. so, like, it's not just like a little notes app or something. And I have found
Matt Stauffer:
Yeah.
Matt Stauffer:
Yeah.
Aaron Francis:
I think the best meta principle is kind of the do one step at a time and call it back. And then the other thing is you get this sense, and I think this is new in the AI first world, you get this sense of when something feels brittle. And we never got that sense before because we would just write it such that it wasn't brittle, right? But now you get to this point where you're like, wow.
Matt Stauffer:
Yeah. Yeah.
Aaron Francis:
I fix one thing and this other thing happens and then I fix the other thing and that other thing happens. Like, hang on. And the trick there is you have to hang on. Like you can't just, there are a couple of things you could do. You could power through and just write if statements until the cows come home and be like, great, I fixed it. You know, it's.
Matt Stauffer:
Yeah.
Matt Stauffer:
handle the conditions. Yeah.
Aaron Francis:
30 if statements, but none of the bugs happen anymore. The other thing you could do is say, I knew it, AI sucks. Those are the two things you could do. And I think the third thing that you should do is say like, hey, why don't you, Claude or whomever, why don't you draw me an ASCII diagram representing our code structure here and what's going on. And it'll draw you an ASCII diagram and you're like, yeah, that's insane. That is bat.
Matt Stauffer:
Yeah.
Aaron Francis:
crap insane, like of course this doesn't work. And then you say, all right, I noticed that we have two code paths, one for if this is set and one for if this isn't set, and we're doing a lot of if statements all the way through the code. Why don't we introduce a unified object that handles both of those things and then we eliminate, we just.
Matt Stauffer:
Yeah.
Aaron Francis:
fundamentally get rid of the problem. We don't fix it, we just make it go away such that now this object is responsible for everything throughout the code and we're not doing these switches to be like, if it's this but it's not that and it's that, then let's show this. You're like, that's never gonna work. And so I think there's a big part of like, there's a big part of feeling when something,
is harder than it should be and thinking, okay, I'm a developer, this is not that difficult. And instead of screaming all caps, fix the god dang thing, you say, hey, tell me, put it in plan mode and say, I don't want you to do anything, I want you to tell me, what's our layout, what's our structure, draw it out for me. And it'll draw it out and it becomes abundantly clear that it's absolutely insane and then you say, all right, fix it.
Matt Stauffer:
Yeah. Yeah. It's funny because I joked when people started asking me about AI stuff, I got asked that so often, that's what led me to make this podcast, but people started asking me, and one of the first things I said was, I'm not worried about my career because we've spent so much of our time fixing up bad code from offshore work and bad code from junior whatever, that like, we're just gonna be fixing up AI code, whatever. And I said it jokingly, but it also is true.
And what you just mentioned there, that experience is very similar to the experience I've had when I'm responsible or my team is responsible for cleaning up work done by people who aren't experts. You know, we, we, we get a client and said, I paid some offshore team $10,000 to build this a hundred thousand dollar project for me. And now it's everything's falling apart into security holes. And it's the same I'm looking at. like, there's 17 if loops, your whatever else it ends up being or control structures. It's the same of like, when you don't have this level of engineering knowledge, but you just have.
Aaron Francis:
Mm-hmm.
Matt Stauffer:
the brute force like attack the problem until it gets there. What do you think, especially as an educator, as a self-taught developer, what do you think the process looks like in our industry for people developing this skill set in the world of AI? Do we know? Because that's the thing that you developed over 23 years, right?
Aaron Francis:
Yep.
Aaron Francis:
Mm-hmm.
Aaron Francis:
I don't, yeah, yeah. I don't know if we know. And I'm very sensitive to it. Because I am a self-taught developer. And I did get...
I did break into this world because I was able to learn things and I was able to start at the bottom of the ladder. If the bottom of the ladder doesn't exist anymore, what do we do? One thought that I've had is, there well on faster.dev, will have a boot camp that is like, here are computer science fundamentals, but at a level that
Matt Stauffer:
Yeah.
Aaron Francis:
matters today because I think all boot camps up until this very moment have been focused on here's how you write a React component and that has been really valuable for a really long time and I don't think it is anymore. But what I think is very valuable is you need to learn what state management is. Is stuff like
popping up and going away in your application, what is your state management? And they're like, what's state management? I just have it write it down. And you're like, okay, let's talk about state management. Let's talk about state machines. Let's make impossible things impossible. And those are the things that I think people will need to learn, not necessarily at a syntax level. Like, we don't need to learn ZooStand, or we don't need to learn how to introduce
state for state machines, but you do need to know state machines exist, here's how you do transitions, here's how you do impossible states, here's why they're valuable. We need to know what atomicity is, like a thing happens or it doesn't happen. We need to know what idempotency is, like it happens once.
Matt Stauffer:
Yeah.
Yes. Yep.
Aaron Francis:
And only once, and you can do it again, but it's not gonna happen again. And so I think those are the things that people are going to need to learn from a level that is very different than in the past. We can show them code examples, because it's a helpful illustrative point, but I don't think they need to know necessarily. Here's how you introduce an atomic counter in Lua for Redis. It's like.
Matt Stauffer:
Yeah. Yeah. Yeah.
Matt Stauffer:
Yes.
Aaron Francis:
I'm just gonna ask the gods to do that, but I need to know that that concept exists and I need to know the category of problems that that concept solves. Because if you don't know that that exists, you're gonna run in circles because the AI is never gonna suggest, well least not yet, it's never gonna suggest some of these things because it's trying to please you to finish the task and adding an if statement is way easier than refactoring to a state machine. And you have to be smart enough, not smart
You have to be informed enough to know state machines exist. And I remember that this problem feels like something a state machine might solve. So let me ask the AI about it. And the AI will be like, ooh, that's a great idea. It may take four hours to implement. Do you want to do it? And you hit enter and it's done in two minutes. You're like, okay, all my state transitions are actually correct now. And so that's what I think the future of education for people
getting into development is going to be and I think the I think a focus of the future of development education could or maybe should be less in like here's how you get a job at a big co and more on hey
your dad is in the window and blind industry, why don't you go talk to your dad and use these things that you've learned how to build and see if like the windows and blinds industry still runs on Excel spreadsheets? Because at the top, like Anthropic and OpenAI, they're coming after...
all, they're coming after Salesforce, right? But Salesforce is never coming after the local property tax company in Dallas, Texas, right? Salesforce has never reached down that far, they never will, it doesn't make sense. But if you know, if you know through your friends, your uncles, your cousins, whatever, if you know that window washing, they run on paper and the paper gets lost in the guy's truck because he's running from job to job, there's an opportunity to come
Matt Stauffer:
Yeah. Yeah.
Yep. Yeah.
Aaron Francis:
and be like, I can actually build you an app.
and an iPhone app that'll keep your guys on track and make you a bunch more money and you can get rich doing it too. So my hope is that in the future there will be fewer behemoth software firms and there will be a proliferation of mom and pop software firms that are solving problems that historically nobody wanted to solve, which is why we invented Excel, because it solves every problem. But historically, Salesforce doesn't want to come down to the three person operation.
Matt Stauffer:
Yeah.
Matt Stauffer:
Yeah.
Matt Stauffer:
Yes.
Mm-hmm.
Aaron Francis:
operating in Waxahachie, Texas, but your little nephew might, and it might make him or her 60, 70, 80 thousand dollars a year, and it might make your business 30, 40 percent more profitable, and that helps everybody. And so my hope is, yes, we'll have a few big AI firms, because that's terribly expensive, but we'll have a ton of mom and pop software shops, because all of those niches can finally get filled. That may be optimistic, but that is my hope.
Matt Stauffer:
Yeah.
Matt Stauffer:
That's the dream. Well, you definitely made two groups very happy in everything you just said. The first was the computer science professors. And obviously we're talking about IT specifically and for folks who are listening who are in the broader business world. I still think there's an element here of in our industry, you've got the big theory, you've got the architectural systems and the theory, and then you've got the practice. And the practice has become more and more the focus of education for our particular niche for the last while, because it's like,
You could learn all those things, but, you can go to a bootcamp. Bootcamp inherently means learn the, that on the ground application of specific lines of code and specific languages to the exclusion of four years of training of theory. And I think that what you just said about kind of bootcamps falling apart and what they're really gonna need to learn is the basics. I'm like, there's a lot of computer science professors are gonna go, that's what I've been saying all along. And I think they're increasingly right with this change.
Aaron Francis:
Hmm?
Matt Stauffer:
I think the other group that you're going to be making happy is the Laravel world, which for those who don't know, that's kind of the world that we're in because we've often been about enabling individuals to ship software rather than allowing individuals to join a team of 600 where they are a tiny piece of a cog. And I'm not saying you can't use Laravel for, you know, 600 person teams, but I do think that it really, you know, systems, tools, frameworks, people that allow an individual to create something are really excelling right now.
because that individual can now say, great, so I just tell my AI use Laravel and it's gonna do the whole thing the whole way through versus my AI has to orchestrate 17 different systems because this was never built to be run by one person. And granted, it's probably easier to do those 17 systems as one person with AI than it was before, but I still think this is just like this match made in heaven of full stack batteries included and an AI, you can really go far with those.
Aaron Francis:
Yes.
Aaron Francis:
Mm-hmm.
Aaron Francis:
Correct.
Aaron Francis:
Mm-hmm.
Aaron Francis:
Yeah, all the old head funny daddies like you and me, all the Laravel and Rails folks are finally getting their day in the spotlight because, you know, Laravel and famously Rails were built to be the one person framework, which means it includes all the way from the database, all the way up to the front end, which includes jobs, queues, emails, everything. And so now,
Matt Stauffer:
Yes.
Matt Stauffer:
Yeah.
Matt Stauffer:
Yeah.
Aaron Francis:
This is literally what I'm thinking when I talk about the 17 year old in Waxahachie, Texas, which nobody's ever heard of, which is why I chose Waxahachie.
Matt Stauffer:
Nope.
Aaron Francis:
they can sit down with Linda in the front office of their local firm, whatever that may be, and she can describe the process, and that little kid can build out a full web application from soup to nuts that does Linda's job so that Linda can do more important things, right? And so Linda now becomes like, she becomes the superhero because she can orchestrate everything because this kid has taken these disparate data sources and plugged them into a single application
Matt Stauffer:
Mm-hmm.
Aaron Francis:
that is well tested and hosted and managed and it's like.
great, now the economy is becoming more efficient and people like the kid and this local firm are becoming more efficient and more profitable and life gets better and easier. And yes, it is like a hopeful dream, but I don't think it's too far off of reality. I mean, if anyone can take a 20-year-old framework and apply alien intelligence to it and then go talk to somebody that nobody's talking to, that's a pretty good setup to have some sort of
Matt Stauffer:
Mm-hmm. Yeah.
Matt Stauffer:
Yeah.
Aaron Francis:
economic value come out of it that is captured by these individuals.
Matt Stauffer:
Yeah.
So I want to make sure that one of the things I do in this podcast is I never allow myself to go too far in one direction, the doomer, the gloomer, or the maximalist. And I want to poke on a couple things that you've said, not because they're wrong, but because I want to make sure we're kind of looking at the broad thing. And I think you have a broader perspective than we've been able to talk about so far in this podcast. So you said two things. One of them is a 17-year-old writing software that runs a possibly multimillion dollar local companies, operations, finances, everything like that. And the second one, as you mentioned a couple of times, being 20 to 50 times faster. And I know listening back
Aaron Francis:
Yeah.
Aaron Francis:
Mm-hmm.
Aaron Francis:
Mm-hmm.
Matt Stauffer:
Those are gonna be the two things that get the most pushback from the people I'm not gonna say do-mers, but people who are more critical Programmers who are using AI as a part of their day-to-day job and saying I see a 5 % improvement, know And so I want to kind of push on both of those So let's start with the 20 to 50 X 1 because I think there's specific implementation details of how you're using it but also what you're doing while you're waiting for it to run and You know like your relationship with AI that that make you say 20 to 50 when somebody else might say I'm seeing a 5 % to 10 %
Aaron Francis:
Mm-hmm.
Aaron Francis:
Mm-hmm.
Aaron Francis:
Mm-hmm.
Matt Stauffer:
So imagine you've just been pushed back on that. Talk to me more about this 20 to 50 time speed improvement that you're saying you're getting from using AI.
Aaron Francis:
Mm-hmm.
Aaron Francis:
Yeah, totally. mean, it's a reasonable pushback because it sounds insane. So while you're waiting for the AI to run, I think you have a couple of options. One is you work on another tab on the same project. I've never been able to mentally stomach that, if that's a metaphor. Like, I can't...
Matt Stauffer:
Yes.
Aaron Francis:
There's been a rare case where it's like, hey, let's work on this copy on the homepage and let's work on this backend implementation. And I'll dangerously pop open two tabs in one git checkout and I'm like, good luck everyone, hold on tight. There are people who will check out the same project in a specific technical way that's unimportant. They'll check it out five times and work on five things in one project.
Matt Stauffer:
Yep, and just hope they don't conflict too much,
Aaron Francis:
Depending on the size of the project, guess that's doable. I don't know that my little brain can handle that. That's not something that I do. What I do is...
Matt Stauffer:
Same, yeah.
Aaron Francis:
I've got multiple projects going on. so things that like, and whether that is, you know, whether that is money making ventures or just open source or just, and this is a huge category, tools to make me go faster on other things. And that's not like, let's build AI tools that are AI tools to build AI tools, right? That's stuff like, hey, let me build out a newsletter builder, right?
Matt Stauffer:
Yeah.
Matt Stauffer:
Yeah.
Matt Stauffer:
Yeah.
Aaron Francis:
newsletter every week and it's like, let me build out a thing where I can have like a block editor and a newsletter that's built into my personal CMS and doesn't make sense to like go out and buy somebody's software and like this is a tool that makes my life better. So I always have a couple of, you know, a couple little things that I'm tinkering on and so.
Matt Stauffer:
Yeah.
Aaron Francis:
I will have one Claude working on project A and it's like, all right, I'm talking to it, I'm talking to it, I'm talking to it, I feel good, go. I'll switch to project B, see where it's at, scroll back up to see what it did and see what code it wrote, because I'm not gonna open my editor. And so I scroll back up to look at it and I'm like, wow, you're not very smart.
Scroll down and be like, all right, hey, we did this. I see what you were doing. I think the better thing to do would be this. Enter plan mode, hit enter. It's gonna churn for five minutes. Then I hop over somewhere else and I do something else. I also will have Claude AI open on the web and I'll have it like doing some deep research. One thing that I often do,
is I'll have Claude on the terminal. I have a skill for it called DeepStack. And what it does is I'll invoke DeepStack and I'll say, am like, I'm running into this issue.
I'm running into this issue with scroll bar overlays on Safari, like just the worst of the worst stuff that you don't want to do. I need you to write me a prompt for deep research. what deep stack does is it looks through my whole code base and it says like, hey, we're using this library, this version, this library, this version. Here's the DOM layout. Here's the issue that we're running into. And I just yoink and take that into Claude.ai on the web. And I hit a deep research.
and goes off for 30 minutes searching old Stack Overflow and Reddit and it comes back with like, oh yeah, this is because of WebKit 25432 and it's like.
Aaron Francis:
All right, great, let's take that back. And so I'm always kind of like orchestrating a bunch of different things. And the way that my brain works is I'm okay if a bunch of different things are typically different domains or projects. I get wrapped around the axle if I'm trying to do four things in one project because guy number A may have just done something that affects guy number C and I'm like.
Matt Stauffer:
Yeah.
Matt Stauffer:
Yep. Yeah. Yeah.
Aaron Francis:
shoot, like we're mid-flight with so much stuff I can't handle it. So copy changes and backend, little tools and different projects, open source, fixing my development environment. Sometimes I'll go in.
I'll go to the root of my system and open Claude and be like, hey, when I open my terminal, takes a second and a half to boot. Can you fix it? And it looks through and it's like, it looks like you're using NVM and NVM loads really slowly. You should switch to FSN or whatever the other one is. And it's like,
Matt Stauffer:
What gives? Yeah.
Matt Stauffer:
Yeah. Cool.
Aaron Francis:
All right, cool, let's do that. And so just feel like I'm in this loop of improving things and fixing things further down the list that would have never gotten fixed because I do have this in-between time of clogs working on something.
Matt Stauffer:
Yeah, man fixing things that never what I can't fix. That's a whole podcast episode of its own and I wish we weren't coming up on time because it's like, my God, the things that I'm doing now that I'm building internal tools at Tighten for that I've wanted for a decade and I'm like, yeah, it's not worth it. I'm like, it's worth it now. There's also a whole conversation to be had about, I don't know if you have or are diagnosed with ADHD, but I do. I just had a conversation with a good friend yesterday talking to her about, hey, you know what, you know how you always have 17 different things running and you've got six businesses and you just started a new one? Might be ADHD.
Aaron Francis:
chewed.
Yes.
Aaron Francis:
Yep. It is.
Aaron Francis:
Mm-hmm. Mm-hmm.
Matt Stauffer:
One of the benefits I've always gotten from ADHD is the ability to abnormally easily context switch multiple times a day within very short periods of time. So I'll go to a business development meeting with somebody I've never met before and have a deep, thoughtful conversation with them where I have insights into what's going on with them. I come up with a plan for my team after I hear the call. I write the whole thing up, and then a minute later, I'm on somebody's annual review and I'm thinking about what their work has been for the last year. And then a minute later, I'm doing technical architecture for a new plan or reviewing something I haven't touched for three years.
because of my ADHD, I'm just like, that's fine. I don't mind that world. And I imagine, because one of the things you mentioned often is the way my brain works. So ADHD or not for anybody, I would say some people's brains are gonna guide them to interact with these tools differently. And this 20 to 50 X thing, I think one part of it is a function of.
Aaron Francis:
Mm-hmm.
Aaron Francis:
Correct.
Matt Stauffer:
Aaron being a constant context switcher, who is an entrepreneur who also has a million things going at once. If you were to say, I'm sitting down on a single code base and working on it or a single writing project, whatever else, you wouldn't be saying I'm 45 times faster on this one code base. You're talking about your whole output. Am I understanding that right? Yeah.
Aaron Francis:
Yeah, yeah, yeah, totally.
even if you scope it down, like let's just scope it down to solo the desktop app. I've been working on that since Thanksgiving. And so it's not like, and where are we now? We're in like early February. So it's been a couple months, right? And so again, back to the one-shot thing, that's a lie. Anything that like, anything that you, I'll speak for myself, anything I wanna put my name on is gonna take time, because I want it to look good and I want it to be right and I want it to feel good.
Matt Stauffer:
Mm-hmm.
Matt Stauffer:
Yes. Yeah.
Aaron Francis:
Could I have gotten this project done in the two months pre-AI? I don't think so. Even if this was the only thing I was working on, I don't think so. I think it would have taken me a lot longer and importantly, I think I would have given up. And that's a big part of it. Yeah, that's a big part of it is like, boy, this part is gonna be a slog. What if I did something fun instead and started a new project? And now,
Matt Stauffer:
Yeah.
Matt Stauffer:
Yeah, or maybe never started.
Matt Stauffer:
Yeah.
Aaron Francis:
that part, like that feeling has not gone away, but the bar is so much lower, where it's like, it's gonna be a slog to ask Claude to do this. And I'm like, that's actually okay. That's not that big of a deal. And so, yeah. And so there's a big human emotion aspect of it, of like,
Matt Stauffer:
Yeah.
Matt Stauffer:
I'm like it's gonna be a slog. Thank God. I'm not the one writing the code
Aaron Francis:
the things that were not too hard but were too freakin' tedious or like soul draining, those things are now on the table as well. And those contribute to my speed up, which maybe is not pure technical speed up, but is like human wall clock time speed up. Maybe this would have taken me two years, because I would have gotten discouraged so many times in the process, but now it's just like, brother, keep going, that's not my problem, you do it.
Matt Stauffer:
Yeah.
Yeah. Yeah.
Aaron Francis:
There's some aspect of that.
Matt Stauffer:
I have six more questions written down and 20 more questions I'd like to ask, but I've got, I said I was going to try and do 30 minutes, it's 45. I gotta a wrap 45. So I don't know if I'm having you back on or what. The good news is all of us can learn from you further at faster.dev. But I have the two questions I want to end every single one of them with. So the first one is, what is one thing you wish more people would understand about AI?
Aaron Francis:
That's right.
Aaron Francis:
I wish more people would understand that it can help you dream bigger. So I wish that people, I understand that it's a terrifying time. I've got.
four kids at home and a wife that works very hard but makes no money because she's a mom and an amazing one. And I don't have a W-2. If anybody understands it's a terrifying time, it's me. But my hope is that...
Matt Stauffer:
Yeah.
Aaron Francis:
I think maybe my fear is that people are gonna be so scared and terrified that they stick their head in the sand and then get walloped. And so my hope is that people will realize this is a once in a generation opportunity to maybe dream a little bit and think, what could I do? Like what have I always wanted to do and how can this enable me to do it? Because I think,
Many, if not all, many of those things are within reach now. Not easy, but within reach.
Matt Stauffer:
Yeah, that's really good. And my last question is if people like learning from you and want to keep learning from you, how should they follow you?
Aaron Francis:
You can follow me at faster.dev or on x.com, the everything app. am Aaron, A-A-R-O-N, D as in Daniel, Francis, Aaron D. Francis on x.com.
Matt Stauffer:
I have insisted on calling it Twitter and I will continue to do so forever, but if anyone ever says X, I hear immediately your voice. X.com, the everything app. So thank you because it makes me less annoyed when people call it X.
Aaron Francis:
It's the only way, it's just, it highlights the absurdity of it. To call it what they want you to call it is so funny that it's a good bit every time. So x.com the everything app.
Matt Stauffer:
We gotta laugh so we don't cry, right?
I love it. Well, thank you so much for hanging out with us. I want to end every single one with sharing two tips that people shared with us on the Internet about how they're using AI practically. And I'm not sure if I should let the guests go first or not, but since you're my first guest, you're going to hang out while I read them really quickly. And if that's awkward, then I'll let future guests go. So I just went on the Internet and I just said, hey, you know, what is a one way that you're using AI practically? And again, I'm not trying to.
Aaron Francis:
Perfect. No, I love it. That's great.
Matt Stauffer:
talk about all the weird edge casey things but like what actually changed your life so two people I think you actually know both of them first one is Jason Beggs sweetheart of the Laravel community said I've been rebuilding our stairs it was huge help calculating the measurements for the rise and run cuts on the stringers calculating the placement of the balusters figuring out the gaps and laying everything out much better than normal calculators because you can chat back and forth and ask good questions and I was like yeah
Aaron Francis:
Mm-hmm.
Aaron Francis:
smart.
Aaron Francis:
Mm-hmm.
Matt Stauffer:
I've done some, like I built some things using plans and still had to do all these crazy mathematical calculations so I can imagine. So Jason, thanks for the idea. And then Alex6 says, using it for coloring pages has been a massive hit in our household. We still like to get the kids coloring books to use, but sometimes they want to color a very specific type of picture for a game they're playing or whatever else and I could just whip them up into a man. Have you done that, Aaron? Have you made stuff for your kids with AI? Really? Okay.
Aaron Francis:
Mm-hmm.
Aaron Francis:
Mm-hmm.
Aaron Francis:
we've got so many coloring pages of dinosaurs riding fire trucks. Yes. It's like the stuff that honestly, honestly, the companies should be producing because that's an awesome coloring page. But yeah, no, we've done it. We've taken pictures of like, you know, mom and daughter and said, make this a coloring page. And it's like turns them into princesses. A hundred percent. It's perfect.
Matt Stauffer:
that's brilliant.
Matt Stauffer:
huh. Right. Yeah.
Matt Stauffer:
cool.
I love that. Okay. Well, thank you, Jason. Thank you, Alex and Aaron Thank you so much for hanging out and sharing with us. And I look forward to seeing kind of where this journey takes you,
Aaron Francis:
Yeah, thanks for having me. I'm honored to be here.
Matt Stauffer:
Awesome, for the rest of you, we will see you next time.