Aaron Francis

Things Aaron Says on Podcasts

Software developer and database educator, co-founder of Try Hard Studios with Steve Tenuto, making video courses on SQLite, Postgres and MySQL for developers. Built up the PlanetScale YouTube channel before starting the studio. Started his career as an accountant before moving into software. Co-hosts Mostly Technical and hosts Database School, and co-hosts Token Town. Regular guest on shows including Whiskey Web and Whatnot, The Panel, The Nathan Barry Show, The Bootstrapped Founder, PodRocket, Remote Ruby and Syntax, mostly talking about databases, developer education, and building a business around technical content.

Where to Find Them

Aaron Francis hosts and writes Mostly Technical , hosts Database School and hosts and writes Token Town . They have also been a guest on Whiskey Web and Whatnot (6 times) , The Panel (3 times) , The Nathan Barry Show (2 times) , The Bootstrapped Founder (2 times) , PodRocket (2 times) , Remote Ruby (2 times) , Syntax (Wes Bos & Scott Tolinski) , The Business of Laravel , North Meets South Web Podcast , Laravel Podcast , Pragmatic AI with Matt Stauffer , Indie Bites , Software Huddle , Talking Businessly , Scaling DevTools , Build Your SaaS , The Art of Product , Justin's brain and Offsite .

Recently: “154: Ian Laravel New'd Again!” on Mostly Technical (October 2026); “153: 13 Billion Tokens” on Mostly Technical (October 2026); “152: Feel The AGI” on Mostly Technical (October 2026); “152: Feel The AGI” on Mostly Technical (October 2026); “151: Tokenmogged” on Mostly Technical (September 2026); “151: Tokenmogged” on Mostly Technical (September 2026).

What They Said

“What I want is I don't want you to build AI into your app, I want you to make AI usable from outside your app. I want you to give me the best CLI, API, MCP in the world, and give me access to my data, and I'll take care of the rest. That's what I want as a very stupid developer.” — Aaron Francis, Mostly Technical

Francis, while admitting he doesn't represent the market, on how B2B software should handle AI. Ian Landsman pushed back that most people work inside one app all day and won't run their own AI setup outside it, so products will need both an MCP and built-in AI.

Mostly Technical · 2026-09-29 Permalink → Listen →
Mostly Technical Around 58:53 into the episode
Aaron Francis

third-party agent, HubSpot, increase, own agents, a long time CRM, extra agent, API, another niche product. Yeah, so everyone is uh raising prices. Yes, yeah,

Ian Landsman

…is the time to raise the prices, and then you know, who knows? Maybe you won't be able to sustain that raised price. Maybe in three years, right? The AI is everywhere, and there's tons of downward price pressure. Then you'll cut prices, and that's fine. But real now, it's the time to raise prices.

Aaron Francis

You know what I want as a very stupid developer who's not representative of the market at all. What I want is I don't want you to build AI into your app, I want you to make AI usable from outside your app. I want you to give me the best CLI, API, MCP in the world, and give me access to my data, and I'll take care of the rest. That's what I want as a very stupid developer. That's exactly what I think this is the

Ian Landsman

standard developer take, right? This is what you see. Everybody, they're like all day on Twitter. It's like, oh, I'll just use the MCP and whatever, and that's fine, and you will. But that's not how most people are. Most people are in this app all day, they don't have their own local thing, or if…

Aaron Francis

Bob. He should get a dream a little bigger. He's gnawing, but he's not.

Ian Landsman

So you got, so I think all these B2B apps are going to have both, right? You're going to have your MCP and you're going to have whatever comes next and all this stuff, but you're also going to have your internal AI stuff because that's where honestly most of the people using it are using it. And so…

Speaker names from our own diarization · position estimated from where the line sits in the episode
“Out of the hubris of my heart, I said, but I'm a software developer and you're a property tax company. I don't know if you can afford to hire me. And they're like, brother, we can afford to hire you. So, first learning, real businesses do have money. It's not just Silicon Valley.” — Aaron Francis, Mostly Technical

Francis was the first and only programmer at a Texas property tax firm, which he helped grow from 16,000 to 72,000 properties before leaving as COO. He calls it one of the most valuable experiences of his career, because the software only had to serve the business.

Mostly Technical · 2026-10-06 Permalink → Listen →
Mostly Technical Around 32:48 into the episode
Aaron Francis

One Aaron Francis.

Ian Landsman

Hey, look at that. I didn't go to school

Aaron Francis

…my brother-in-law's a programmer. That could probably help. And so I came in first on a contract basis and took their existing systems and made them more robust. Still Google Sheets, still Zoho. I just did a bunch of Google Sheets like app script, which is Google's like macro thing. And I made their processes way better. And then they were like, hey, why don't you come work for us full time? And out of the hubris of my heart, I said, but I'm a software developer and you're a property tax company. I don't know if you can afford to hire me. And they're like, brother, we can afford to hire you. So, first learning, real businesses do have money. It's not just Silicon Valley. Then I spent the next four years building internal software that we scaled from 16,000 to 72,000 properties under representation. I hired a couple of programmers. I left as the COO. So I moved all the way up to like running the company. And it was one of the most valuable learning experiences of my professional career. Because you see, like, you see what matters and what doesn't matter. Like…

Ian Landsman

You know what I mean? Well, that's always the best way to get a job, to be fair, but also, but like having some kind of connection, obviously. But yes, you have to get through now. There's whatever. That's a whole separate discussion of how you get in the door, right? So, okay, you got to get in…

Aaron Francis

how you get in the door. And second, like more, how do you even know that the window and blind company is out there slash needs help? You know what I mean? Like, they're not saying I need to hire a software developer. They're like, oh, our business is on fire. That's the thing where, how do you…

Ian Landsman

Well, I think that it's, I'm, I'm saying, even sort of a step farther than that. Not even just finding the company that's on fire, but I'm saying applying to jobs that are just explicitly manager jobs. Like, it's not even a programming job at all. There's no expectation of programming. It's like…

Speaker names from our own diarization · position estimated from where the line sits in the episode
“I do think the value of unit tests has always been less than the value of feature tests. We just didn't want to write end-to-end tests because they were kind of a nightmare to set up and run.” — Aaron Francis, Mostly Technical

Francis and Ian Landsman were complaining about slow test suites that coding agents run over and over. He now tells his agent to delete low-confidence tests and favour end-to-end tests, since the agent takes on the setup work that used to put people off.

Mostly Technical · 2026-10-06 Permalink → Listen →
Mostly Technical Around 1:14:36 into the episode
Aaron Francis

…expected outcome? Yes or no? That black box testing. That's what I want. I want the thing to work. Now, because of what Solo is and does, I have thousands of unit tests to make sure that like this mutex doesn't get poisoned. This, you know, when I parse this, it turns into that, that kind of stuff.

Ian Landsman

Yeah.

Aaron Francis

But I do think the value of unit tests has always been less than the value of feature tests. We just didn't want to write end-to-end tests because they were kind of a nightmare to set up and run.

Ian Landsman

Yeah, and that is the vast majority of my tests. Certainly on a web app, right? Like you don't generally, you shouldn't, I don't think, often have a ton of unit tests. Like I know people in the past probably did, but it's like, yeah, I'm not mutating a lot of data from A to B. It's like, I just…

Aaron Francis

Yeah, I think pest browser is probably the best way to go. And you, you know, even pest browser is still flaky because it's browser. Right. And so you just tell the agent, hey, we got to harden these flakes and it, it will.

Ian Landsman

Yeah, I know. It's also slow, though, too. You know, it's slow and the whole thing. Where

Speaker names from our own diarization · position estimated from where the line sits in the episode

Collections They Appear In