# 116: Inside The Token Factory  ## Episode metadata - Episode title: 116: Inside The Token Factory - Show: Mostly Technical - Owner / Host: Ian Landsman and Aaron Francis - Episode publish date: 2026-01-20 - Episode AI description: Join Ian and Aaron as they recap their exciting trip to San Francisco, where they explored the buzz around AI and showcased the Laravel AI SDK. Discover the inside scoop from their visit to Anthropic, complete with meet-ups and scooter adventures. They also delve into the launch of Faster.dev, discussing innovative product strategies and the creative logo design process. Along the way, enjoy their light-hearted musings on podcast listeners and a cost-saving email solution with Bento. It's a blend of tech insights and amusing anecdotes! - Duration: 01:06:19 - Episode URL: [Open in Snipd](https://share.snipd.com/episode/dffa86f2-6c4f-4c6a-85e3-f1d4e18c7d7c) - Show URL: [Open in Snipd](https://share.snipd.com/show/7ad45fb4-8aa6-49c2-83cc-bc2327d14b03) - Export date: 2026-02-17T23:35:46 ## Snips ### [Inside Anthropic Offices](https://share.snipd.com/snip/2328f959-b2c5-4c3b-9f6d-50dee693fb64) 🎧 04:20 - 06:34 (02:14) <iframe src="https://share.snipd.com/embed/obsidian-player/snip/bf873e4f-899a-4cb0-98ea-6232ad725a1d" 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 visited Anthropic, toured accessible areas, and met a Claude Code engineer. - He described strong security, champagne happy hours, and high excitement about AI work. #### 💬 Quote > I went to Anthropic and met up with an internet friend there. I saw where they're cranking out the tokens. > — Ian Landsman Ian Landsman describing his Anthropic visit #### 📚 Transcript **Ian Landsman:** run it down. I went to Anthropic. I went inside the token factory. I saw where they're cranking out the tokens. Oh, man. I saw where the Ralph Wiggum loops are happening. Yeah, so I went to Anthropic and met up with an internet friend there. I'm telling you, man, the internet is undefeated, just absolutely undefeated. So went and met up with an internet friend there, got to go in their offices, only to like the, you know, the pedestrian areas, of course. I didn't see the machine gods. Like where **Aaron Francis:** they have all the humans who could type really fast and just like reply back to you with **Ian Landsman:** answers and write your code really quick. I didn't see that. Okay. I went to the floor where they have like their cafeteria, which is underselling it quite a bit. They were setting up for happy hour and there was like champagne bottles. And I was like, yeah, this is, **Aaron Francis:** oh, of course it's Wednesday. Why wouldn't you? These are the two ends I may interrupt you for a second I only want to be on these two ends I either want to be on what I have which is like we are super laid back everybody works wherever we we just work normal hours I want to be that or I want to be we all just live at the office we bring in champagne yep we have a gym we serve four meals a day. There's a chef 24 hours. And that's just what we're all doing here, people. Nobody goes home ever. That's what we're doing. Those are the only two ends. I **Ian Landsman:** am a young man. If I were a younger man, **Aaron Francis:** I would be out there. I get it. I get it. I don't know how you're not out there. You were built for that out there. I was built We're going to be out there more. That's my prediction. You're going to be out there even more than me, I think. I think so. You're going to be out there. **Ian Landsman:** Yeah, I think so. So got to hang out with one of the guys that builds Claude Code. And it was just a sweet hang. No agenda, no deal flow. But it was a lot of fun. Security there. Tip top shape. I'll tell you that. I will tell you that much. You look up the Anthropic office on Google, not the office I went to. --- ### [AI Raises The New Developer Baseline](https://share.snipd.com/snip/35943f35-db47-4e20-b3d2-e9f8486e09b4) 🎧 10:48 - 13:10 (02:21) <iframe src="https://share.snipd.com/embed/obsidian-player/snip/14b43c39-3e8f-4713-9d73-b351dc64c181" width="100%" height="100" style="border: none; border-radius: 12px;" sandbox="allow-scripts allow-same-origin allow-forms allow-popups allow-clipboard-write" ></iframe> - AI will change developer workflows and raise baseline productivity even if hype is overstated. - Many companies will tinker first, then build practical internal tools that matter. #### 💬 Quote > As much as I'm prepared, I am not prepared for what's coming. > — Ian Landsman Ian Landsman on the scale of AI advances seen at labs #### 📚 Transcript **Ian Landsman:** And just to hear him talk about the stuff they're doing internally, it's like, oh my gosh. As much as I'm prepared, I am not prepared for what's coming. So hopefully I can escape the permanent underclass. But there are like, even if individual companies go bust, which I think many will, I don't think this is, I don't think it's going to be over. I think this is like leveled up the new baseline of how people will work in the future. I still don't buy all the hype that we see on Twitter. But I think if you cut that in about half, that's pretty close to reality, in my opinion. Yeah, **Aaron Francis:** I still think there's a lot of like, there's just where things hit the real world. I think people like maybe you can increase productivity in certain zones. I mean, you will, but like you still have to, people only have so much attention to give anything. And so there's still a lot of places where the real world's going to intercede all your productivity and be like, yes, but we've made 8,000 products. How do we get enough people to know about these things and want to buy them? There's still going to be only so much time for people to even learn all the things that are out there, blah, blah, blah. you're going to have to have You're going to have to have advertising. You're getting, there's only so much room for those things to exist in. So I don't know. There's still like an end of the day where like, it's not just going to be just because you're more productive doesn't mean you're going to be successful. It doesn't mean your product's going to work. There's just a lot of places where it all breaks down where they're like, we're going to be a times more productive and blah, blah, blah. That means everything will be run by one company or whatever. I don't know. Because there's other types of advantages you can have besides just software development speed. And those are not so easily overcome with more software developer speed. I **Ian Landsman:** think the advantages are unevenly applied. I think the technological advantages of like being able to build stuff quickly are extremely real. And then how do you get people to care is still very much the same. Still the exact same amount of difficulty it always was. And I think right now, most people, at least judging on the timeline, most people are like, I don't know, tinkering or navel gazing. It's like I'm using my setup to improve my --- ### [Start With Internal AI Tools](https://share.snipd.com/snip/4c22f88c-5a03-4896-b0b6-2aed8685a181) 🎧 19:55 - 21:08 (01:13) <iframe src="https://share.snipd.com/embed/obsidian-player/snip/f3bf8e3e-338b-4a12-afee-2f4f32ca1bf0" width="100%" height="100" style="border: none; border-radius: 12px;" sandbox="allow-scripts allow-same-origin allow-forms allow-popups allow-clipboard-write" ></iframe> - Build internal AI agents for sales and first-line support to avoid heavy marketing needs. - Prioritize internal tools because they require less UI polish and easier adoption. #### 💬 Quote > Internal tools is such a good one... you don't have to market it. > — Aaron Francis Aaron Francis on focusing AI efforts on internal productivity tools #### 📚 Transcript **Ian Landsman:** So that's the thing I got most excited about. **Aaron Francis:** Yeah, I think that stuff. This is like a cool thing, I think, for faster.dev to focus on to me, too, is just like being out there, finding what cool things people are doing and surfacing them, like cross-pollinating the ideas, because it's like, yeah, you can build stuff faster now, but what should you be building? It's, again, another issue with this whole thing. Like, we're a thousand times faster, but like we don't know where we're going or what we should do with that. Like, so coming up with good ideas of you should emphasize. I think internal tools is such a good one, too, because there's so much. of that other overhead problems. It's like you don't have to market it. It's an internal tool or at least not in quite the same way. The UI, yeah, the AI doesn't do as good a job with the UI, but who cares because it's an internal tool and you're just going to run the components straight the way they come out of the box for whatever framework you use. And you don't have to worry about that too much. So that's all great. So it's like, okay, how can we just increase our internal, you know, productivity and things like that? So really like that as like an angle for fast drive dev is just like surfacing those things and all that. But yeah, I hope he does that. I think that'd be cool to see what they're doing. --- ### [Making The Faster.dev Video](https://share.snipd.com/snip/21622895-d30e-4d4d-a139-d851891d38b5) 🎧 31:05 - 32:19 (01:13) <iframe src="https://share.snipd.com/embed/obsidian-player/snip/2acdd4dc-5222-49c4-8f34-d13d27ddf897" 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 described creating the faster.dev launch video using Claude, Moises.ai, Adobe Audition, and a producer. - He iterated audio and visuals with human help to finalize the track and edit. #### 💬 Quote > I started by, of course, talking to Claude about songs. And then I went over to moises.ai and it spits out 15 different tracks. > — Aaron Francis Aaron Francis recounting the video production workflow #### 📚 Transcript **Ian Landsman:** You know, you win if you do, you win if you don't. That's the way we like it around here. Yeah. All right. Okay, what else we Should we do it? Yeah, let's do it. Bastard.dev has pre-launched, and we pre-launched with a great video come on it's a great video very **Aaron Francis:** good video we love the videos so **Ian Landsman:** a couple weeks ago i was getting hounded by everyone that i know and love dearly bro you just gotta ship bro you gotta ship bro like what are you doing stop doing anything just ship and I'm like guys give me a second and so I decided all right let's put together video let's let's claim some ground here put our stake in the ground get a video out buys me you know maybe I don't know how long February 1st is from now like fewer than two weeks, I think. So I put the video out. So I started by, of course, talking to Claude about songs. I'm like, hey, I need a song that can work for this kind of like hype video. And it gave me, you know, --- ### [Use Pre-Launch Funnels To Learn](https://share.snipd.com/snip/5821959a-e4fe-4003-9201-029a9b9d4b83) 🎧 37:03 - 37:57 (00:53) <iframe src="https://share.snipd.com/embed/obsidian-player/snip/f7dc1421-20de-4c1b-86db-79d9abf9777a" width="100%" height="100" style="border: none; border-radius: 12px;" sandbox="allow-scripts allow-same-origin allow-forms allow-popups allow-clipboard-write" ></iframe> - Pre-launch content and email automations help convert early interest into sales and product direction. - Faster.dev used Bento automations to survey signups and refine content plans. #### 💬 Quote > We've used Bento Automations to send out an email that's like, hey, where are you at? What's your stack? > — Aaron Francis Aaron Francis on collecting buyer signals during pre-launch #### 📚 Transcript **Ian Landsman:** Sales are selling. Sales are selling. I got numbers this morning from Kelsey. We have made 210 sales. **Aaron Francis:** Wow. Yeah. Holy cow. All right. So pretty good. Pretty good. And this is a subscription, which is very different than a course. Yes. It's a very different thing to sell. Yes. Yes. 100%. I think that's fantastic. **Ian Landsman:** I think so. And we've gotten even more email signups from people who weren't ready to buy. We've used Bento Automations to send out an email that's like, hey, where are you at? What are you hoping to learn? What's your stack? What's your current comfort level with AI? And so that'll help inform, one, one, who I'm reaching, but two, like, the kind of content that I should produce. So just, like, very, very pleased with the launch and the reception. --- ### [Ship Fast, Learn Faster](https://share.snipd.com/snip/b2831962-705e-457d-b6d8-4b57b7516234) 🎧 41:16 - 43:50 (02:33) <iframe src="https://share.snipd.com/embed/obsidian-player/snip/bd926770-9ba1-4cef-930e-6e41a2e4e1b3" width="100%" height="100" style="border: none; border-radius: 12px;" sandbox="allow-scripts allow-same-origin allow-forms allow-popups allow-clipboard-write" ></iframe> - Ship early and iterate publicly to create content and feedback loops. - Use projects as a lab to discover useful parts and teach those findings. #### 💬 Quote > The 90% case of any of these products is give me a lab in which I can discover new things and give me content to talk about. > — Ian Landsman Ian Landsman explaining the purpose of shipping early projects #### 📚 Transcript **Ian Landsman:** And it gives me lots of live streaming content. And like if you want to go deep on how was this part of the app created, there's probably going to be a live stream that I'll put on the site so you can go back and see, oh, there's first sale here, and then he did this live stream, and then he hit this revenue, and then it was like gave up on this project, and then great, we're done. So it'll be interesting. It's all in pursuit of creating a spectacle. Do **Aaron Francis:** you worry? Let's push into this a little bit because one of the things i my bigger concern with this is first of all you have situations like it's not going to be less work because you have things like we just had a whatever that open source thing you did that ended up being a lot of work and you were like holy shit there's a lot of work right and it's a free thing that's downloadable. Fusion, right? Yeah, but see, nobody **Ian Landsman:** liked that. So I just gave up on I understand, like **Aaron Francis:** there was pressure there. So that gets to my second point. If nobody likes the things you release, is that going to make you frumpy? And I don't want you to be frumpy. I don't want to be frumpy. I don't like being frumpy. People like faster.dev. The idea of it, I think they're going to like the stuff you ship with it. And so I think making it more about the products to me is potentially negative. Whereas like, I agree you need something to talk about. So I'm on board with that. That is the **Ian Landsman:** 90% case of any of these products is give me a lab in which I can discover new things and give me content to talk about. So I can go off and do stupid stuff on one of these products and then realize this was stupid, but I actually found, you know, one piece that works. Let me pull that back in and teach everyone else the one piece that works. **Aaron Francis:** Yeah. **Ian Landsman:** i can't just educate in a vacuum because there's like then you're just a you're just a think fluencer and i don't want to be that yeah **Aaron Francis:** yeah i think it's just the balance of like how much is this a real product where i'm going to be on top of customer service and right there and i'm getting bugs and fixes and all the real stuff that goes into like really shipping a real piece of software that people really pay for and like all that stuff first like how much is it content and riding that line I mean it's a little bit of a weird thing from a customer acquisition standpoint to be like this is not real so you kind of have to make it real which implies some realness I just think there's a lot of complexity there. There's a lot of complexity. love complexity. Mental complexity. I think there's a lot of practical complexity. --- ### [Teach Principles Over Specific Tools](https://share.snipd.com/snip/3d57663d-d902-4b06-a292-06f9b320f362) 🎧 46:56 - 47:55 (00:59) <iframe src="https://share.snipd.com/embed/obsidian-player/snip/05d4f47d-94a0-4552-896e-0a273898a030" width="100%" height="100" style="border: none; border-radius: 12px;" sandbox="allow-scripts allow-same-origin allow-forms allow-popups allow-clipboard-write" ></iframe> - Teach enduring principles and workflows, not tool-specific code, because tools will change. - Skills and guidelines transfer across providers like Claude, Gemini, and OpenAI. #### 💬 Quote > I'm trying to teach principles and guidelines because the tools themselves will change. > — Ian Landsman Ian Landsman on the pedagogy behind faster.dev #### 📚 Transcript **Ian Landsman:** I'm trying to hit the meaty middle. I'm trying to hit the people who are not, like, ricing their, you know, their tuis or whatever. It's like... They're **Aaron Francis:** not buying a Linux desktop to run open code. No. like, do all the things. **Ian Landsman:** They probably don't have Omarki or whatever DHH's thing is. It's like, we're just developers trying to get some work done. Trying to get by. That's my mental avatar. We'll see if that **Aaron Francis:** holds up to reality. Just a man in the galaxy trying to make his way through the galaxy or whatever that you're telling me yeah **Ian Landsman:** i don't know what line that is but i like that's **Aaron Francis:** from a terrible star wars movie uh probably so all right i don't know so what's next let's see it's january 19th today it is so give me some timelines here what are you looking at? **Ian Landsman:** So I'm not going to release, you know, there are like eight or nine series on the homepage. I'm not going to release all of those the first week of February. --- ### [Designing The Particle Logo](https://share.snipd.com/snip/68bc75d6-e8b6-4159-9c99-40779d2bf2a9) 🎧 50:00 - 52:19 (02:18) <iframe src="https://share.snipd.com/embed/obsidian-player/snip/6f686970-eb1c-4848-8ea2-e2b8b35c7ef9" 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 bought faster.dev, used MidJourney and Claude to design a motion-particle logo, then made a React HUD to tweak it. - He exported paused animation frames for OG images and reused the component across the site. #### 💬 Quote > What if we take that same motif of like motion and instead of doing like speed tearing lines, what if we did like particle disintegration? > — Aaron Francis Aaron Francis describing the creative process for faster.dev's logo #### 📚 Transcript **Ian Landsman:** So I found the domain. The domain is great. Great domain. It's Originally, I had like move faster.dev, which is an order of magnitude worse. I hate that. I don't like that name at all. At all. And I saw faster.dev was available aftermarket. So it cost me a pretty penny. You got do what you got to do. But you got to do what you got to do. So I bought it. And then I started thinking like, I need like a motif. You know, with database school, I've got the chalkboard. It's like I can kind of build everything around that. I thought I need like a motif. Let's go with something faster. And so I went over to Mid Journey and started talking to Mid Journey about like making a logo that looks like it's in motion. I was like, okay, coming up with some cool stuff there. It was doing a lot of like kind of like speed tearing, like motion blur to one side. And I was like, yeah, that looks cool. So I took those images to clawed on the web. It was like, hey, I want you to build me out in HTML, CSS, JavaScript, SVG, whatever you want. Something where I can play around with these settings. And it did. And I was playing around with them and I just wasn't moved by anything. And so I was like, what if we take that same motif of like motion and instead of doing like speed tearing lines, what if we did like particle disintegration? And so like it's driftier on the left and more solid on the right. And that's when it hit. That's when it was like, yeah, this is cool. This is a motif I can get behind. And so I had it build it out as an artifact on the web. And then I downloaded it and plopped it into the repo and told other Claude, its brother, I said, hey, here's, you know, here's a reference. I need you to build me out like a React component that's all nice and tidy where I can like, you know, have props and put in different words and control the drift on various parameters. And it just came together. And so that's what I ended up using for the hero. And then it is a component, so I can use it for other stuff, like the post-purchase thanks page. --- ### [Make Tools To Tune AI Outputs](https://share.snipd.com/snip/672c1fb1-8248-415b-a446-39e27e2adceb) 🎧 54:13 - 55:02 (00:49) <iframe src="https://share.snipd.com/embed/obsidian-player/snip/e7f2e3aa-7440-49d9-935f-7aa2593fff81" width="100%" height="100" style="border: none; border-radius: 12px;" sandbox="allow-scripts allow-same-origin allow-forms allow-popups allow-clipboard-write" ></iframe> - Build small tools (HUDs) around generative outputs to iterate quickly instead of repeated prompts. - Ask the model to produce adjustable interfaces so you can fine-tune values visually. #### 💬 Quote > Just build me a heads up display and I will do it. And then I'll tell you what values to use. > — Aaron Francis Aaron Francis advising to construct interfaces for rapid iteration with AI #### 📚 Transcript **Aaron Francis:** It's like just build me the tool so I can figure it out. But I have all the little dials right there so I can get the spread right and the speed right and the rotation right and all the stuff. And yeah. Yeah. **Ian Landsman:** Or like just constantly banging your head against a chat and being no, make it a little bit, a little bit more drift. No, a little bit less drift. No, move it up a few pixels. It's like, no, dude, just build me a heads up display and I will do it. And then I'll tell you what values to use. That's definitely going to be one of the tips and tricks inside of Faster.dev is like, you can just ask it to build you a HUD that has 15 sliders on it, and you can play around with it for 10 minutes until it's perfect, take a screenshot and give it back, and say, here are your prop values. Use these. It's awesome. --- ### [Appearing In A Stripe Campaign](https://share.snipd.com/snip/f2e0aaeb-a910-4abd-bceb-f6e6a84f1ad0) 🎧 57:22 - 58:47 (01:24) <iframe src="https://share.snipd.com/embed/obsidian-player/snip/9b75fb4f-cbe3-421e-a0c9-302e21cb9a90" 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 filmed short Stripe influencer videos after an agency reached out and the company paid him. - Stripe co-posted the short-form content across social platforms without his direct posting. #### 💬 Quote > Some agency reached out and was like, hey, we work with Stripe. They want you to make an influencer video. > — Ian Landsman Ian Landsman recounting how he shot paid short videos for Stripe #### 📚 Transcript **Ian Landsman:** like two months ago or something, some agency reached out and was like, hey, we work with Stripe. They want you to make an influencer video. And I'm like, who are you? What kind of scam is this? Yeah, exactly. I have to send you how much money? But turns out they were real. And so the deal was like, I think I made, I forget, it's like five videos. Wow. And they're all short form, so they're like, you know, under 30 seconds. They're vertical. And they gave us kind of like some talking points of things they want to hit. And so I just like one day, you know, set up a bunch of cameras, used my phone a few times, and just like filmed these 30 second shorts, basically. And then I edited them. So like they had, you know, windows sliding on and, you know, fancy, like kind of screen studio style recordings, but using a different tool. And they finally like made it back through, you know, the Byzantine, you know, structure inside of Stripe. And it's interesting because part of the deal is they connect, like you saw in the screenshot, it's like on Instagram, it says, you know, Aaron Francis and Stripe. It's kind of like a co-post. So I connected my social accounts to whatever, their tool or whatever. --- Created with [Snipd](https://www.snipd.com) | Highlight & Take Notes from Podcasts