# 94: Full Sleazy Lock In  ## Episode metadata - Episode title: 94: Full Sleazy Lock In - Show: Mostly Technical - Owner / Host: Ian Landsman and Aaron Francis - Episode publish date: 2025-08-12 - Episode AI description: A lively debate kicks off with the differences between trips and vacations, highlighting humorous travel mishaps. The hosts tackle sponsorship strategies for tech podcasts and ponder the challenges of monetizing content. GPT-5 takes center stage, sparking mixed reactions about its capabilities. They also dive into personal anecdotes about driving adventures and family milestones, all while exploring innovative tech solutions and the quest for a personal life operating system to boost efficiency. - Duration: 01:27:10 - Episode URL: [Open in Snipd](https://share.snipd.com/episode/a1c84dc3-068e-4938-90e7-0f3890763d1c) - Show URL: [Open in Snipd](https://share.snipd.com/show/7ad45fb4-8aa6-49c2-83cc-bc2327d14b03) - Export date: 2026-02-17T23:31:33 ## Snips ### [Vacation Turned Medical Trip](https://share.snipd.com/snip/e6611d7c-2c53-41d7-9292-60a665cf5608) 🎧 00:20 - 02:31 (02:11) <iframe src="https://share.snipd.com/embed/obsidian-player/snip/5377e10b-5da3-424d-ac56-b8ecf509f938" width="100%" height="100" style="border: none; border-radius: 12px;" sandbox="allow-scripts allow-same-origin allow-forms allow-popups allow-clipboard-write" ></iframe> - Ian Landsman had two urgent cares and a hospital visit on this so-called vacation. - He concluded that when kids come, it stops being a vacation and becomes a trip. #### 💬 Quote > So it was two urgent cares and a hospital visit on this so-called vacation. > — Ian Landsman Ian Landsman on a family trip that became medical visits #### 📚 Transcript **Aaron Francis:** We're back. You're back. I'm back. Back vacation, guy. I'm back. Air quotes. Oh, no. That's a rough one, man. Yeah. one. **Ian Landsman:** Woo! We had a—I think on the last episode I told you about how we did urgent care. Right. And we did a doctor or hospital. And then the back half, Jamie got sick. So we're back to the urgent care. Jamie, what are you doing? So it was two urgent cares and a hospital visit on this so-called vacation. That's not a vacation at all. Yeah, it was quite something. Matt Stenson, who works with me at Uduscape, he has this theory that if your kids are with you, it's just a trip, not a vacation, which I think I buy that. I think that makes sense. I definitely need a vacation. But yeah, it was a trip. It was good, though. Brother got stuck in the sand. So that was exciting. A bunch of people getting stuck in the sand. There was like, I don't know, something about the wind and stuff and the thing. People getting stuck in the sand everywhere, even big trucks and stuff. I didn't have any trouble, **Aaron Francis:** luckily. Yeah, what was he driving and did you pull him out with the Rivian? **Ian Landsman:** So he was in a Volkswagen SUV, but see, I think it's all-wheel drive and you're really supposed to be four-wheel drive. **Aaron Francis:** All right. How many wheels is all? That's my question. I've never understood this. All-wheel drive, there are only four wheels. So what are we doing? That's I think **Ian Landsman:** the difference is a four-wheel drive vehicle can send, I can't remember if it's up to 25% of the power, maybe it's up to all the power to any of the four wheels. Okay. Whereas all-wheel drive vehicle only can send like 10 of the power to the back whatever it is like it can't just shove all the power to one wheel so you know it can't balance as well but uh wow **Aaron Francis:** that's very confusing yeah **Ian Landsman:** we did get them out um actually somebody else came behind him and pulled him out with my strap. So you still get credit. Yeah, I bought gadgets, baby. So I had the gadgets. Of course you did. I bought a strap, a toe strap, and I bought a collapsible shovel thing, both of which came in handy, so that was good. Yeah, the toe strap pulled him out. --- ### [EV Charging Matched Family Breaks](https://share.snipd.com/snip/98434dd3-0b2f-40ae-90b1-86c8f93b4f86) 🎧 03:59 - 05:11 (01:12) <iframe src="https://share.snipd.com/embed/obsidian-player/snip/bc6db8a4-c746-4674-bc67-37b728b1790e" width="100%" height="100" style="border: none; border-radius: 12px;" sandbox="allow-scripts allow-same-origin allow-forms allow-popups allow-clipboard-write" ></iframe> - Ian Landsman found EV charging perfectly matched family rest stops on a 12-hour drive. - He noted the car was charged by the time people finished bathroom and food breaks. #### 💬 Quote > By the time everybody goes to the bathroom and you get something to eat, the car is charged. > — Ian Landsman Ian Landsman on long EV drives and charging rhythm #### 📚 Transcript **Ian Landsman:** We are in an electric vehicle. **Aaron Francis:** Oh, so we're stopping every 14 minutes. Okay. **Ian Landsman:** So this, I mean, it's a whole other side little thing to talk about. But basically, that turned out so good that Jamie's like, we, you know, whatever. Her car's like paid off, whatever. She's been thinking if she wants to get a new one, whatever, blah, blah. She's thinking about getting the Rivian SUV because she was always scared about the long drives with the EV. And it went so good because here's the thing. You got a bunch of people. You are longer doing whatever you're doing on that rest stop than it takes for the car to charge. So not once did we have to wait for the car. The car was always done by the time people were done. It's not what I would have expected. Yeah. So by the time everybody goes to the bathroom and you get something to eat, the car is charged. It's fine. So it's totally fine. So that was great. It wasn't an issue. You know, you have to stop every three hours anyway. I have the smallest battery in my truck because I was like, whatever, it's once a year it's an issue. Yeah. But it was fine around town, you know, but I was like, I was worried about this trip because it's like the once a year where you're like, oh man, I wish I had the big battery. But it was fine. Like, I don't think the way down didn't cause any extra stops at all. And then the way home, we probably had one extra stop. We stopped three times. We probably could have stopped twice if I had the big battery pack. --- ### [Kids' First Real School](https://share.snipd.com/snip/9298d04c-f96a-4d32-ad1e-613d564a4faa) 🎧 05:50 - 07:57 (02:07) <iframe src="https://share.snipd.com/embed/obsidian-player/snip/b79ecffd-e61a-4371-bc6b-e0d2ad30609d" width="100%" height="100" style="border: none; border-radius: 12px;" sandbox="allow-scripts allow-same-origin allow-forms allow-popups allow-clipboard-write" ></iframe> - Aaron Francis got both kids into the same full-day 'real school' class and felt relieved. - He valued the neighborhood public school and immediate access to therapists and community support. #### 💬 Quote > This is their first real school thing. > — Aaron Francis Aaron Francis on his children's transition to full-day public school #### 📚 Transcript **Aaron Francis:** tomorrow. So on, uh, I think Southerners. **Ian Landsman:** It's crazy. **Aaron Francis:** know. I know. Um, yeah, I think on Friday or Thursday we had meet the teacher and it's just, you know, it's crazy. They're going to school. They're going to pre K. Okay. This **Ian Landsman:** is their first real school thing. Real school. **Aaron Francis:** Yeah. So they've been going to like mother's day out, you know like three four days a week but now we're going to real school with real teachers real lockers real you know thing cafeteria i know it's just it is uh it's weird makes me feel like i'm a thousand years old i managed to get the kids every day, Monday through Friday. Full day? managed to get them full day, like 7.50 to like 2.50 or something. So early. I'm going to be drop-off guy. I'm so happy to be here for this. This is a big moment. We got them in the same class, which is important for them. Because one of them is very helpful to the other um and so that that will be very good to have them both together so that you know if one of them needs help uh the other can interpret you know for the teacher um so that was very stressful but we got them into the same class because we got an email like the night before that was like hey here are your kids teachers and they're two separate emails with two separate teacher names. And we're like, **Ian Landsman:** oh. Always. You know, that's 100% of the time that's going to happen. There's no chance you're going to have the right thing you want happen. **Aaron Francis:** No, no, no. But fortunately, you know, it's a small neighborhood, you know, public school. And we sent them an email. And we're like, hey, here's the situation. We can get a doctor's note if we need to, but regardless, hopefully we can make this work out. And the next day they were like, oh yeah, sure, sorry about that. It's all fixed. It's like, whoa, whoa, wait, what? crazy. What's going on? I expected them to be like, oh, these parents are always with these problems and here are the educators and they were just like, oh, no, that makes sense. Sorry about that. Our mistake. Cool. And it's fixed. Yeah. So that's good. --- ### [Make Poll Voting Members-Only](https://share.snipd.com/snip/83345d23-0f3a-45bc-bb8c-321ca28fa39a) 🎧 14:00 - 19:58 (05:58) <iframe src="https://share.snipd.com/embed/obsidian-player/snip/ca3b8e7e-b276-4562-ad6f-2aa04a332e36" width="100%" height="100" style="border: none; border-radius: 12px;" sandbox="allow-scripts allow-same-origin allow-forms allow-popups allow-clipboard-write" ></iframe> - Make polls visible publicly but restrict voting to authenticated members to prevent bot abuse. - Use voting as a membership incentive to convert casual viewers into signed members. #### 💬 Quote > I think it starts as a member thing. > — Ian Landsman Ian Landsman on building polls as a member feature in Outro.fm #### 📚 Transcript **Aaron Francis:** right so i added this to the last mega episode and we didn't cover it obviously because it was a mega episode but i think outro needs a sort of like a sort of hot takes polls section where we can put in the comments, should Aaron get a golf cart? And everybody can go say, yes, Ian's an idiot on Outro FM. I think that's important. Totally agree. So I'm going to pause you there and say that this episode is brought to you by. Hey, that's called a tease. That's called a tease. Let's talk about polls. But first, this episode is brought to you by a new sponsor. Friend of the show. Friend of the family. Laracon AU 2025. AU stands for Australia. Level up your Laravel skills. Connect with brilliant people and experience two unforgettable days this November at Laracon AU in Brisbane, Queensland. Queensland, plural. That's interesting. Queensland, Queensland. You can find out more at Laracon.au, of course, organized by our very good friend Michael Durenda. So check them out at Laracon.au. Also brought to you by the Laravel family, Laravel Nightwatch. Nightwatch sees what other monitoring apps can't in Laravel applications. Take two minutes to set up the free plan right now. And before this episode of Mostly Technical is done, you might uncover rogue queries, misconfigured middleware, hiding in plain sight, or the dreaded N plus one. Find out more at nightwatch.laravel And also brought to you by Flare, by Spotsy. The all-in error tracking and performance monitoring for your Laravel application. You can find them at flareapp.io slash performance monitoring. All these links will be down below. you to our sponsors for making this possible. Sponsors this month. I love it. Lots of sponsors. Got to pay the bills. Got to buy a golf cart. All right. You like **Ian Landsman:** the polls idea. You like it. I like the sponsored golf cart idea. This is great. Sponsor the show so Eric can get a golf cart. Let's do it, people. Let's do **Aaron Francis:** it. Yeah, I'll claim it's putting food on my family, but really, I just want a golf cart. You **Ian Landsman:** get a little sticker on the side. It's like a NASCAR car, but it's a golf cart. Now we're talking. And your ad is seen in the little suburbs of tens of people in the pickup line. So imagine the ROI. You never know. You only need to get one. You need to get a single customer, most of you, to make it pay off. So it's fine. Because we're too cheap. That's just the reality. But it's a separate issue. All right. Oh, man. Phone's going on. Let me tell you, before we get to outro, I cannot wait. I can't wait for this new. The Google people have had this ahead of us. It makes me mad. But in the next iOS update, they're going to have the thing where you can turn on the AI to answer the phone and take a like pre-filter the message thing. And then just give you a text thing. It's like, hey, do you want to actually answer this or not? Because they say this is what they say they want to get to you for. Because I'm so sick of all these spam calls. I'm so sick of the spam calls. Sounds **Aaron Francis:** cool. It's not going to work very well, but it sounds cool. Tell me about it. No, it's going to be a disaster. It's Apple. It's Apple and AI. It's Apple and AI. They don't know how to do it. I **Ian Landsman:** don't even know if it's real AI. It's not doing anything AI-ish. It's just answering the phone and being like, what do you want? And then the person says what they want. it and says, do you want to answer this call? Have **Aaron Francis:** you ever in your life asked S-I to do anything and it got it right? No. I don't even think this is S-I **Ian Landsman:** I think this is like just, this is just like, this is ML. Maybe it's ML. Maybe **Aaron Francis:** it's ML. Maybe it's ML. They're good at the ML. They're **Ian Landsman:** not good at the **Aaron Francis:** AI. The mathematician's AI, ML. I will take that. **Ian Landsman:** All right. Polls, outro. I mean, I don't have a ton to say here other than that. I totally agree with you. One interesting thing, I guess, about it would be I was thinking about it as this is a type of. So there's like you're going to members can have posts, right? And send posts out and have public posts and whatever. So I'm thinking it's a thing that where you have a public post that's a poll and then i guess the only question would be do you have to be a member even a free member to actually vote in the poll which i kind of think yes otherwise you just have like the just any rando on the internet can vote as many times as they want like then there's no sort of you know it's like a Twitter poll. **Aaron Francis:** Kind **Ian Landsman:** of useless. Twitter polls even now you have to be logged in, I think. But I don't know. So I guess that's my other question. It being part of the member feature set, or is it like a freestanding polling thing? Which I think, I don't know. **Aaron Francis:** Why not both? Or either. both. Could be either? You could have a public poll. You could have a members-only poll. You could have a poll that splits it out by how members voted versus the public voted. Come on. Now you're just trying to let the scope creep. Yeah, do it. I think it starts **Ian Landsman:** as a member thing, but maybe I still think it has to be like in the feed of your public posts, which are like kind of like the same thing as the member posts. If it's public, it's just like it's not only members can see it. Everybody can see it. So in that same vein, I guess in that sense, a poll that is public, you know, it's like a public poll would just be votable by anyone. **Aaron Francis:** I think there's like weird spam stuff with just said a poll that is public would be a public poll. And I think that is like, I think I agree with that. I just want to go down and say, yes, I agree with **Ian Landsman:** that. Don't you hate these things, though? When it's just open to the internet, you know, you know. The first five minutes, it's just going to be, oh, there's a poll with seven million answers because some bots in there answering polls. Why? Why? This thing that's got no users, the bot's going to be there answering polls, but you know they're going to be doing it. So that's the part that annoys me. I'm **Aaron Francis:** fine with members only. I don't care. Members only is fine. I'm fine **Ian Landsman:** with that. think only, because it's also a draw that you should become a member. You should become a member. To vote in the polls, right? --- ### [Charge Per Member, Not Per Podcast](https://share.snipd.com/snip/c304244b-eee1-403b-b9f5-4748630997d4) 🎧 25:30 - 29:45 (04:15) <iframe src="https://share.snipd.com/embed/obsidian-player/snip/a53426ef-3a8f-41ac-9ae8-4953af264cb4" width="100%" height="100" style="border: none; border-radius: 12px;" sandbox="allow-scripts allow-same-origin allow-forms allow-popups allow-clipboard-write" ></iframe> - Offer the core app free and charge a tiny per-member fee (cents per member) so fees scale with audience size. - That pricing aligns costs with the value podcasters get from audience management features. #### 💬 Quote > Full free, all the features are free. You pay for members. > — Ian Landsman Ian Landsman describing Outro.fm's proposed per-member pricing model #### 📚 Transcript **Ian Landsman:** like it's just not like the amount to make it even useful amount of money to me in terms of like if this becomes a thing and is busy and requires like developer like you get if it becomes a thing like i need a lot of people at nine dollars a month to like make all that **Aaron Francis:** nine dollars is a mess right what's the minimum that's interesting to you in terms of either monthly uh all **Ian Landsman:** right so let's say let's want it i mean you know listen i mean even like 20 000 a month is interesting right but it still even is not that much money in terms of like once you put a developer on it well now like that plus the other baseline fees and stuff that you're uh you've eaten up all the money pretty much so like um i've been thinking about it as like 50 000 i feel like 50 000 not that day one, it's going to be 50,000, but the idea of like- 50,000 a month? The idea of getting it to $50,000 a month is where it's like, okay, like now it's worth, it's worth several million dollars. And that's sort of like, to me, the goal. Like if it can get to $50,000 a month, that's a big success for me. Like I consider **Aaron Francis:** that- Do you think there are 1300 podcasts that would pay you 39 a month for this no **Ian Landsman:** like yes and no right like i think yes in the world of podcasts there are but can i reach all of them to actually then get 1300 of them to pay 39 a month do you know what I'm I do know what you're saying. So what do you think the price is? I just think there's not a good price like that. So I was going, so I guess the way I was thinking about it, I think 39 to me is like almost the worst kind of pricing because it's like expensive. It's going to more than their podcast hosting. Correct. But it's also not, it's also nothing to like a podcast that makes a lot of money right like it's over we're not maximizing on the big fish and the little fish are going to be like ah this is costing twice what transistor costs me like you know that end of it so like that's where i've kind of and maybe this is wrong so i'm not i don't have all the answers here but i've been sort of pushing that pricing out of my mind a bit because i'm like i just don't know people are gonna pay 39 49 so i've been thinking about it more like 99 because it's like hey either you want to pay me because if we do some feature thing you just like the features and you're like i don't care even though maybe i'm not making money on it that kind of thing or obviously your bigger podcast you make money and you're like yeah it's worth 99 bucks because i'm way more organized than having 10 tools to do all the stuff um and then obviously you have the free tier on the other end i think you could do the like nine dollars a month five dollars a month twelve dollars a month maybe thing and just try to get the broad base there i think that part i don't love as much either because, again, it's like $9 a month. So you need like kind of a lot of customers. So that part's not super great. Although how many customers do you need? **Aaron Francis:** The question is how can you make it valuable enough that anyone would pay $99? $9 a month. It's **Ian Landsman:** not as bad as I thought I'd say. $9 a month. **Aaron Francis:** 5,555 **Ian Landsman:** repeating customers. That's a lot. Yeah. That's not good. That's not good. Yeah. Yeah. That's not going to work. That's not going to work. Yeah. 5,000 customers, too many. yeah. So that's where, okay. So then let's say we go full free for a second. Let's go down that path, right? Full free, all the features are free. Okay. You pay for members. Let's go with my preferred membership model, which is two cents a member, three cents a member, one cent a member, whatever. Something like that, right? Okay. I think it has a couple advantages I enjoy. How **Aaron Francis:** many members do you we have for this show in this fake universe? **Ian Landsman:** I mean, we already have like we already have like 300 signed up just randomly to follow the outro **Aaron Francis:** stuff. So like, **Ian Landsman:** I think we could get a thousand members. Yeah. **Aaron Francis:** And free for the members, right? Yes. Okay. So yeah, we could get a thousand members at two cents. That's 20 bucks a month that we would be paying. Okay. Carry on. **Ian Landsman:** And that, so that has to be worth 20 bucks to us, right? How is it not worth 20 bucks? Like if we ever ask them to do anything, it's worth 20 bucks a month. I think that's correct. Yeah. Right? So, and I think it like, show, like there's a direct correlation to that value, right? --- ### [Let Creators Export Member Data](https://share.snipd.com/snip/2d9fb1fc-6ff1-4213-9202-7c3be43ae18d) 🎧 33:07 - 35:52 (02:45) <iframe src="https://share.snipd.com/embed/obsidian-player/snip/896fb3a2-a4b0-4930-824f-5544c0f4052c" width="100%" height="100" style="border: none; border-radius: 12px;" sandbox="allow-scripts allow-same-origin allow-forms allow-popups allow-clipboard-write" ></iframe> - Allow full exports of member lists so creators can leave without losing audience data. - Avoid adversarial lock-in; make switching friction low while building features that keep users engaged. #### 💬 Quote > No, we're not trying to do the full sleazy lock-in now. > — Ian Landsman Ian Landsman on avoiding adversarial product lock-in for Outro.fm #### 📚 Transcript **Aaron Francis:** one question for you does this set you up for adversarial lock-in? Are you going to allow people to export their member lists? Oh, yeah, yeah, yeah. No, I would **Ian Landsman:** let people export. Yeah, yeah, yeah. No, we're not trying to do the full sleazy lock-in now. **Aaron Francis:** So the bar is just, hey, if you quit using this, you're going to have to export it. Go to some other tool. Import it. Pay for that tool. So it's more of like a friction **Ian Landsman:** thing. Yeah, it's a thing. It's a like, I think there will be features that where it's like, if you just use ConvertKit or Bento or whatever, it's not going to be as integrated with the idea of what your podcast is doing. So I think there can be some features there for the podcasters that are a bonus, even things like obviously very simple things like your new episode comes out, everybody gets emailed, right? Like just automated, it's just like whatever, very easy to set up all that stuff. You don't have to go set that up in the other platform, all those things. But I think there's even more we can do with that. So yeah, my idea is to have some more specific features. also think there's all the whole membership management of like, maybe members get a private podcast that doesn't have ads, right? There's other, there's a PDF you give people, like all this stuff, which again, you can always set up with one of a million other tools, but it's usually a little bit more complicated process because it's not a tool that's like inherently designed with this concept, right? Like, so you can do it, but you have to jump through whatever hoops to make it happen because it's a generic mailing list manager and not designed with the idea of actual members who you are giving special perks to. And the idea of perks is an actual concept that the platform understands. Right. So I think there's some of those niceties that would be in there. I already have a little bit of that in there. so yes i think that's kind of the aspect of that there is that and i think i do think that that's a risk on the other side of it like you do get members who are you know podcasts that build up a big list and they're like hey like bento convertkit mailchimp whoever um now i need the other stuff those tools provide that you don't. Because I want A-B testing and I want whatever. Because what really doing is selling courses. And this is not the main platform for that or whatever. So I think you'll lose people on that side potentially too. But I feel like there's possible answers to that. And it's like a thing to not worry about until like, oh, wow, we lose a lot of people. When they hit 3,000 members, they all leave, right? Well, then like we could figure out what we do about that and why are they leaving? And maybe we could add those features or we have to change something else about the system to make it more valuable in some other way. --- ### [Podcast Download Metrics Are Noisy](https://share.snipd.com/snip/da457b54-ca21-4437-b349-ab80bb0ba8b7) 🎧 39:27 - 40:45 (01:17) <iframe src="https://share.snipd.com/embed/obsidian-player/snip/39aef377-c708-4dc7-87e7-08d8c374d847" width="100%" height="100" style="border: none; border-radius: 12px;" sandbox="allow-scripts allow-same-origin allow-forms allow-popups allow-clipboard-write" ></iframe> - Podcast audience metrics vary wildly by source, so single-platform medians can mislead. - Validate listenership across platforms before making product or pricing decisions. #### 💬 Quote > Median is 27 downloads. > — Ian Landsman Ian Landsman quoting a reported median downloads figure and noting discrepancies #### 📚 Transcript **Ian Landsman:** Well, there's different numbers here. Buzzsprout's data for downloads within the first seven days. Well, it's a little different, but I guess that's kind of similar. No, that's close. Median is 27 downloads. **Aaron Francis:** Oh, no. **Ian Landsman:** This is according to ChatGPT5, but top 5% is 1,000 downloads. Top 1% is 4,000 downloads. Hey we're almost one percenters yeah you know it's a little tricky now with the youtube they're probably not counting that because it's just a podcast platform buzzsprout right so that's kind of interesting too but um all right dscript has now this is a huge discrepancy the median show according to dscript is 421 downloads **Aaron Francis:** per episode massive order of magnitude discrepancy. **Ian Landsman:** It's discrepancy there. Hold on. Okay. Yeah. Yeah. So, don't know. So, that is completely useless. I **Aaron Francis:** don't know which one to trust. That's totally useless. Jackson, if you're listening, please leave a comment on the YouTube episode and tell us what the actual number is. Yeah. I would like to, I would like to, he's got, he's got his fingers on all the data. He knows, he knows all this stuff. 27 is not great. It's not great. 400 is incredible. --- ### [GPT‑5 Is Incremental, Not AGI](https://share.snipd.com/snip/6a0e360b-42a3-435c-8516-aa47cd0dca5c) 🎧 47:29 - 53:24 (05:54) <iframe src="https://share.snipd.com/embed/obsidian-player/snip/15d21af5-dd23-44e5-b64c-ce6b6aafe355" width="100%" height="100" style="border: none; border-radius: 12px;" sandbox="allow-scripts allow-same-origin allow-forms allow-popups allow-clipboard-write" ></iframe> - Aaron Francis felt GPT‑5 was incremental and not AGI, easing existential fears about immediate job replacement. - Incremental releases give people and tools time to adapt rather than cause instant disruption. #### 💬 Quote > They release it and it's like you can change the color of your chats and also it's marginally worse. > — Aaron Francis Aaron Francis reacting to GPT‑5's initial release and capabilities #### 📚 Transcript **Aaron Francis:** Let's do it. GPT-5. Yes. Big, big sigh of relief for me. I'll be honest. All right. You're happy. So I'm happy because it kind of sucks. It's **Ian Landsman:** not whatever AGI or whatever. **Aaron Francis:** No, it's not super intelligence. It's not AGI. TBH. It's not better than GPT4 like it's fine it's totally fine and so some of my existential angst is like you know am i going to be able to feed my family um and then you know sam altman's tweeting like pictures of the death star which i don't think that's a good idea just like from a pr perspective but basically being like what if you had the smartest person in all of history in your pocket at all times for free it's like oh shoot damn that was kind of compelling um and then they release it and it's like you can change the color of your chats and also it's marginally worse and can't do graphs very well i'm like hey this is great news like yeah we're still in it. We still have a chance. So I don't know. You weren't supposed to be worried. We're not doomers. I thought we were cool. We're not doomers. **Ian Landsman:** I'm just, you know. a secret doomer. **Aaron Francis:** You're a secret doomer. not a secret doomer. I'm not a secret doomer. I am a, how should I brand? I'm a reasonablist. I'm a rationalist. another ist. Positive words. Good branding. Okay. No, I like the AI. I use the AI. Yeah. I don't want knowledge work to become obsolete because have you seen my delicate hands? I am a knowledge worker. I sit in a cold, air-conditioned room all day. So, yeah, I'm not a doomer but i did worry i did have angst that like if this is what they say it is education is dead right programming may be dead like what are we going to and then they released it and it sucks and that's kind of like the general feedback i've seen is people being like this is not better what are we doing like it's slower it's dumber give me back the model picker so i'm sure that they will you know improve it over time but it just felt like we might be witnessing a little bit of the plateau of the scale ups of like oh we can just throw more hardware at it and it gets infinitely smarter i don't know if that's true anymore and that like feels kind of like a relief if i'm being honest yeah **Ian Landsman:** oh that's interesting so i haven't been following much of the takes because i was on ostensibly uh vacation but um yeah i would say i i thought it is better i thought like i'm using five more than i use like i never used 4.0 like i always used oh three and the more thinking based ones because i found like they were just so much better than 4.0. And now I feel like 5 is pretty good for things I was using that for. And it's worth the speed. Now I was like, well, the trade-off is now I'll just take the better speed and it's close enough to like what O3 did or whatever. And I think the model picker is better. Like I like it's like, okay, three logical choices instead of here's 42 models. And I have to figure out what the hell is the difference between O3 and 4-O Mini and like all these things that are like, their description is like reasoning, more reasoning. And you're like, well, I don't understand what that is. Yeah, stupid. So I thought that's, I like that better. But yeah, I also think totally no way in the world, like this is just normal software. We released version five and it's better than version four. Right. Because I was thinking, people are saying it's worse. Maybe they're right, whatever, I don't know. For me, I'm like, ah, it seems a little bit better, but it seems like your normal software, incremental've **Aaron Francis:** reached iphone a little better seven or eight phase where it's like it's the thinnest phone we've ever made you're like cool all right what else does it do it's the thinnest phone we've ever made oh okay well i guess that's fine yeah so i feel like yeah cool like it's got **Ian Landsman:** a little better i like the picker better I had less things for me to think about. And great, fine. Not life-changing. Not life-changing. I thought it'd say the cool thing, it has the agent mode. I don't know if that was there before or not. I feel like it wasn't in the UI, which I think makes it a little bit better. I only use it a little bit, but I had to do like a multi sort of build multiple files type of thing. And it did a better job than I feel like it normally does in terms of it actually built all the files like in the past like it would like not build all the files right so it actually built all the files now what it put into those files was not it wasn't pure garbage but it wasn't you know it wasn't like oh I could just take this I had it like build a website basically. Yeah. And I definitely could not take this website and put it on the internet and be like, this is my new website. Like it was nowhere near that. It did have a couple of interesting ideas. It did do the task for the most part of building out the correct number of files. So I was all good. But yeah, I feel like I'm not, I totally don't think they're going to come out one day and be like, look at it. It's basically a human. Like I think it's, I just don't think there's the training data to do it. I don't think, you know, I think it's going to be more stuff like they really figure out how to agentify it in their fancy language. Right. And so if it gets faster and the agents get better, then you could start to simulate more human-like interactions because now it can spin off a bunch of different tasks at once and come back to you the way a human might more, you know, with like having thought about different things and tried a couple different things and whatever. I think that will get better. But yeah, **Aaron Francis:** I don't know. I'm still not worried. I'm not worried. Given the current architecture, I'm not worried. LLMs, I think, are not a path to AGI or super intelligence. --- ### [Agents Need Specialization And Time](https://share.snipd.com/snip/3323ecbf-66ca-4f85-b16b-d7895e59d9ba) 🎧 50:53 - 55:41 (04:48) <iframe src="https://share.snipd.com/embed/obsidian-player/snip/2f7ba1f2-21b7-423a-bf50-f8bb9990a0b6" width="100%" height="100" style="border: none; border-radius: 12px;" sandbox="allow-scripts allow-same-origin allow-forms allow-popups allow-clipboard-write" ></iframe> - Ian Landsman argued reliable AI assistants need specialized models, agents, and extensive engineering to match expert humans. - Broad, human-level AGI is unlikely without task-specific systems and long, careful rollouts. #### 💬 Quote > It really has to be specialized for the task more so to actually do it at anything close to the real human who knows what they're doing level. > — Ian Landsman Ian Landsman on why AI assistants require specialization to reach expert performance #### 📚 Transcript **Ian Landsman:** But yeah, I also think totally no way in the world, like this is just normal software. We released version five and it's better than version four. Right. Because I was thinking, people are saying it's worse. Maybe they're right, whatever, I don't know. For me, I'm like, ah, it seems a little bit better, but it seems like your normal software, incremental've **Aaron Francis:** reached iphone a little better seven or eight phase where it's like it's the thinnest phone we've ever made you're like cool all right what else does it do it's the thinnest phone we've ever made oh okay well i guess that's fine yeah so i feel like yeah cool like it's got **Ian Landsman:** a little better i like the picker better I had less things for me to think about. And great, fine. Not life-changing. Not life-changing. I thought it'd say the cool thing, it has the agent mode. I don't know if that was there before or not. I feel like it wasn't in the UI, which I think makes it a little bit better. I only use it a little bit, but I had to do like a multi sort of build multiple files type of thing. And it did a better job than I feel like it normally does in terms of it actually built all the files like in the past like it would like not build all the files right so it actually built all the files now what it put into those files was not it wasn't pure garbage but it wasn't you know it wasn't like oh I could just take this I had it like build a website basically. Yeah. And I definitely could not take this website and put it on the internet and be like, this is my new website. Like it was nowhere near that. It did have a couple of interesting ideas. It did do the task for the most part of building out the correct number of files. So I was all good. But yeah, I feel like I'm not, I totally don't think they're going to come out one day and be like, look at it. It's basically a human. Like I think it's, I just don't think there's the training data to do it. I don't think, you know, I think it's going to be more stuff like they really figure out how to agentify it in their fancy language. Right. And so if it gets faster and the agents get better, then you could start to simulate more human-like interactions because now it can spin off a bunch of different tasks at once and come back to you the way a human might more, you know, with like having thought about different things and tried a couple different things and whatever. I think that will get better. But yeah, **Aaron Francis:** I don't know. I'm still not worried. I'm not worried. Given the current architecture, I'm not worried. LLMs, I think, are not a path to AGI or super intelligence. There may be other architectures that they're working on that are different than just like, let's turn all these words into numbers and then go make stuff up based on the numbers yeah i don't think we're getting there that that way now are they working on like you know a neural net that's more like a brain with neurons and like the ability to learn things i don't know that may be a pretty good idea um but this you should go yeah i'm not worried about it. I'm not worried about it anymore. Oh, good. We've plateaued. Yeah, I **Ian Landsman:** don't know. You know, I think, yeah, I'm not too worried about it. I listen, I think there's going to be disruption. I think it does a lot of stuff good enough, right? And it's helpful and all that stuff. So I still think it's going be quite disruptive. Even if **Aaron Francis:** it doesn't improve any further, it's going to very disruptive as people adopt it i agree **Ian Landsman:** right i think it's gonna i think it's gonna take the standard technology like whatever like i've said before like we have people still coming to us the reason they're using help spot is because they were just using a shared mailbox shared mailbox has been around 50 years help that software has been around 40 years they're still coming from the old 50 year old thing to the 40-year thing, right? People **Aaron Francis:** are still doing lift and shift for their cloud workloads. It's like, oh, I'm a cloud migration expert. People still need that? What have we been doing the past 20 years? We customers who **Ian Landsman:** have been using HelpSpot for 15 years, and they're in the on-premise version. And they're just now being like, we're still doing this migration to cloud. You know, all the time we have migrations to our cloud platform from people who are not in the cloud platform. It's like, there's all this stuff. It takes so long. The arc of like enterprises shifting is decades and decades, right? So like some things will happen quicker, but a lot of it is going to take a long time. And a lot of it has to be specialized. Like I think if this is, again, if they come out with the, like, I'm just a human but super smart, that changes everything. But if they don't, then it's like, well, okay, this is a really smart thing, but you really have to guide it. And it needs agents. And it really has to be specialized for the task more so to actually do it at anything close to the real human who knows what they're doing level. So now you're on a bunch of specialized software and specialized, maybe even models and whatever else to do so that now it can be something like a physician's assistant, right? --- ### [Exclude Competitor Sponsors For Trust](https://share.snipd.com/snip/dbb8c6ad-3746-499a-aca1-82426d0e852f) 🎧 59:10 - 01:03:57 (04:46) <iframe src="https://share.snipd.com/embed/obsidian-player/snip/ad246f45-8314-4a01-9ae6-44907026b2e4" width="100%" height="100" style="border: none; border-radius: 12px;" sandbox="allow-scripts allow-same-origin allow-forms allow-popups allow-clipboard-write" ></iframe> - Avoid taking sponsorships from direct database competitors to keep guests comfortable and preserve editorial neutrality. - That policy makes it easier to book high-profile database guests without conflict concerns. #### 💬 Quote > I don't think I want to allow database companies to sponsor the show. > — Aaron Francis Aaron Francis explaining Database School's sponsorship stance #### 📚 Transcript **Aaron Francis:** All right. I think I want to get sponsors for Database School. So here's what I'm thinking. The podcast Database School. It's a good show. We like the show. You know, I probably get three to four hundred audio subscribers or audio downloads through Transistor. And then the average video is kind of all over the place, but it can do, you know, three to 10,000 with DHH being an outlier at like 140,000, you know? So here's, here's my thought is I would to do, I'd like to do sponsorships very similar to how we do them, although probably fewer per episode and more expensive. But here's my thought that you may hate. I don't think I want to allow database companies to sponsor the show. **Ian Landsman:** Yeah, **Aaron Francis:** it cuts both ways, right? Cuts both ways. I want to, and I'll give you my reasoning. So far, you don't hate it. So that's already like, you know, it wasn't a visceral reaction. I don't know I hate it or not, but I get it. I get why you wouldn't want to. That's good. We're on the road to understanding. So my thought, which is probably very clear, is I don't want to influence the content, right? And I also don't want the CEO of Prisma to be on and then to read a Superbase ad or vice versa, right? That just feels like, what the hell am I doing here as the CEO of a competitor giving you content so that you can then cut in and advertise my competitor? It's just the whole thing seems misaligned. And so my thought is the audience is highly technical, probably only senior engineers that would be willing to listen to an hour of like database stuff that a lot of companies would probably want access to outside of even database companies. So that's my pitch. What's your take? **Ian Landsman:** I do like it from the like, yeah, keeping it clean for you. Like no weird stuff like that, of course. Keeping you the impartiality, I guess you'd say, right, of the content. Like you're not favoring because Oracle paid you a bunch of money and now everything's MySQL or whatever. **Aaron Francis:** I don't If Oracle wants to pay me, now that's a different guy. They have more money than God. We'll see. Okay. **Ian Landsman:** So here, I do think, though, this is getting to my downside, which is that I feel like you are defining, first of all, databases quite broadly. I think right is your goal here right like redis whatever all kinds of stuff right vector databases so and who's left yeah who's left like it's sort of who's left and the people who are left are now not so well aligned right it's like you have this great alignment like these are people super interested in databases i get why i would advertise on this but now i can't advertise my database stuff on this. So now I have to advertise like, okay, so I'm Laravel. I'm going to advertise Nightwatch or Cloud. But are these even software developers? Like maybe they are database administrators, right? So now I'm definitely less aligned. Like there will be some programmers for sure, but it's not going to be as aligned as prisma advertising or whatever um so that's the part where i wonder if i just feel like you're throwing away a lot of revenue oh **Aaron Francis:** yeah for sure practical angle to that um and i don't know the connections like all these people that have come on the show i could email and be like hey do you want a sponsor yeah now i'm now i'm stuck going to like digital ocean or headsner people I don't know, like compute. I mean, any sponsor that we've had on here would work. You know, Nightwatch, Bento, Century, Flare, Honey Badger, anything would **Ian Landsman:** work. I don't know. I just say it's yes and no, right? A little bit. Because they're still way less aligned. Because how many of them are even programmed? think the programmer oriented ones might not be so interested in database school potentially now that's you so like the people literally on our show who sponsor our show like they know you very well and they know that probably most of your audience is actually programmers who are watching that or programmer adjacent but once you get out of that little bubble to people who don't necessarily know Aaron Francis very intimately and know about the Laravel connection and all the stuff. Now it's just like, well, the database school, it's going to be database administrators and people who are super in databases and how a percentage of them are actually programmers might only be 20% or whatever. I think for you, it's much higher than that, but I don't know. So, so I guess I don't oppose it, but I do think it's worth considering. Could you do something different to give yourself the opportunity for those people to still advertise? Like, could you just structure it to get some of the benefits of if you had none, but without doing, going that far with it? Like when you have somebody from a database platform, you do not run other database advertisements on their interviews. --- ### [Segregate Ads To Avoid Guest Conflicts](https://share.snipd.com/snip/65f4412a-5c6a-488d-8445-96500fc02c8f) 🎧 01:02:44 - 01:05:10 (02:25) <iframe src="https://share.snipd.com/embed/obsidian-player/snip/9f504cfe-61e0-49a5-beb5-facde19268f9" width="100%" height="100" style="border: none; border-radius: 12px;" sandbox="allow-scripts allow-same-origin allow-forms allow-popups allow-clipboard-write" ></iframe> - If you accept industry ads, structure them so competitors don't run against a guest's episode. - Use intro-only or segregated ad placements to avoid mid-interview conflicts and preserve guest trust. #### 💬 Quote > When you have somebody from a database platform, you do not run other database advertisements on their interviews. > — Ian Landsman Ian Landsman advising how to handle sponsor alignment for guest episodes #### 📚 Transcript **Ian Landsman:** work. I don't know. I just say it's yes and no, right? A little bit. Because they're still way less aligned. Because how many of them are even programmed? think the programmer oriented ones might not be so interested in database school potentially now that's you so like the people literally on our show who sponsor our show like they know you very well and they know that probably most of your audience is actually programmers who are watching that or programmer adjacent but once you get out of that little bubble to people who don't necessarily know Aaron Francis very intimately and know about the Laravel connection and all the stuff. Now it's just like, well, the database school, it's going to be database administrators and people who are super in databases and how a percentage of them are actually programmers might only be 20% or whatever. I think for you, it's much higher than that, but I don't know. So, so I guess I don't oppose it, but I do think it's worth considering. Could you do something different to give yourself the opportunity for those people to still advertise? Like, could you just structure it to get some of the benefits of if you had none, but without doing, going that far with it? Like when you have somebody from a database platform, you do not run other database advertisements on their interviews. So you could... **Aaron Francis:** That's **Ian Landsman:** like most of them. Could you it where you do a thing at the beginning and it's only in the beginning, but they're never interrupted with an advertisement? Yeah, I can do that. You know what I mean? Where it's like segregated so it's not so much like here's the planet scale guy. Now we've interrupted him with AWS. I feel **Aaron Francis:** weird being like, hey, welcome to, you know, as an ad, like an intro before the guest is even there. Welcome to Database School. This episode is sponsored by Supabase. All right. Now, the CEO of Ne let's go it's like yeah feels that feels a little weird to me but it's like how much leverage do you have though **Ian Landsman:** like i mean at the end of the day how many database specific podcasts are there right and like very few i'm guessing very few i don't know if that's true but i'm guessing very few so if they want to reach the people who are super into databases, this is the cost to them. Is that like, first of all, they should be advertising that they are there. And then second of all, you know, this is just if you want to reach these people, your competitors are also going to be there, which is always true. Right. So but I do see it does get weird. --- ### [Use Voice Memos For Ubiquitous Capture](https://share.snipd.com/snip/b130e996-3e4e-4a71-8932-d745d2bc0a10) 🎧 01:13:17 - 01:14:35 (01:17) <iframe src="https://share.snipd.com/embed/obsidian-player/snip/5ec21377-ebe7-4a24-9dc3-62f7d21105d6" width="100%" height="100" style="border: none; border-radius: 12px;" sandbox="allow-scripts allow-same-origin allow-forms allow-popups allow-clipboard-write" ></iframe> - Capture ideas everywhere by using voice memos because iOS records and syncs instantly. - Defer heavy processing to desktop, where transcriptions and integration into workflows are easier. #### 💬 Quote > Voice memos are very helpful for me because they are so well integrated into ios. > — Aaron Francis Aaron Francis on using iOS voice memos for quick idea capture #### 📚 Transcript **Aaron Francis:** and about experience, the mobile experience, is it enough to capture and send while you're out and about and then process when you're back on the desktop? I'm trying to figure out what is, like how robust does mobile need to be? **Ian Landsman:** Yeah, it's mostly, that could be, mostly like hey i have this thought while i'm doing whatever i just want to like be able to log it and have the thought saved somewhere and then sometimes i'll do more with it sometimes i won't kind of depends what exactly it is uh but yeah potentially okay **Aaron Francis:** so we need a we need like ubiquitous capture that's what we're going for i need to be able to capture anywhere, not necessarily process everything everywhere. Yes. Yeah, I don't know. I'm notoriously anti-second brain, so I'm not very deep into this whole PKM, this personal knowledge management thing. Me **Ian Landsman:** either, really. I'm not super deep on it. Yeah, I **Aaron Francis:** do like the idea of capturing while I'm out and about. started using voice memos for that um which i may have a youtube video on soon because there's some interesting there's some interesting things there um okay but voice memos are very helpful for me because they are so well integrated into ios that like you can just push that action button record a thing and then let go and it's just it's done it's there where do they go it's **Ian Landsman:** just in the voice memo app that's --- ### [LifeOS Automates Content Workflows](https://share.snipd.com/snip/fead44a3-a636-4ba5-83de-88575614819c) 🎧 01:24:10 - 01:26:32 (02:22) <iframe src="https://share.snipd.com/embed/obsidian-player/snip/79f3f961-4c2c-4def-8695-38c465fc1cda" width="100%" height="100" style="border: none; border-radius: 12px;" sandbox="allow-scripts allow-same-origin allow-forms allow-popups allow-clipboard-write" ></iframe> - Aaron Francis built a LifeOS pipeline that ingests tweets, bookmarks, videos, transcriptions, and more into one workspace. - He uses automation to surface content, create thumbnails, and prepare drafts so he can focus on higher-level work. #### 💬 Quote > Under the hood is everything else and i've i've a really good pipeline. > — Aaron Francis Aaron Francis describing the automated pipeline inside his LifeOS #### 📚 Transcript **Aaron Francis:** a really good, so like I have a really good pipeline, I have a really good system set up through which I can have the computer reach out to me and ask me human input questions, and I can respond and it goes back to work. And that pipeline has proven very, very profitable. So like it's pulling in all my tweets including like my public tweets and my private bookmarks which is very helpful i keep a lot of information in there it's pulling down all youtube videos from this channel mostly technical channel my personal channel and any like um like uh guest videos i do so i have a few on the laravel channel and then i have podcasts, guest podcasts scattered everywhere. It pulls them all down, transcribes them, makes moving thumbnails for them, like asks me who the speakers are, pulls down comments now so that I can feature certain comments on AaronPrancis.com as like, you know, testimonials, social proof. Right. And then the voice memos thing is the thing I've been working on recently where i can take a voice memo and then icloud will sync it and then sync it down to the mac studio and obviously life os is running on the mac studio as well and through a little trickery i can get at those files transcribe them and then like integrate them into the rest of the workflow so it's working I'm very happy with it. The entire goal is to automate more and more things. I think for me, it primarily automating content creation or like parts of the content creation workflow. So at some point, I will try to like, when I'm done editing a video just export it into the mac studios you know storage directory of life os and then life os will wake up and say like hey here's a new video let me just upload that to youtube as a draft because then then like i have one fewer thing to remember so i'm very happy with it and it's making me think like oh this is really interesting content maybe i should do something around this whole thing, automating your life with Laravel and AI. Like, well, it's pretty interesting. So I don't know. Oh, don't let people hear that. I know. So yeah, I'm happy with it. I'm excited about it. And it's going well. --- Created with [Snipd](https://www.snipd.com) | Highlight & Take Notes from Podcasts