# p99 Podcast with Sam Lambert: Taylor Otwell of Laravel

2026-09-03 · Recording: https://x.com/i/broadcasts/1OGwbnpNEnmKB · archived copy: https://archive.sirupsen.com/recordings/p99-taylor-otwell.mp4

_Lightly edited automatic transcript: speech-to-text, cleaned up by an LLM. It may contain errors; the recording is the source._

## Welcome and Laracon events

<!-- t=28 -->
**Simon:** We're very lucky to have a very special guest today: Taylor Otwell, the creator of Laravel. Good to meet you.

<!-- t=38 -->
**Taylor:** Thanks for having me.

<!-- t=39 -->
**Simon:** How's it going?

<!-- t=41 -->
**Taylor:** Pretty good. Pretty busy with work and life and everything, trying to ship here at Laravel, remotely from Arkansas. It's been a busy summer. A big conference in Boston, and I was in Japan for a Laravel Live event, so it's been a packed summer.

<!-- t=59 -->
**Sam:** You do huge Laravel events, right? Your India event was massive. Do you want to tell us a bit about that?

<!-- t=65 -->
**Taylor:** We've got a few Laracon events around the world. We have the one in the US every year, which is 900 to a thousand people, sometimes a little bigger. We have one in Amsterdam, around a thousand people. We have one in India, about 1,200 people. Then we have an Australian Laracon, which is smaller, around 400. Those are all official Laravel events. Then we have Laravel Lives, which are community-run events. We have one in Copenhagen and one in the UK. Those are usually in the 400 to 500 person range. I just got back from Copenhagen, actually. We had DHH there and did a roundtable.

<!-- t=110 -->
**Sam:** The man. Simon and I have both worked at two of the largest Ruby on Rails websites on planet Earth, GitHub and Shopify. Shopify is one of the largest on Earth, and their absolute load and QPS is unbelievable. Laravel was the Rails moment for the PHP community, right? There was such diversity in the PHP community, with lots of different frameworks. I know you took a lot of inspiration from Rails. We love Rails, we love the Rails way. PlanetScale uses Rails. It's something that's ingrained in both our cultures, and something Simon and I bonded over is that way of systems engineering. How did you get to doing this?

## Laravel's origins and Rails inspiration

<!-- t=163 -->
**Taylor:** It goes way back. I was doing enterprise .NET and COBOL development at a company here in Arkansas, and I didn't want to build my side-project startup ideas in COBOL or .NET. Rails was really hot at the time, around 2008, 2009, basically at its peak. I knew PHP from college, and the Rails ecosystem was very Mac-focused. I just had a Windows laptop, and I could not get Rails to work on it. So I was like, screw it, I'll just do PHP, I already know that. That moment was a big turning point in my career, because I started building this Rails-like framework in PHP. For pretty much every major feature, I would Google how Rails does it. How does Rails do routing, or validation? I would take inspiration from it, sometimes modify it or put a bit of a different spin on it. It inspired so much of what Laravel is.

<!-- t=233 -->
**Sam:** To me, Rails is a bit like an iPhone moment, even for PHP. Every framework that came after it for a while looked like that. It had a models directory and a controllers directory, and Laravel was no different. The structure was very similar. What were some of the early moments in Laravel's history where it really took off, where people noticed it and it clicked for them?

<!-- t=259 -->
**Taylor:** PHP was, and some would argue has always been, in a dark spot. Back then it was really in a dark spot, where Node and Ruby and other popular languages seemed to be where all the cool stuff was happening. I was hanging out in IRC channels back then and people were depressed. Should I change the language I use at my job? Should I change jobs? They were worried about their career. When Laravel came out, it felt modern and fresh. It was also an opportune time in PHP's history, because PHP 5.3 ushered in a lot of new language features that made building something Rails-like even possible. It really wasn't technically possible before. So when Laravel came out, it was a breath of fresh air for a lot of developers. There's so much PHP in the world that people thought, "Oh damn, this is pretty sick, actually. I can keep using the language I already know and have this really modern framework that feels very fresh and Rails-like, and is similar to what other ecosystems are enjoying." I think that really clicked for people.

<!-- t=331 -->
I also put a shitload of time into the documentation from day one. The 1.0 release of Laravel had insane documentation from the very first day, very clear and very detailed. That was huge, I think, in kicking off mass adoption of it in PHP.

## Early adopters and who uses Laravel

<!-- t=351 -->
**Sam:** Who were the first big adopters? The big names that started coming in, where you noticed they were using it and thought, "Oh wow, this is real, this is serious"?

<!-- t=360 -->
**Taylor:** Laravel has never been a Silicon Valley, San Francisco darling child in terms of adoption. We have a bunch of really random companies. Early on, a lot of the people reaching out to me were government agencies that had started using Laravel, sometimes for classified projects, sometimes for non-classified web projects they were doing internally. They'd ask for feedback or ask how something works. Those were some of the really early moments where I thought, "Wow, this is pretty crazy, actually." That was within the first couple of years of it being released. A lot of the typical customers of Laravel are mid-market and large companies that are doing well and making a lot of money, but aren't necessarily household-name brands. If you look at our top 10 customers on Laravel Cloud, it's a wide variety of random, kind of interesting stuff.

<!-- t=424 -->
**Sam:** You do have some pretty amazing ones, though. Kick.com is a Laravel site. We've got some pretty good shared customers at sizable scale.

<!-- t=436 -->
**Taylor:** Laravel also wiggles its way into larger companies through smaller internal tools or through acquisitions. That's happened at Zillow, or even Apple or Pfizer, where Laravel plays some role in the company. It's obviously not the whole company, but large companies have found themselves using Laravel that way as well.

## In defense of PHP

<!-- t=458 -->
**Sam:** You kind of slightly apologized for PHP, or referenced it in a light that some people have as negative. I just want to say that the people who hate on PHP are wildly wrong and should try doing anything better. We have Facebook and the modern web because of PHP. And 70% of the web is still served by PHP and MySQL websites. Now most of it's WordPress or whatever, but PHP and MySQL together built the early web. The impact is wild, and the breadth of what it served was crazy.

<!-- t=507 -->
**Taylor:** People say the same about Ruby now and hate on it, but it's silly. The productivity you can get from these things is huge.

<!-- t=517 -->
**Sam:** We still use Rails for our API and web app. People ask us how we ship things so quickly, and Rails makes this extremely easy. You just use a new gem for this and it works. Sharding in Rails is natively supported. I think this generally speaks to the power of really good frameworks. It seems like the JavaScript community has settled a bit now, so you start to build these sedimentary layers. But the sedimentary layers of Shopify and GitHub and all these people working on Rails are crazy. They built sharding logic and all these things that other companies would spend years on at much lower quality, and it's just free and open source through a framework.

<!-- t=562 -->
**Taylor:** It's huge. It's a similar story in Laravel. For a lot of the companies that work with it, their Laravel teams ship incredibly fast because of what's available to them in the framework or through the package community.

## From open source to business: Forge

<!-- t=583 -->
**Simon:** When did it turn into a business? I was asking what the inflection point was for the growth and popularity of the framework, and it sounds like it was an aha moment for a community that felt this thing the other languages were getting didn't exist for them. There was no Django or Rails equivalent. And then you had the intrinsic motivation to write documentation. I feel the vast majority of open source projects stall out because people genuinely don't know how to use them. They're sponsored by some company whose experts just don't need to document them. Very few open source maintainers actually want to do the documentation, but the ones that take off just have great documentation. Rails had great documentation, right? You over-invested in the docs. At some point it sounds like you had a day job and started writing this for the day job, but then decided maybe there's a business here. Tell me about that transition.

<!-- t=641 -->
**Taylor:** I released Laravel on GitHub in the summer of 2011. For the first three years of Laravel's existence I made no money from it. It was a pure open source project that I volunteered on. About six months after I released it, a guy named Ian Landsman emailed me. He said, "I saw your new PHP framework. We're on PHP and we want to modernize onto your framework. Can you come work here?" Laravel was only about six months old, so this was a pretty risky move on his part, to rewrite his company on this very new framework. So I joined the company. That was an opportunity to build a lot more real-world things into Laravel that we learned from building real production applications, very similar to Rails with the Basecamp situation. I worked there for several years, and it was a great maturing time for the framework.

<!-- t=699 -->
**Simon:** Was that the only company running on it at that time?

<!-- t=705 -->
**Taylor:** Oh no, there were thousands and thousands of companies running on it. But I didn't have a business around Laravel. I worked for someone else using Laravel.

<!-- t=714 -->
In 2014, I had the idea to build a deployment platform for Laravel. Unlike Heroku or Vercel, which manage all the infrastructure, you could bring your own AWS account, your own DigitalOcean account, whatever. You link it to us and we give you a dashboard for creating servers, managing sites, configuring PHP or MySQL, and so on. I built that in 2014 and released it at Laracon that summer in New York City. I really had no idea what was going to happen. But within the first few weeks I was making more money than my day job, and within a few months I was working on Laravel full time. It was still just me at the company, and it was definitely over a million in ARR before I hired anyone. The next year I hired someone. The company was bootstrapped until 2024. We grew to nine or ten people and made north of $50 million bootstrapped, with 20,000-plus customers. It was me, eight programmers and one customer support person.

<!-- t=789 -->
**Simon:** What were you selling? A script?

<!-- t=797 -->
**Taylor:** No, it's a SaaS platform. You sign up, and they would pay us 19 bucks a month or 39 bucks a month, depending on the tier, to have that dashboard they log into. The underlying infrastructure was on their own AWS account, which was really convenient for us because we weren't managing the real infrastructure. The business was really high margin. I was doing about 92, 93% profit margin on Forge the whole time it was bootstrapped.

<!-- t=831 -->
**Simon:** Amazing. This is the original BYOC, right? It's impossible to beat the margins on something like this. And it sounds like you had incredible revenue diversification because it was so cheap, since you're just charging for the dashboard. That's extremely high-quality revenue. That's very impressive.

## Running developer conferences

<!-- t=849 -->
**Simon:** Another thing I wanted to ask you is that you've run conferences for a really long time. We're starting to talk internally about running conferences. I'm not going to do it personally. I have a team that's going to do a much better job than we would. What are some of the things you've learned that you would pass on to others who are starting to dip their feet into running large conferences for dev community companies?

<!-- t=879 -->
**Taylor:** It's a good question. We've always tried to keep our conferences authentic and true to the dev community. They do not come across like corporate events or opportunities to pitch you things. There are lots of talks about real engineering topics, open source libraries people are working on, cool stuff people are doing. Laravel has always had a really passionate community. I've seen Laravel tattoos. I've seen kids named Laravel. One guy named his daughter Laravel.

<!-- t=915 -->
**Sam:** That's amazing. No shade on baby Laravel, but that must be a surreal experience.

<!-- t=926 -->
**Taylor:** It is insane. We did our first Laracon in 2013. There were like 90 people there, a pretty small event in Washington, DC. It grew every year. The next year was a couple hundred, then 500, and now a thousand or so.

<!-- t=940 -->
**Simon:** Let me ask the question differently. Say you're brought into a new company and you have to run the first meetup, 100 or 200 people. What's happening on that Tuesday to make it the authentic event you're speaking to, with everything you've learned over the past 15 years concentrated into a day?

<!-- t=964 -->
**Taylor:** It's hard to answer, because there's so much groundwork that has to happen beforehand, building out a community that even cares about such an event.

<!-- t=983 -->
**Simon:** There are a lot of slop conferences, right? If you don't have a community and you throw a conference, it's just going to be a bunch of vendor talk. You've done it too early. So let's start there. What's the prerequisite?

<!-- t=1000 -->
**Taylor:** The prerequisite, and there's no easy way around it, is that you as a community or a company put a lot of value into the world over a period of time, to earn the street cred to throw an event where people think, "This is something I'd be interested in going to." For me, that was a lot of open source work, video content, blog posts. I think companies like PlanetScale or turbopuffer that put out really high-quality engineering content are earning street cred in that way, so developers think, "I would like to go to this."

<!-- t=1042 -->
**Simon:** I like the way you said it: value into the community. It's a really fair value exchange. When you really distill down marketing and dev tools, the only way to do it is to educate people, to teach them how you see the world. If they take that because they think it's a great way to see the world and they feel educated by it, then they'll remember you. That's really all you can do.

<!-- t=1072 -->
**Taylor:** And there's no trick around that. There's no scheme to trick the community. It has to be real.

<!-- t=1082 -->
**Simon:** Agreed. I've attended one Laracon, last year's in Denver. It's large, right? Two thousand people or something at that point. But it was still single track, and there was something about everybody having the same experience. You all saw the same talks and learned the same things, and there was a lot more camaraderie than at the ones where you're walking around and everyone's doing different stuff. So the prerequisite is that you've brought real value into the world, you have a real customer base, a real community, real community love. You have people who in the hallways don't attend a single track because they're meeting their buddies from Slack or IRC, people they've met through this community. That's a milestone in the community, when it's strong enough that friendships are formed in it that wouldn't have existed otherwise. So then on the day, what happens? I feel the weight and authenticity of the event is there, but what happens on the day?

<!-- t=1150 -->
**Taylor:** The single track is an interesting point. We've always done single-track Laracons. Laracon roughly looks like eight to nine speakers per day over a two-day event. They're essentially all engineering talks. Sometimes we'll have one or two business-related talks or something relevant to growing a software company, but most of the talks are engineering-related. They actually have code on the screen. It is a very engineering-focused event. It is not a sales event for Laravel Cloud hosting or anything like that. It's about real engineering problems that are interesting to the developers there.

<!-- t=1189 -->
**Simon:** And what happens outside of the event?

<!-- t=1191 -->
**Taylor:** Oh man, all kinds of crazy stuff. This year we did dodgeball. We did a Lara Prom. We've done golf tournaments. We've done basketball with terminal.shop and ThePrimeagen and all them.

<!-- t=1207 -->
**Simon:** Where can they find you after 6 PM? Where are you at Laracon?

<!-- t=1213 -->
**Taylor:** I go out with everyone else. That's something I'm very intentional about. Wherever the people are going, I will go. I'm not in some VIP room, I'm not hiding in the back with all the other speakers. If they're at a bar, I'm at the bar. If they're at a bowling alley, I'm at the bowling alley. For me, that's how it should be.

<!-- t=1240 -->
**Ben:** Sorry to cut in, I just want to say that watching you two talk is so fun for me, because I know both of you. I know your businesses, and I know how similar they are from the origins and the things you've done. Having you guys discover this together is amazing.

## What's Arkansas about Laravel?

<!-- t=1253 -->
**Simon:** Well, I'm about to ask a question that is very much that. What is Arkansas about Laravel? I'm in Quebec right now. Is there something you can trace to Arkansas culture, you being there, in Laravel, the way you run the company?

<!-- t=1292 -->
**Taylor:** I think the common-person feel of the way I like to interact with the community is very Arkansas. This is not a stuffy, elite state by any means. It's a very working-person state where there's not a lot of pretension. I really see myself, honestly, as an average programmer, and in many ways that was an advantage for building Laravel. I'm an average programmer, so I'm going to build tools that are understandable and let other average programmers get a lot of leverage and build pretty cool things pretty fast. That's how I interact with the community. I'm just trying to get shit done and I'm in the trenches with you building stuff myself, and we're figuring it out together. To me, that's a bit Arkansas.

<!-- t=1356 -->
**Simon:** I think it's more counter to the Silicon Valley VC playbook, which often isn't that. It's the same Walton approach to running a company, of being in there and popping into every retail store you pass by.

## Why raise venture capital?

<!-- t=1383 -->
**Sam:** I was going to touch on that. We all have profitable businesses, but we've all also raised VC funding. You both had a really impressive, almost bootstrapped start, but you've also both raised VC money. So how did you think about doing that, Taylor? What was the time, what were the motivators in the business? Simon, it became public that turbopuffer hit 100 mil of revenue, which was an exceptional win for such a tiny team, and you still raised as well. You obviously had higher revenue numbers out the gate, so why did you take the money?

<!-- t=1420 -->
**Taylor:** For me, it was a very personal decision. This was late 2023 coming into 2024. I'd been getting emails from VCs for years and deleting them. I didn't even know which VC firms were legit or good or bad. I deleted them all, didn't care. Life was good, I was making a lot of money, but at the same time I had built everything I'd set out to build with Laravel. I'd built all the packages, built the businesses. The only thing remaining was bigger, ambitious swings, which ended up being Laravel Cloud, our fully managed hosting platform. And similar to what you said about why you raised, Sam, building a fully managed database or fully managed compute offering is just not something I wanted to bootstrap.

<!-- t=1469 -->
For me personally, I was at a crossroads. With Laravel, I could essentially semi-retire, just coast, maintain the framework, make sure the bugs are fixed, but not actively build new ambitious things. That was a fine option, I guess.

<!-- t=1491 -->
**Sam:** When you looked at that fork in the road, what did life look like for you? Did you think you'd get bored? It's not a terrible life to retire at your age and chill.

<!-- t=1500 -->
**Taylor:** I don't know if I would have gotten bored, but really it was about the fact that people had been passionately devoted to Laravel for 10-plus years at that point. To coast or semi-retire at 38 years old just felt weak. It felt like letting down all of those people who'd say, "Dude, I got a tattoo of your logo," and you're just going to coast, instead of trying to go harder and make Laravel as big and great as it could possibly be. I felt like I still had gas in the tank. I'd been successful bootstrapping, and I also wanted to see whether I could be successful in a whole other world of VC-backed companies. So it was a mix of a personal challenge to test myself and a commitment to the community not to let them down and just coast away and let things fizzle out. That would have been really lame and not the ending I wanted for Laravel.

<!-- t=1571 -->
**Simon:** It's an interesting question to talk about the retirement point. I'd love to hear your perspective on this too, Sam. It's this tech fantasy, but I think it's really hard to get really good if that's what you always wanted to do. After I left Shopify, I thought, "I'm going to spend some time in the woods, get a six-pack," and went all in on that. But I got good at the thing in the first place because I loved it so much, so I stepped away to rediscover how much I loved it. Then I consulted for a little bit, because it felt like maybe that was the right set of compromises. I'd help other businesses and get to take some breaks in between. But I found out that I missed the intensity. I also found out that the people who are really good at the stuff I really cared about want to work really hard, because that's how they got good in the first place. By ablation, I rediscovered what it is that I love so much. Now I can't imagine doing anything else. This is it. And if it's not fun, it's a skill issue on my part for not structuring it in a way that's more enjoyable.

<!-- t=1671 -->
Maybe speaking a little to the venture side, there are six or seven different reasons to raise capital: fund R&D, fund growth, fund the founder's ego (a very popular reason to raise), fund employee liquidity, strategic partnerships, M&A, and maybe one other that I forget. You have to be very honest with yourself about why you're raising. It has to be one of those reasons. It can't be "because I got a term sheet from this person on a call and it was really great." That's not a real reason to raise. Our first raise was to fund R&D. We had spent way too much of our own money and had to spend more to test the idea. When you say yes to VC money, you're saying yes to a particular growth rate. What became equally clear in my mind is that if I want to work with some of the best people in the world, they also want to be on that growth cycle. That's not always true in every context, but for the domain I'm in, I think it's true. So the first raise was to fund R&D, and the second was to fund liquidity for the employees. These are some of the best engineers in the world, and they deserve to be compensated on that growth trajectory, contingently so over time. I think that trajectory is compatible with the one I want to have.

<!-- t=1759 -->
But there are so many businesses that would be amazing businesses if they had not taken funding, and we see them all the time. We see these massive raises from companies, and they read like an obituary. They're like, "This doesn't make sense." These founders are getting ambition-bullied by the VCs: "You're going to do this whole thing." I did a tweet about this, about the gap between your proven execution and what you say you're going to do. If that gap is too great, it's not ambition anymore, it's just dilution. And the bad VCs can talk you into that. Anyway, that was a long rant about VC funding.

## The dangers of over-raising

<!-- t=1802 -->
**Sam:** I think we should dig in there a bit, because the engineers listening to this don't understand the dynamics of what happens when companies raise funding. When you see these giant mega-rounds, like companies raising $10 billion or raising at a $10 billion valuation, the Twitter takes are always wrong, and then all the CEOs are back-channeling, and we all know what it means. You obviously want liquidity for people at high prices, but it comes with an expectation that can absolutely finish you as a company if you screw up. You'd better hit that number. Whenever I've gone to raise, I've thought there's an upper and a lower band that I would accept the valuation at. I never want to go over, because that sets you up for much more pain, potential down rounds.

<!-- t=1861 -->
When we really got going, around the first time Taylor heard of PlanetScale, we had a competitor that we never name. We have since far surpassed their revenue growth and everything. They were valued at $5 billion, and their growth slowing destroyed their company. They're just not getting that money again. We're watching companies do it in AI now. I don't think there's one big AI bubble, but there are going to be lots of mini ones. When crypto popped, even companies we know that just host websites dropped revenue 40%, or missed quarters by 70, 80% the first quarter after crypto. If you can't keep the revenue momentum going, you are in trouble, and it's revenue momentum that means the most for continuing forward. You can't backtrack.

<!-- t=1921 -->
PlanetScale Postgres and Neki just would not exist if we didn't have the safety to deliberately take our foot off the growth pedal to build two new products. We were profitable, we had our destiny, and now the business is re-accelerating extremely rapidly because we had that time. If we'd over-raised, that time is not yours. You're never getting it back, because you're off pace. If you're at a company right now that's valued at $20 billion and your growth is slowing, you're in really serious trouble. People don't understand that. They look at that big number and think their equity is worth $20 billion. It surely is not. ClickUp was the lesson, right? Masses of money in the bank.

<!-- t=1968 -->
**Simon:** Was it ClickUp? The one that just sold for like...

<!-- t=1974 -->
**Sam:** Airtable. Airtable, sorry. Masses of money in the bank, huge revenue base, growing too slowly. You've got nothing to spend that money on to re-accelerate your business. You've tried, probably done tons of marketing. You're just wedged. People don't understand that. They look at the top-line number and all the glamour and don't realize they've set themselves up. You'd better win. You'd better not lose your biggest customer, whatever.

## Preference stacks and employee equity

<!-- t=1998 -->
**Simon:** The top-line numbers are also really problematic. This sometimes puts us at a disadvantage when a candidate is comparing two offers. If we're handing out equity at some price and another company is handing out equity at some price, and you get this many shares, you have some idea of how much you're getting. But I've never seen a company disclose its whole preference stack. When you raise money from VCs, generally, not always, they get their money back at some multiplier before anyone else with equity sees that money.

<!-- t=2040 -->
**Sam:** Explain that through the eyes of an acquisition, because that's another thing people find very confusing. Let's imagine a company gets bought for a billion dollars. This is heavily simplified and I'll probably screw up the math, so check me. You think you have $10 million worth of equity. The company has raised, let's say, $100 million in capital. Well, maybe $300 million, $400 million, $500 million of that has to go back to the investors before you see anything. By the time you pay the taxes and everything is cleared, you might have a much, much lower number than what's actually represented. So I think it's borderline not a scam, but dishonest, not to represent the fully diluted number after the pref stack.

<!-- t=2100 -->
I think often it's not done because the math is quite complicated, and you're disclosing all kinds of terms that most people don't disclose. People know this, which is why the liquidation preference exists: you can focus on a headline number. The equity that we have, we represent after everything, but it puts us at a disadvantage. It's one of those things where you don't know what to do, so you do the most transparent thing, which is to show the fully diluted number after the pref stack.

<!-- t=2128 -->
**Sam:** It's really nuts, some of these dynamics, and how employees can get screwed by bad raising strategy. These things are very unintuitive when you set out to build a tech business. You start thinking the way we all did: build something people want. We're all doing infrastructure at some level, so it seems intuitive to build those things. Then you have to learn to market a company and sell a vision publicly. And then you not only have to execute perfectly on what you build, how you sell it and what corner of the market you choose, you also have to raise exactly right so you don't screw yourself up there as well. With timing in the market on top of all of that, it's really tough.

## AI adoption in the Laravel community

<!-- t=2179 -->
**Sam:** Speaking of the market, Taylor, how has your audience adopted AI? What's the AI zeitgeist in the Laravel community?

<!-- t=2188 -->
**Taylor:** I don't know if it's much different from many other communities, but I haven't written code by hand all year, basically, and I think most Laravel devs are probably similar. We were pretty early in releasing local MCP servers to hook into your agents to make building Laravel apps with AI better, like feeding them all the latest docs. Early on we put out skills from Laravel.

<!-- t=2215 -->
**Sam:** I remember Laravel Boost, I think you announced it like a year ago, right? That was before this was a popular thing.

<!-- t=2224 -->
**Taylor:** Yeah, definitely. So it's everywhere, same as everywhere else, I guess.

<!-- t=2233 -->
**Sam:** Have you met people who are truly never using AI? My Twitter mentions seem to be full of people who just won't touch it, which is obviously absolute insanity.

<!-- t=2243 -->
**Taylor:** Super rare for me. In my experience it's really rare in Laravel.

<!-- t=2250 -->
**Sam:** After talking to people who go to conferences all around the world, you'd be absolutely horrified by the number of people who are just really resistant. It's unbelievable. I do think there's a skills extinction event coming, and it's completely optional. I've said this before: there's not a single software developer who has to be replaced by AI. They're choosing to be. I got into a back and forth on Twitter with someone who refused to believe you can do QA with AI as well as a human, and that's completely nonsense. I said to them, and it was mean, that I don't think you can call yourself an engineer anymore. I think you have to have some level of modernity to still call yourself an engineer. In tech, especially software, there are tools that have made people orders of magnitude better and require so much less human toil to keep quality high or remediate incidents. If you aren't aware of those, you're probably already close to not being classed as a contemporary software engineer, in the same way that if you're still only writing COBOL, you're a COBOL engineer. That means something, but we all know what it means, right? That's the meme. We're about to get the biggest COBOLization of the tech industry, when it splits between barely any AI use or primitive AI use and what very modern AI users are doing. I think it's going to be one of the most fascinating things to watch live, even from a sociological perspective.

<!-- t=2360 -->
**Taylor:** Crazy. I do wonder, though, because at least from my perspective it's way easier to catch up and learn. If you refused to learn some new programming language a decade ago, it might take you a year to get good at that thing. But you could get up to speed and get very good at using AI in days or weeks, if you're open-minded about it. So there's also the flip side: if someone changes their attitude, they could catch up very quickly and be relevant again.

<!-- t=2386 -->
**Sam:** I think overcoming obstinance is a very hard personal journey for people. It's really tough.

## Does AI coding feel sad or fun?

<!-- t=2394 -->
**Sam:** Taylor, you spent a long time handwriting Laravel and getting into the code, and I know you love the craft of coding. You say you're not a very good engineer, and we all know that's not true. You think very deliberately and craft very deliberately, and I know you care about your users and what they experience. From the conversations I've had with you, you thought very thoroughly about how tactile Laravel was. Do you feel any sadness from not doing this anymore?

<!-- t=2419 -->
**Taylor:** I was kind of sad at first. I think Laravel and Rails are very similar this way, and I asked DHH about this in Copenhagen. We both built our identity around handcrafting these delightful code interfaces meant to be consumed and enjoyed by humans. I got hooked into AI last December, when the coding agents really blew up, and stopped writing code by hand. At first I was really uncomfortable with it, but I found myself having more fun and getting more done. I could still steer the architecture of the code and lay the foundations how I wanted, and let the AI do all of the junk I didn't want to do, like writing a bunch of tests or filling in boilerplate once the architecture is defined. I learned that the fun for me was never about the individual keystrokes of the code. It was about solving problems, shipping products, building really great experiences. AI didn't stop any of that. In fact it helped me do more of it. This year has been the most fun I've had coding in recent memory, for sure.

<!-- t=2508 -->
**Sam:** That's incredible. I'm having insane fun. Simon, is Taylor's experience similar to yours?

<!-- t=2516 -->
**Simon:** Frankly, I find the topic almost a little boring, because it's so obvious to me that I'm not riding a fucking horse and buggy. I'm going to use the greatest and latest tools. It's opened up a lot of creativity. There are lots of things I wouldn't have done before: tests I wouldn't have run, benchmarks I wouldn't have run. I redesigned the docs, which is probably the most major engineering thing I've done recently, and it was so fun. It's a creative outlet that's 10x in the amount you can do. I think it's changing the way we build product. Handcrafting little syntactical things is just not as important anymore, and it means we can move faster on some product things. We're more explicit in the APIs we build. I spend more time thinking about where the experience goes. Do people care about a dashboard? What do they care about in the dashboard? So I feel more downstream from this. If someone shows up and says they ride a horse and buggy, I'm like, okay, what do you do for work? It's not particularly interesting to me to discuss the tools. All the best people I know are using them. They adopted them at different rates and use them in different ways, and I trust the best people in the world to use them correctly.

<!-- t=2604 -->
What's interesting is what it does to product. I don't think everyone wants to consume everything through the ChatGPT interface. You want to see some graphs, you want some opinion. I don't think you want to ask AI everything. You still want an expert to decide what matters and what doesn't, and then at some point the interface is good enough that there can be progressive disclosure, exposing more details. I hate reading other people's AI output. This is the new boomer thing. It's the opposite of "have you Googled this?" Someone just sends you their Claude link, and I react with an inappropriate emoji because I'm not reading that.

## MCP, models and system complexity

<!-- t=2650 -->
**Sam:** I'm glad we've normalized around that, and glad people got sick pretty quickly of reading other people's AI slop, because it was extremely wearing for a while. I'm glad people are getting booed out of doing AI slop posts. But I'm with you. Our whole product surface area has changed dramatically and will continue to. Our MCP server is my favorite thing ever. It has created so many magic moments for our customers that I will never fall out of love with it. We expose so much to the MCP server that people do one line and fix all their database problems. We could never do that for them before. We could never write code for them. It's actually crazy how much use it gets.

<!-- t=2696 -->
**Simon:** It's the end goal for infrastructure.

<!-- t=2701 -->
**Sam:** Every day you wake up and there's a new person tweeting a picture of our Insights graph: "I hooked it up to MCP, look at all my P99s." That's our dream, our favorite picture to see on Twitter. It does it for them now. It's amazing. You get the insights from any of these products, and you have the best systems engineers in your pocket now.

<!-- t=2724 -->
**Simon:** The only thing that bothers me about the engineering of the LLMs is that they're still very bad at systems architecture. You and I talk about that a lot, Sam. Their trade-off between simplicity and complexity is not, in my view, where it is with most senior engineers. The cost to implement complexity has gone down substantially, but there are so many second- and third-order effects to complexity that these models don't understand, because they know they can just type it out, a hundred thousand tokens, and run it. They are not properly weighing complexity versus simplicity.

<!-- t=2765 -->
**Sam:** That's also our role in the way we teach people to use the database. We have to do a lot of education on what you can do with our database, what a good query looks like, what a good schema looks like. We want to get people to the fastest, most efficient implementation as fast as possible, just like you. This is where sometimes we have to fight the models that are trying to do something incredibly complicated, and that's a little tougher to do. I do wonder, does turbopuffer speak SQL, or is it a custom query language?

<!-- t=2803 -->
**Simon:** Not yet. If you're building a database company, you end up speaking SQL at some point, unless you want to go down some obscure query path, which I don't think is a good idea. But there's an idiosyncratic point you hit along the number of query plans you've implemented. A query plan is where you take SQL and decide what it's going to do on the indexes, what pages it's going to visit. You learn more and more query plans over time, and at some point you have enough of them that exposing them through SQL is the easiest thing for the human, for the model, for everyone. We haven't reached that point yet, but we're getting closer by the day. turbopuffer can do more and more. For now it's an idiosyncratic but very simple query language, so the models are probably quite good at it now, actually.

<!-- t=2852 -->
**Sam:** That's why I brought it up. When things speak SQL, models know SQL so well, especially something like Postgres where all the performance behaviors are well known by the model. It's a different challenge if it's a new thing the model is trying to learn. The skills will be more important.

<!-- t=2868 -->
**Simon:** The sophistication of our customer support is so much more fun now. Being close to the customer and having domain expertise is a fun job now, because it's not "oh, you mistyped this thing." That doesn't happen anymore. From the customer channels I'm in, the basement has been raised so substantially, and the roof has been raised so substantially. Linear rebuilt their sync engine on turbopuffer because they realized it's about a bunch of set intersections, and a full-text search engine is really good at set intersections, which B-trees are not.

<!-- t=2912 -->
**Sam:** This is dying for a Ben post.

<!-- t=2914 -->
**Simon:** Yes, a Ben post. There are particular queries that B-trees are not good at, and very long WHERE conditions is one of them. I don't know the guy who works on it that well, but presumably he sat and talked to an LLM about this, or somehow had that insight. I think that's super exciting. I think documentation still matters a lot. We make ours very example-heavy, but I think it's your duty now to use these LLMs as much as possible so you can figure out how to make the product the best.

## Smart but not wise

<!-- t=2956 -->
**Simon:** The one thing I find very interesting about the models now is that they don't seem bottlenecked on intelligence anymore. The very visceral image I have in my mind is that some of these models are a bit like Einstein in solitary confinement. They need to be hooked up to the company, all the rules of the company, the context of the company. That's where we play in as database companies, supplying the right context for these models to reason with. Intelligence does not seem like the bottleneck for these models anymore. They are so smart.

<!-- t=2998 -->
**Sam:** I say they're smart, but not wise. Going back to what you said about system complexity: I'm almost certain the model companies are trying to address the wisdom issue, but that is much harder. Getting the code knowledge was not difficult, or at least getting the data right, because it was GitHub. There was so much out there. The wisdom has never been written down. It lives in people. The whole reason you hire for experience and an engineering culture is to transfer wisdom or bring it in. When you're working at the edge of something, you cannot buy a book that is going to help you. Now, thanks to VMG, there are amazing articles out there on Git hosting. But in the GitHub days, we didn't say, "Okay, we've got a problem, let's go get the O'Reilly book on Git platforms." There wasn't one. It's the same with everything.

<!-- t=3053 -->
To help people understand where I'm coming from: at Meta, there are teams of 10 to 15 people who work on network drivers and protocols, because when you own many millions of servers, any change anywhere has a massive impact. All of this deep knowledge and experience is just not leaving these big companies. It's the same with database knowledge. Why is PlanetScale the company it is? We have a hundred-person company with 600 years of database experience, and none of us have spent any of that time writing it down to get it trained anywhere. So I think the gap is widening: the people who have gained this experience and knowledge, and for whom writing code was the thing that got in their way, are now so powerful. If you're a little older in your career, have experience, and feel like you're lagging behind, know that you're sitting on a bank of skills that are not in the models, and the models will let you go so fast. It's kind of amazing how much more value experience has now. Until a model says "no, you shouldn't really be using Spanner for this, you should be using sharding or whatever," it's not a wise model. People were worried about competition, whether it's all going to round to nothing. Nope, because my competition's experience means they'll type something completely different into the LLM than we will. That's where competition and edge come from.

<!-- t=3150 -->
**Simon:** I like "smart but not wise." That's some good thought leadership, Sam. You should have a podcast or something.

<!-- t=3162 -->
**Sam:** I'm not a podcaster.

<!-- t=3167 -->
**Simon:** When I had to explain LLMs to people early on, I said it's kind of the average of the internet. That's how it works. There is no PageRank, right? PageRank being that some website is more important if a lot of people link to it. That also is so gamified now that it's not even a ranking signal. You can't bootstrap anything useful with that anymore. So what is that today? I was tweeting the other day that a random fireside chat, recorded on someone's phone at a 50-person random meetup, with someone deep in the trenches with their team building software, is in my personal PageRank worth 100x a polished TED talk. The model doesn't know how to do that. If I'm designing a database, it doesn't know "these are extremely credible people who know how to do that." The weights don't say that they must know something about it. They do say things that are, you know, not that the earth is flat. So there's some of that, but for the wisdom, you have to figure something out. Maybe that's what we bring.

## Don't let models pick your infrastructure

<!-- t=3246 -->
**Sam:** It becomes obvious when you start asking where to host your database that they're not very wise. We know that famously, and I'm sure Laravel has exactly the same problem. If you ask it to start a framework, it's going to choose a JavaScript one. Right now the models basically only recommend based on ease of getting going, which means how much breadth of examples they've seen, or whether there's a free offering. It's optimizing for the wrong thing. If you ask an experienced person how to design a really big, complicated problem from first principles, they're not going to say, "First of all, let me find someone that has a $25 free tier that this will fit into, otherwise the problem's wrong." That's silly. And no one is going in saying "I'm optimizing for long-term maintenance." They just ask where to start. That's where I think the skill gap is going to widen. If you're taking recommendations from a model, you're going to get stuck. If you're choosing infrastructure purely because of what a model says, you're going to get stuck at the bottom end of tech basically forever. The point of creation is not in any way instructive of where your architecture will end up. It's cool if you can create a database super fast. I remember some of the other Postgres hosts glazing each other on Twitter about how fast they can start a database. Great. The GitHub database was started once and was a continuous running system for 15 years. Well, actually they created it a couple of times because it got dropped famously, but you know what I mean. Same with Shopify. Day one is immutable and it doesn't matter. Those day-one infrastructure choices are not relevant, and the fact that it took 10 minutes or 10 seconds is not why you should pick them. Taylor's Lamborghini goes extremely fast around the track, but it doesn't mean you should be using it for shopping.

<!-- t=3385 -->
**Taylor:** But I do, sometimes.

<!-- t=3390 -->
**Sam:** You do. You can and you should, for fun. But you get my point, right? The point of creation is a silly time to optimize for.

**Taylor:** Totally.

## Audience question: mobile apps for Laravel

<!-- t=3398 -->
**Sam:** Do we have any questions, Ben, from folks?

<!-- t=3402 -->
**Ben:** I wasn't tracking all the Twitter stuff, it doesn't come through on here. This one's from way earlier and it's for Taylor: thoughts on mobile apps for Laravel?

<!-- t=3414 -->
**Taylor:** I've always been a web person, so I've never done any mobile development or built mobile apps. There's NativePHP, and companies do it. They have a PHP team and build native apps in PHP. It's pretty wild. It can be done, but I don't have any firsthand experience with it.

<!-- t=3441 -->
**Sam:** I'm sure there were a bunch of other questions on Twitter. Unfortunately this app doesn't pull in the Twitter questions. I just opened up Twitter and there are no questions. Taylor, is there anything you want to say to the audience?

## Staying intense as a remote company

<!-- t=3465 -->
**Taylor:** Honestly, I was going to ask you guys a couple more questions. We're wrestling with this at Laravel. One of our biggest challenges has been that we're a remote company. PlanetScale is not remote at all, right, Sam?

<!-- t=3485 -->
**Sam:** We're pretty remote. We've got maybe 20% of our employees in the Bay Area, and they don't all come into the office. The office is a place to go and do remote activities in a nice place.

<!-- t=3498 -->
**Taylor:** One of the things I've found challenging as the company has grown, and as a remote company, is maintaining startup intensity. Whenever we do a hackathon as a team, it's awesome. We're all there in person, everyone is super locked in at all times, we're shipping like crazy. It can be hard to feel that same intensity remotely. How are y'all staying on top of that? What are your day-to-day engineering practices? How are you making sure stuff is moving quick, that there are no distractions, that people are shipping? What have you learned there?

<!-- t=3542 -->
**Sam:** I'll let Simon go first.

<!-- t=3547 -->
**Simon:** I believe in the church of the weekly meeting. In the weekly meeting that sets up the beginning of the week, you as the CEO, the CTO, as leaders, have to beat the drum and talk about what's possible. At the beginning of the week we also have a channel called Priorities, and everyone posts what they're going to do that week. They comment on each other, and I read every single one, comment on it, and ask questions. You hold the drumbeat in that meeting. I do a meeting with the engineering team on Monday, with the go-to-market team on Monday, and with the growth and marketing team on Monday, every single Monday. You're in there. I think that helps a lot with pace.

<!-- t=3598 -->
The other thing we do for pace is in-person time, which is great. Everyone in a remote company has a different temperament for how much they want to be in person. Some people want to go to the summits twice a year. Other people are fine being on planes every two to three weeks to spend more time in person with the team. So we encourage that. As far as I'm concerned, and maybe the CFO has a different idea, there's not really any travel budget, and we encourage it to the point where, especially for the Europeans and people crossing coasts, if you write a blog post, you get a turbo credit. With a turbo credit you can upgrade your next flight. That's in service of people spending time together and getting rewarded for extracurricular activities. All companies say writing blog posts is just part of the engineering job, but we all know it's something you somehow fit in with everything else, pushing through a couple of evenings, with review cycles, putting all this compute into writing a good blog post.

<!-- t=3664 -->
We call the concept the campfire. The framing is that you come warm up at the campfire as often as you need. If someone goes to the New York office or the Ottawa office or San Francisco or Paris, other people think, "Hey, I want to go. It's a good time, I don't have anything going on, I'm just going to hop on a plane and go." One of the things we're starting to do with these campfires is formalize a little mini, what we call a hack stay, when people are in person.

<!-- t=3699 -->
But I think the majority of the drumbeat comes from the leaders. I'll drop into a project and say, "If you were to ship this tomorrow, what would blow up? If we're okay taking some risk here, what would it look like?" As founders and CEOs, it's our job to help people take risks in the business and to beat the drum as hard as we can, but also not beat the drum when some people need the space. Towards the later part of the week we have our all-hands, where I'll call on people randomly to come present demos. Sometimes people just sign up for demos of whatever they're working on. I think Sam has formalized that even further than we do. I don't think remote is a disadvantage, but you have to be able to command a lot of energy on a Zoom call, I think, to beat the drum.

## The CEO sets the pace

<!-- t=3755 -->
**Sam:** I want to emphasize that being the key one. There's great process that underpins this, like weekly demos. And I'll never share it, but Ben's team has put an insane blogging process together. It's quite regimented, but it produces content at such a pace. It's a very similar thing that trades off against everything you just talked about, but the irreplaceable thing, I think, is the CEO's drive. Usain Bolt and I can both run a hundred meters. The speed each of us does it at is the differentiator. The pace your company ships software at is entirely arbitrary. CEOs who bemoan the pace of their team sound terrible to me. It's naughty to blame anyone but yourself. You can decide how quickly things have to be built, and the pressure is on you to be right about that. The best CEOs choose the soonest possible moment they could ship the right thing at the right quality, push their team towards that, and are extremely unreasonable about trading that off.

<!-- t=3832 -->
It's hard, though. It's emotionally difficult. Sometimes someone's going to have to work a weekend and you feel terrible about that, but you have to know that you're out front. We're in a very frenetic environment in terms of how tech is. Ultimately you're doing the right thing for your whole organization by pushing as fast as you believe is reasonable without burning them out. You've got to be right. No one wants to follow a CEO who's making silly mistakes day in and day out, whose efforts burn away to nothing. People are going to feel safe if it feels like "if we're dying in battle, I'm dying with you." There's no individual blame. But I've worked for, and I know, people who get all of this advice, watch the founder mode talk or whatever, which everyone took the worst lessons from, and show up on Monday and turn into a demanding prick. It shatters trust in your competence. If people look at a goal and go, "Wow, that is intimidatingly quick, I'm going to have to work extremely hard to get there, but I know it's doable," you've got them. That's the exact fine line. If they go, "That's impossible" from day one, you just lose them. They think you're stupid and you can't go anywhere. And then you'd better bring back the reward. It had better do what you said it was going to do for the business, the product, the users. There's a lot to drive that pace, and everything else comes second in my mind.

<!-- t=3927 -->
**Simon:** The thing I have an allergic reaction to is leaders who only have one tool call, which is to tell people to work more. That's such lazy leadership. When you've been around the block a few times, you know it's not the only tool. The way to show pace is not "I want this, I want it tomorrow." The way to show pace is: how can we cut scope? Can we go talk to a customer and figure out if we can cheat on this and this and this, and then give it to them sooner? The greatest motivational tool is to make it feel like it matters, and nothing feels like it matters more than giving it to a customer as early as possible. Then you tell people, "It's okay if this fails, we're taking the risk together. Just talk to the customer, let's get it in their hands. We know the compromises. It might err, but they want it." So one tool is cutting scope. Another is to give them more resources. You say, "You're working on this thing over here, it's really important, but if we ship D right now it speeds everything else up, so let's put some more resources on this." These are not calls the individual engineer can necessarily make. There are lots of tools we have available that they don't necessarily have, and we can help them grow and they can ask for it. That's also part of pace: prioritization and putting more weight behind the arrows you're actually doing. I think it shows you care, it shows intensity, it shows pace, and it still means people can log on and log off and have dinner with their kids. If your only weapon is "you're just going to have to work harder," that is lazy leadership. How do you do it, Taylor?

<!-- t=4035 -->
**Taylor:** Very practically, it's super similar to what Simon was saying. We don't have a priorities channel, but in the general channel, the last few weeks, it's been "what do we want to ship this week? What do we got cooking?" And then in individual project channels, I hop in and say, "Where are things at? Are there any blockers we need to remove? Are we confused about anything? Is there any ambiguity we need to resolve around this project?" Either by talking to the customer or ironing out the scope, as Simon was saying. I was mainly asking because sometimes I'll hop in these channels and say stuff like that, and I wonder, is this what I'm supposed to be doing? What are other remote companies doing?

## Advice stops applying, and founders should own the product

<!-- t=4087 -->
**Sam:** Simon and I went hours on this on the phone the other day. As your company grows, the advice you get from other people is garbage. At the stage the three companies involved in this podcast are at, no outside advice is relevant anymore.

<!-- t=4105 -->
**Taylor:** How big is your company? Like 100?

<!-- t=4115 -->
**Sam:** About 100. But I don't think that's a function of team size, it's a function of revenue. I genuinely think that if the CEO has to get in a room and do a rain dance for money to come in the door, you do that. You just do whatever is working and don't stop doing it. I'm going to say the majority of companies start exceptionally on something and get something right. They have this one-in-a-million event where they break through all the filters. Simon, remember when I told you how many filters you'd made it through at one point in the company? I think you were at just $1 million of revenue, and 99.9% of companies never get there. A customer paying you $1 million is even more rare. You're getting through filters that nearly every startup never gets through. By the time you're through there, most CEOs look around and think, "What does everyone else do? I'm going to hire a professional product person from the outside, a professional engineering leader from the outside, and do all the things that didn't get my company here. I'm going to do the sales and marketing shit." I'm glad the founder mode realization came about, though getting there seemed so stupid to me. If you create a generational company and you were the one who did it, you should protect the product and engineering and the interface with the users and the marketing and own every part of that, because you made it. Everyone you hire is meant to scale that. But most CEOs hire a VP of product, VP of marketing, VP of engineering, give them everything, and the company falls apart. And this VP of engineering was the fourth VP of product at some boring company, and why are they running the company now? It's ridiculous. They're friends with the recruiters and the VCs, right?

<!-- t=4245 -->
**Simon:** They're friends with the last board member that got on the board. They run their LinkedIn like a VC. It's like their personal cap table.

<!-- t=4256 -->
**Sam:** Correct. The worst fucking people you can work with in this industry occupy that tier. So, Taylor, the answer to your question, which Simon and I figured out, is: whatever the fuck you want to do. We're building towards a launch happening in a month or so with the marketing team, so poor Ben has had to hang out with me. I've been writing posts, reviewing, just in the trenches because I care about it. I care about it the absolute most. I'm not going to ask this team to spend two weeks presenting to me what they're going to do for a launch. I'm just going to work on it with them. What else could I possibly be doing?

<!-- t=4301 -->
**Simon:** Sam put it really well to me, so I'll explain my general framework. First, I decided very early on that I wanted to be able to do a good job at everyone's job at the company. That's the kind of company I wanted to build. Part of the reason I'm in this is that I'm a growth addict and I want to learn how to do marketing, sales, all of these different things. I wanted it to be true that you could ask anyone in the company, except maybe the CFO, because I don't have anything to do with NetSuite, "Would you want Simon on your team? Do you think he'd be a good member of this team?" and have them say yes. Otherwise, how am I going to know if someone is exceptional at sales or exceptional at deployed engineering? How are you going to call bullshit if you don't know how to do it?

<!-- t=4367 -->
Part of my journey to that is that V1 of Simon was engineering Simon, right? Same for you, Taylor. V2 of Simon is what I call B2B Simon, learning how to sell the product. Recently I had to do the phase transition into V3 of Simon, company-building Simon. You've entered this phase, both of you earlier than I did: you're no longer building and selling the thing, now you're building a company that builds and sells the thing. So what does that look like? Sam said something to me that I think is a very strong mental model for the right way to do it, and Sam and I agree on this. If you have a project, and a project is defined as something that takes one person more than two weeks to do, then at least until some scale you want to be inserted in all of those projects. You just change the definition of a project you need to be involved in. Sam's unlock for me was that the time for you to get involved in that project is a collective intellect in the company about the best possible point. I think Sam gets very involved early on in writing blog posts at PlanetScale, is my impression. I used to just review them at the end, which is four hours and not the right point. So what is the right point in the project cycle to get involved? If you've been an in-the-weeds founder, which is my impression that you have been, then your team probably has an incredibly good instinct for when that insertion point is. We have a channel called Simon Input, and you go in there when you need Simon's input at some point. Then people have a conversation about the project timeline of this thing and what the optimal point is to insert Simon. Some of these are shorter term, like some pricing exception for a particular customer who has a weird workload that doesn't fit the existing pricing.

## Placing bottlenecks where iteration is cheapest

<!-- t=4485 -->
**Sam:** I remember this conversation. The framing I used was bottlenecks. CEOs try to remove bottlenecks, but that reduces the quality of everything you ship to garbage. Put the bottlenecks in at the cheapest place to iterate. The reason CEOs let bad things slip past them is that it feels too late to change it, and it feels terrible to redo too much work, so they let it go. They think, "I'll get this next time," and they don't. You shunt that bottleneck to the cheapest place where you could logically walk into a room and say "we're redoing the logo," and it would be the cheapest time to do that. You have to be able to say "we are 100% changing what we're doing," and this was the least effort, emotion and tears that has gone into this before we do that.

<!-- t=4539 -->
Ben's team has built this process for making content. What people don't know about Ben is that AI has allowed a lot of inferior companies to catch up, and Ben is one of the architects of the animated article blog post as a medium. Ben's IO post, where he made an SSD in JavaScript, was done by hand, with a little bit of LLM help but nothing like we have now. Ben basically popularized this way of blogging that is now the default for PlanetScale and the default for everybody.

<!-- t=4575 -->
**Ben:** I would say I popularized companies doing it. Individuals did this on their personal blogs before.

<!-- t=4582 -->
**Sam:** Yes, but companies doing it has to be done repeatedly, right? It has a different mechanism to it, and it's our number one way of driving pipeline. So look at my problem for next year: we have to keep growing at the pace we're growing. How do you increase pipeline? We don't have more Bens. So how do we hire more people to learn to do this? But also, where do we find the place where Ben himself gets to hold that bottleneck and change what his team is doing really rapidly, almost violently, in terms of what could change? That really made me think about the role of process in a company like ours that is inherently chaotic, where you want the CEO to be able to come in and change everything, because the CEO talks to customers and can take the blame for problems. The company-building element is handing that to people and asking every leader to obsessively think about where that bottleneck is. Removing all the bottlenecks is crazy. You don't have a bottle anymore, you have a beaker, and the liquid goes everywhere. So find those bottlenecks and insert them, but make sure they're in the cheapest place, and then you can scale while being very chaotic.

<!-- t=4660 -->
**Simon:** I'll add one thing to what Sam said, since you brought me up in creating content. For a while I did not lead a team of people who create content, and now I do. One of the things I always keep in mind, and it seems like Taylor, you're very much this way, is that I should never be someone who couldn't do what I'm asking the team to do. If I need to, I can still create a banger blog post that gets to the front page of Hacker News, or a great video. If I'm going to ask the team to do this, I should also know how to do it. And you review every PR going into Laravel, right? You're very much in the trenches while also leading the team, which is very cool.

<!-- t=4703 -->
**Sam:** If that's the magic that makes Laravel work, build an entire company around making sure you have eight hours to do that. Why does any company have the same job roles as other companies? Make it up. If you're that different, just fucking make it up. Seriously, it doesn't matter. The majority of our marketing team is engineers. Everyone can code on that team. The majority is engineers because that's our company. If you asked a professional CMO from some other company, they'd say that's insanity: "Where's your ABM?" No, just do what works and scale it.

## Wrap-up

<!-- t=4739 -->
**Sam:** Sorry, we don't have any more questions, and we've run over a little bit, so it might be a good time to wrap up. Taylor, thank you so much for coming on. Simon, thanks for continuing to listen to me across the years. This was a really good conversation between all of us.

<!-- t=4771 -->
**Simon:** This was awesome. Thank you so much, Taylor. Thank you, Sam. Thank you, Ben.

<!-- t=4773 -->
**Taylor:** Pleasure. Thank you.
