More from Founder's blog
TL;DR The "build vs. buy" equation has flipped. Businesses used to buy SaaS because it was cheaper than building their own. AI has changed that—building your own is now more affordable than ever. The discovery problem. AI recommendations default to well-established solutions. Think SEO is a long game? Try LLM SEO. Everyone worries about AI taking developer jobs, but what if AI wipes out the entire off-the-shelf software industry? The "Why Buy?" Problem Six months ago, we needed an AI-powered code review tool. We explored several options and ultimately "vibe-coded" our own GitHub Action—a simple Bash script that takes a git log, sends it to Claude via curl, and posts the results to Slack. Done. The best part? AI wrote the entire thing faster than it would take to sign up for a SaaS. How long until every company realizes they can do this? Need a simple "CRUD" CRM with JIRA-style tasks? Done. Need a mobile time-tracking app for remote employees? AI will spit out a React Native iOS build in minutes. Why pay for yet another SaaS when you can "vibe-code" something in a week? And mark my words, LLM providers are one step away from actually hosting the code they generate. Who needs to spawn an AWS server if you can just ask OpenAI to host the code it just wrote? - "Hey Siri! build me a Basecamp, but with green buttons, also register a domain, spawn a server and host it all there, charge this credit card when you're done" - "Absolutely, that'd be $1.17 per hour" The Discovery Problem AI doesn’t just make it easier to build software—it makes it harder for new SaaS products to get discovered. When you ask AI for recommendations, it defaults to the biggest names. And not just in SaaS, by the way, in open source too. Imagine launching a killer new JS framework today. AI coding assistants and tools like Cursor will just default to React anyway. And not even the latest version of it! In a recent tweet Adam Wathan, the creator of Tailwind, asked: "Has anyone migrated to Tailwind 4.0 yet?" The most popular response was "Nah! we're still waiting for LLMs to learn it." AI isn’t just "the next internet moment." It’s more like "the social network moment." Echo chambers get louder, big names get bigger, and smaller ones disappear into the noise. What Can SaaS Companies Do? 1. Become an Industry Standard Or at least a "go-to" product in a niche. If your app becomes something people mention on their CVs or job descriptions, you win. Examples: Slack. HubSpot. Salesforce etc. A salesperson moving to a new company simply expects Salesforce to be there. That kind of lock-in ensures survival. 2. Build Moats: Infrastructure & Vendor Lock-In SaaS products that are just CRUD apps will die. The ones that survive will own infrastructure or at least some part of it. Instead of building another AI voice assistant, create one with built-in VoIP and provide landline numbers to customers. Examples: Transistor.fm – Not just a SaaS, but also a podcast hosting and publishing pipeline. Postmark (or any transactional email service really) – yes, AI can code an email-sending app, but it can't get you a 10-year old high-reputation sender IP address trusted by Gmail and Outlook. SignWell, SavvyCal and similar "inter-business" file-sharing, communication & escrow apps that own the communication part (and frankly, are literally easier to use than vibe-code your own). But prepare for tthousands of clones. Which SaaS Will Die First? Side-project-scale, "one simple tool" SaaS products that used to be easy wins—form builders, schedulers, basic dashboards, simple workflow apps—those days are over. If AI can generate it in an afternoon, no one is paying a subscription for it. Oh, and "no code" is toasted too. The SaaS graveyard is about to get a lot more crowded. I give it 4 years. Software consulting is making a comeback though. Someone has to clean up the vibe-coded chaos.
TL;DR The "build vs. buy" equation has flipped. Businesses used to buy SaaS because it was cheaper than building their own. AI has changed that—building your own is now more affordable than ever. The discovery problem. AI recommendations default to well-established solutions. Think SEO is a long game? Try LLM SEO. Everyone worries about AI taking developer jobs, but what if AI wipes out the entire off-the-shelf software industry? The "Why Buy?" Problem Six months ago, we needed an AI-powered code review tool. We explored several options, tested them all, and ultimately "vibe-coded" our own GitHub Action—a simple Bash script that takes a git log, sends it to Claude via curl, and posts the results to Slack. Done. The best part? AI wrote the entire thing—faster than it took to sign up for another SaaS. How long until every company realizes they can do this? Need a simple CRM with JIRA-style tasks? Done. Need a mobile time-tracking app for remote employees? AI will spit out a React Native iOS build in minutes. Why pay for yet another SaaS when you can "vibe-code" something in a week? The Discovery Problem AI doesn’t just make it easier to build software—it makes it harder for new SaaS products to get discovered. When you ask AI for recommendations, it defaults to the biggest names. Here’s an open-source analogy: imagine launching a game-changing JS framework today. AI coding assistants and tools like Cursor will still default to React. And not even the latest version! Adam Wathan recently asked on Twitter, "Has anyone migrated to Tailwind 4.0 yet?" The most popular response was "Nah! we're still waiting for LLMs to learn it." AI isn’t just "the next internet moment." It’s more like "the social network moment." Echo chambers get louder, big names get bigger, and smaller ones disappear into the noise. What Can SaaS Companies Do? 1. Become an Industry Standard Or at least a "go-to" product in a niche. If your app becomes something people mention on their CVs or job descriptions, you win. Examples: Slack. HubSpot. Salesforce etc. A salesperson moving to a new company simply expects Salesforce to be there. That kind of lock-in ensures survival. 2. Build Moats: Infrastructure & Vendor Lock-In SaaS products that are just CRUD apps will die. The ones that survive will own infrastructure. Examples: Transistor.fm – Not just a SaaS, but also a podcast hosting and distribution pipeline. Postmark (or any transactional email service really) – AI can code an email-sending app, but it can't get you a 10-year old high-reputation sender IP address trusted by Gmail and Outlook. SignWell and similar B2B file-sharing apps (literally easier to use then code your own). Don't just build another CRUD sales CRM, build a CRM with an inbound VoIP number – because AI can’t replace telco infrastructure (yet). Which SaaS Will Die First? Side-project-scale, "one simple tool" SaaS products that used to be easy wins—Calendly replacements, form builders, schedulers, basic dashboards, simple workflow apps—those days are over. If AI can generate it in an afternoon, no one is paying a subscription for it. Oh, and "no code" is toasted too. The SaaS graveyard is about to get a lot more crowded. I give it 4 years. Software consulting is making a comeback though. Someone has to clean up the vibe-coded chaos.
I'm looking for a new daily driver browser on my Mac. Chrome is a non-starter for me due to privacy concerns (Google's tracking empire is alive and well), and Edge is just... too much. Every update shoves another set of “features” down my throat — Copilot, discount coupons, Bing nonsense — things I have to disable again and again. No thanks. I currently use Brave and I really want to like it, but something about it doesn't sit right with me. The constant crypto integration, some of the decisions around their search engine — it just feels like it's got an agenda. Arc? Well, Arc is dying now, so that's out. Someone suggested Zen, which is a Firefox-based browser aiming to be an Arc-like alternative. That got me curious. And since I already had all these browsers installed, I figured: why not run some benchmarks and see how they stack up? Benchmark Setup All tests were run using Speedometer 3.0 on a MacBook M3 Pro. I tested in incognito/private mode with no extensions, except where the browser had built-in blockers enabled: Chrome: Running uBlock Origin Brave: Default built-in ad/privacy blocker enabled Safari: Clean Firefox: Clean Zen: Clean Results Chrome 132.0.6834.160 - 37.7 Brave 1.74.51 - 37.6 Safari 18.2 - 37.6 Firefox 134.0.2 - 34.8 Zen Browser 1.7.3b - 31.6 Browser benchSpeedometer score (higher is better)ChomeBraveSafariFirefoxZen Browser0510152025303540 A few takeaways: Chrome is (unsurprisingly) the fastest. Brave is essentially Chrome with a privacy skin, Leo AI, some Crypto stuff etc, and the Speedometer score reflects that. Firefox holds up well but is still behind Chromium-based browsers. Not awful, but not amazing either. Zen, being Firefox-based, lags a bit further behind. If you want a Firefox alternative that looks different but runs about the same, it's an option. Otherwise, it's just Firefox with extra UI features (see below). Side Note: 1Password Is a Performance Killer One of the most surprising findings was how much 1Password's extension destroys Speedometer scores. Across all browsers, enabling it dropped my score by 10 points. No clue what it's doing under the hood, but it's heavy. Probably scans all inputs to shove a password into. A (tiny) Zen review no one asked for Zen is a very, very nice browser, but it has some rough edges: (nitpicking) Lacks standard macOS keyboard shortcuts — for example, Cmd+W should close a window when no tabs are left. There's a hidden setting to fix this, but seriously, just follow macOS conventions by default. No built-in adblocker, have to install uBlock Origin like it's 2023 again (kidding). The dev tools are Firefox-based, and that says it all. JavaScript debugging is flaky (unreliable variable watch list, breakpoints sometimes get skipped), and reverse-engineering complex CSS can be a nightmare. That said, Zen a very solid contender, and some of its UI design choices are genuinely great! If you'd like to learn more watch Theo's review
After years of working with the "big" Visual Studio, I've had enough. It's buggy, slow, and frustrating, and I've decided to make the switch to Visual Studio Code. While as a C# developer I'm still unsure if I can replicate every aspect of my workflow in VS Code, I'm willing to give it a shot—and so far, I'm really impressed. 1. Performance Visual Studio 2022 performance has been a constant issue. It's sluggish and feels increasingly bloated with every new update. It's like watching paint dry every time I open a project. In contrast, Visual Studio Code feels lightweight and incredibly fast. The first time I opened my large project in VS Code, I was shocked — it loaded in lees than a second, literally, even with extensions like "C#" and "C# Dev Kit" installed. 2. Better Developer Experience Running dotnet watch run in VS Code's terminal has been a revelation. It's fast, responsive, and actually works consistently. Visual Studio's "hot reload" feature, on the other hand, has been a constant source of frustration for me. Half the time it doesn't work, and I'm left restarting debugging sessions over and over again. I can't tell you how many hours I've lost to that unreliable feature. 3. Fewer Bugs, Less Frustration The minor editor bugs in Visual Studio have been endless and exhausting. I remember one particularly infuriating bug where syntax highlighting would break in Razor and .cshtml files whenever I used certain HTML tags or even just adjusted the indentation. It drove me up the wall! Not to mention the bizarre issues with JavaScript formatting that never seemed to get fixed. Since switching to VS Code, I've encountered far fewer bugs. It just feels like an environment that respects my time and sanity. 4. A Thriving Ecosystem The VS Code extension ecosystem is alive and thriving. Need Tailwind CSS IntelliSense? There's an extension for that, and it works beautifully. Want to visualize your Git history for a particular line (better version of git-blame)? The Git History extension has got you covered. In "big" Visual Studio, I'd report issues through the "feedback hub" and wait months — or even years — for a response. With VS Code, the community is constantly contributing new tools and improvements. It's energizing (and sometimes exhausting) to be part of such an active ecosystem. 5. Cross-Platform Flexibility One of the biggest advantages I've found with Visual Studio Code is its true cross-platform support. Whether I'm on my Windows PC gaming rig at home or my MacBook while traveling, VS Code runs smoothly and keeps my workflow consistent. Visual Studio's limited macOS version just doesn't cut it for me. Being able to switch between machines without missing a beat has been a game-changer. I have to admit, I was skeptical at first. I've always had a bit of a grudge against Electron-based apps — they've often felt sluggish and bloated. But VS Code has completely changed my perspective. It's fast, responsive, and flexible enough to let me build the development environment that works best for me. Switching to VS Code has rekindled my passion for coding; it reminds me why I fell in love with development in the first place. While Visual Studio will always have its strengths, I need a tool that evolves with me—not one that holds me back.
More in programming
Here on a summer night in the grass and lilac smell Drunk on the crickets and the starry sky, Oh what fine stories we could tell With this moonlight to tell them by. A summer night, and you, and paradise, So lovely and so filled with grace, Above your head, the universe has hung its … Continue reading Dreams of Late Summer →
I had watched enough true crime to know that you should never talk to the police. And I wasn’t arrogant enough to believe that I was different. While I felt like I knew the interrogation tactics in and out, they were repeat customers of that interaction. I wasn’t going to call. I was going to ignore it. I’m not getting Reid techniqued. Why did they ask for me? This house was owned by my mother, how do they even know I live here? Wait who am I kidding, of course they know. I went to high school here, governments have records of that kind of thing. But still, why ask for me? Another thing was odd. We lived in Brooklyn, aka Kings County. Not Nassau County. These guys must have driven all the way here on a Saturday night. I felt like I was being watched. They wouldn’t drive all the way here to just leave a business card. I felt trapped in the house. Like they were a mountain lion on a rock perch and I was the prey in the valley below. They had the high ground and I didn’t know what they could see. But this was crazy, I didn’t do anything! Should I call them? Figure out what they want? No! That’s exactly what they want. They know I feel like this. This is exactly what they are going for. Another system carefully crafted based on years and years of “user feedback” designed to manipulate you into doing what it wants. But what if I’m doing what they want right now? Maybe they don’t want me to call. Maybe the real goal is to figure out what I do next. Watching and hoping I’ll go check on the body or something. But there wasn’t a body! If I did commit a crime this would all be a lot easier, I’d know why they were here and what they wanted and could plan my next move accordingly. I opened another Bud Light, took my clothes off, and got into bed. Even though there was nobody else home, I kept the sound off on the porn. Just in case they were listening. After I finished, I felt a bit more calm. Dude get a grip, all they did was leave a business card. Coming out of the paranoid spiral a bit, I realized what it must be about. It must have had to do with my Dad’s meeting. That was in Long Island, aka Nassau County. Probably some dumb financial crap. My mother was out with her friends in Manhattan, but she’d be home tonight and maybe she knew what the meeting was. It was now twenty to nine and I texted Brian. He’s like yea bro Dave just got here come through. And you still have that case of Bud Light? I put the beers in a backpack. Is this what the detective planned? Maybe I was playing right into the plot; arrest me for underage possession of alcohol and then get me to talk about what I knew. But I didn’t even know anything! This whole thing was stupid. I thought about how I got the beers, wondering if the whole thing was somehow a set-up. Totally nonsense thought. Kids buy beer with fake IDs all the time. When I got to Brian’s everything was normal. I walked around the back of his house and opened the screen door to his basement. There were three leather couches in a U-shape, two of which were sparsely occupied by Brian and Dave. I took my place on the third empty one and put my backpack on the center ottoman. “Pretty cool, right? Yea I found it in my Dad’s old stuff.” said Brian, referring to the inflated bag atop a device labeled Volcano sharing the ottoman with my backpack. “What is it?” “Bro it’s like an old vape. You put the weed in and plug it in to the wall.” He detached the cloudy bag from the device and demonstrated. If you pushed on the mouthpiece, it let air through and you could breathe in the vaporized drug. “It’s like a bong but chill.” I inhaled. This probably wasn’t smart with how paranoid I was from the interaction earlier, but I felt safe in the basement. It was a summer night, I was with friends, I had drank beer. Life was good. Dave showed us this reel. It was a mouse in a maze, and it started from the mouse’s perspective. Kind of like a skater cam, wow these things could scurry. Then it zoomed out so you could see the maze from the perspective of the experimenter. Then seeing the back of his head looking down at the maze, cutting to sped up dashcam video of him driving home from work. Zooming out again with a sparkling line showing his route through the grid of city streets. AI has done wonders for these video transitions. Maybe this whole video was AI. “What if we’re the mouse,” said Dave in the most stereotypical stoner voice. He’d always find shit like this, in that way that when you are high the thought seems really deep. But if you think about it more it’s nonsense, like that mouse is in a maze constructed by humans, and even if it doesn’t always feel like it, the society we live in is jointly constructed by all of us. Brian showed a video of two girls at some Mardi Gras bead type event licking one ice cream cone. He told us he wasn’t a virgin but I didn’t really believe him. It was a bit after midnight and it was time to go home. I hadn’t really thought about the interaction from earlier, but I started to again when I got outside. It was a half mile walk back home; I was grateful to hear all the noises of the city. Even though I couldn’t see it, it reminded me that there was a society out there. My mom’s car wasn’t in the driveway. Maybe she met a guy. Nothing too out of the ordinary. I unlocked the door, closed it behind me, locked both the knob and the deadbolt, went upstairs into my room, locked that door, and with the blanket of those three locks, a bunch of beers, and a couple hits of the Volcano, drifted off to sleep.
The first Rails World in Amsterdam was a roaring success back in 2023. Tickets sold out in 45 minutes, the atmosphere was electric, and The Rails Foundation set a new standard for conference execution in the Ruby community. So when we decided to return to the Dutch Capital for the third edition of the conference this year, the expectations were towering. And yet, Amanda Perino, our executive director and event organizer extraordinaire, managed to outdo herself, and produced an even better show this year. The venue we returned to was already at capacity the first time around, but Amanda managed to fit a third more attendees by literally using slimmer chairs! And I didn't hear any complaints the folks who had to sit a little closer together in order for more people to enjoy the gathering. The increased capacity didn't come close to satisfy the increased demand, though. This year, tickets sold out in less than two minutes. Crazy. But for the 800+ people who managed to secure a pass, I'm sure it felt worth the refresh-the-website scramble to buy a ticket. And, as in years past, Amanda's recording crew managed to turn around post-production on my keynote in less than 24 hours, so anyone disappointed with missing out on a ticket could at least be in the loop on all the awesome new Rails stuff we were releasing up to and during the conference. Every other session was recorded too, and will soon be on the Rails YouTube channel. You can't stream the atmosphere, the enthusiasm, and the genuine love of Ruby on Rails, though. I was once again blown away by just how many incredible people and stories we have in this ecosystem. From entrepreneurs who've built million (or billion!) dollar businesses on Rails, to programmers who've been around the framework for decades, to people who just picked it up this year. It was a thrill to meet all of them, to take hundreds of selfies, and to talk about Ruby, Rails, and the Omarchy expansion pack for hours on the hallway track! I've basically stopped doing prepared presentations at conferences, but Rails World is the one exception. I really try my best to put on a good show, present the highlights of what we've been working on in the past year at 37signals, and transfer the never-ending enthusiasm I continue to feel for this framework, this programming language, and this ecosystem. True, I may occasionally curse that commitment in the weeks leading up to the conference, but the responsibility is always rewarded during and after the execution with a deep sense of satisfaction. Not everyone is so lucky as I've been to find their life's work early in their career, and see it continue to blossom over the decades. I'm eternally grateful that I have. Of course, there's been ups and downs over the years — nothing is ever just a straight line of excitement up and to the right! — but we're oh-so-clearly on the up-up-up part of that curve at the moment. I don't know whether it's just the wind or the whims, but Rails is enjoying an influx of a new generation of programmers at the moment. No doubt it helps when I get to wax poetically about Ruby for an hour with Lex Fridman in front of an audience of millions. No doubt Shopify's continued success eating the world of ecommerce helps. No doubt the stability, professionalism, and execution from The Rails Foundation is an aid. There are many auxiliary reasons why we're riding a wave at the moment, but key to it all is also that Ruby on Rails is simply really, really good! Next year, with RailsConf finished, it's time to return to the US. Amanda has picked a great spot in Austin, we're planning to dramatically expand the capacity, but I also fully expect that demand will continue to rise, especially in the most prosperous and successful market for Rails. Thanks again to all The Rails Foundation members who believed in the vision for a new institution back in 2022. It looks like a no-brainer to join such a venture now, given the success of Rails World and everything else, but it actually took guts to sign on back then. I approached quite a few companies at that time who could see the value, but couldn't find the courage to support our work, as our industry was still held hostage to a band of bad ideas and terrible ideologies. All that nonsense is thankfully now long gone in the Rails world. We're enjoying a period of peak unity, excitement, progress, and determination to continue to push for end-to-end problem solving, open source, and freedom. I can't tell you how happy it makes me feel when I hear from yet another programmer who credits Ruby on Rails with finding joy and beauty in the writing web applications because of what I started over 22 years ago. It may sound trite, but it's true: It's an honor and a privilege. I hope to carry this meaningful burden for as long as my intellectual legs still let me stand. See you next year in Austin? I hope so!
I hadn’t lost my virginity yet. And it wasn’t for lack of trying; it seemed like the rest of my generation was no longer interested in sex. On some level, I understood where they were coming from, the whole act did seem kind of pointless. But after a few beers, that wasn’t how my mind was working. I turned 19 last week. Dad flew in from Idaho, and it was the first time he was in the house I shared with my mother. He left when I was 12, and it was always apparent that parenting wasn’t the top thing on his mind. There was some meeting on Long Island. That’s probably why he was there, in addition to the fact he knew mom wouldn’t make him sleep on the couch. He had many reasons to be in New York that weren’t me. My birthday was just a flimsy pretense. He’d worked on Wall Street the whole time he was around, a quant. He wrote programs that made other people rich. But something happened to him right before he left. A crisis of conscience perhaps; he was spiraling for weeks, cursing the capitalist system, calling my mother a gold-digging whore (which was mostly true), and saying things needed to change. Then he packed a single backpack and left for Idaho. I visited him out there once my sophomore year. He had a camouflaged one room cabin in the middle of a spruce forest, but instead of the hunting or fishing stuff you might expect, the walls were adorned with electrical test equipment and various things that looked like they were out of a biology or chemistry lab. I didn’t know much about this stuff and that wasn’t what he wanted to talk about anyway. He wanted to talk about “man shit” like nature and women and not being life’s bitch. I tried to act like I did, but I didn’t really listen. All I remember is how eerily quiet the night was, I could hear every animal movement outside. My dad said you get used to it. Brian was having a party tonight. Well okay, party is a lofty way to describe it. He’d replaced the fluorescent lights in his mom’s basement with blacklights, and we’d go over there to drink beer and smoke weed and sit around on our phones and scroll. And sometimes someone would laugh at something and share with the group. I had a case of Bud Light left over from the last party and drank two of them today. Hence the thinking about sex and not thinking that thinking about sex was stupid. People wouldn’t be going over there for a few more hours, so I laid in my bed, drank, and loosely beat off to YouTube. Celebrity gossip, internet gossip, speedrun videos, nothing even arousing. I liked the true crime videos about the hot female teachers who slept with their students. Yea yea yea terrible crime and they all act holier than thou about what if the genders were reversed, but the genders weren’t reversed. Maybe they just don’t want to get demonetized. There were never women at these parties. Okay maybe one or two. But nobody ever slept with them or much thought about them that way. They were the agendered mass like the rest of us. Fellow consumers, not providers. Fuck I should just go visit a hooker. I didn’t know much about that, were hookers real? I’d never met one, and there wasn’t a good way to find out about stuff like this anymore. The Internet was pretty much all “advertiser friendly” now, declawed, sanitized. Once the algorithms got good enough and it was technically easy to censor, there was nothing holding them back. It wasn’t actually censored, it would just redirect you elsewhere. And if you didn’t pay careful attention, you wouldn’t even notice it happening. I tried asking ChatGPT about hookers and it told me to call them sex workers. And this was kind of triggering. Who the fuck does this machine think it is? But then I was lost on this tangent, the algorithms got a rise out of me and I went back to comfort food YouTube. Look this guy beat Minecraft starting with only one block. The doorbell rang. This always gives me anxiety. And it was particularly anxiety inducing since I was the only one home. Normally I could just know that the door of my room was locked and someone else would get it and this would be a downstairs issue. But it was just me at home. My heart rate jumped. I waited for it to ring again, but prayed that it wouldn’t. Please just go away. But sure enough, it rang again. I went to my window, my room was on the second floor. There was a black Escalade in the driveway that I hadn’t seen before, and I could see two men at the door. They were wearing suits. I ducked as to make sure they wouldn’t look up at me, making as little noise as possible. Peering over the window sill I could see one opening the screen door, and it looked like he stuck something to the main door. My heart was beating even faster now. It was Saturday night, why were there two men in suits? And why were they here? It felt longer, but 3 minutes later they drove off. I waited another 3 for good measure, just watching the clock on my computer until it hit 6:57. I doubled checked out the window to make sure they were actually gone, and crept down the stairs to retrieve whatever they left on the door. It was a business card, belonging to a “Detective James Reese” of the Nassau County Police. And on the back of the card, there was handwriting. “John – call me” John was my name.
The first in a series of posts about doing things the right way