Every generation of builders gets one moment when the old way of working becomes a museum piece. David Heinemeier Hansson, the Danish programmer the rest of us simply call DHH, just announced that his moment has arrived. Writing code by hand, he said, is no longer an economically productive enterprise for the vast majority of programmers at the vast majority of companies. By the end of the year, he predicts, that will be true of virtually all domains, all programmers, and all companies. So DHH has retired from coding by hand. He has not retired from building. He intends to build more amazing things than ever, and he describes the future programmer as "a professional maker of things," steering an intelligence that belonged to science fiction until a few moments ago.
He is right. I want to explain why I believe him, and it’s personal, because DHH already changed my life once.
In 2004, DHH released Ruby on Rails, a framework he pulled from Basecamp, the project-management tool his company built and my company used every day. Rails was a ready-made structure for building database-backed websites and web applications. Think of it as the difference between framing a house from raw timber and assembling one from well-made, pre-cut parts. Rails made the sensible decisions for you, letting a tiny team move from an idea to a working product in days instead of months. DHH became a hero of mine almost immediately, because he opened a door that had been closed to someone like me.
By 2005, the second version of the web was exploding in Silicon Valley, and I was sitting in Dallas watching it happen from a distance. In June of that year, I got on a plane to San Francisco to attend Kevin Werbach’s SuperNova conference. I wanted in. More than anything, I wanted to learn what everyone was using to build what we were calling Web 2.0. The answer was Ruby on Rails, which had been public for about 11 months and had not yet reached version 1.0. Standing in those hallways, it was clear to me that this framework would let me launch consumer web services quickly and cheaply. I was right about that.
SuperNova gave me the how, but it also gave me the who. I met, or was introduced to, Evan Williams, who had founded Blogger and was then working on Odeo, the podcasting company that would give birth to Twitter (a service originally built, fittingly, on Rails). I met Caterina Fake of Flickr, Mena Trott of Six Apart, Lili Cheng of Microsoft’s social computing group, and the game designer Amy Jo Kim. They were inventing the social web in real time, and they were generous with a guy from Texas.
Before the conference ended, I was sitting in a ballroom at the Palace Hotel in downtown San Francisco, sketching wireframes for a set of applications and sending them to a developer in the Philippines who had been building business tools for my company. My goal was modest and specific. I wanted to test whether DHH’s framework let a very small team build a real web application faster than anything that came before it. The first test was a URL shortener, a cousin of what you may know as TinyURL, which I called ElfURL. By the time my plane landed in Dallas, a working version was online.
ElfURL did more than shorten links. It added tagging, usage statistics, and RSS support, which turned a shortened address into an information tool. You could organize what you shared and measure whether anyone followed it. Brian Oberkirch wrote at the time that it had been built in a few days to see whether people found it useful, and that was exactly the point. I was testing the distance between an idea and a product, and Rails had shrunk that distance to almost nothing.
Less than a month after SuperNova, we launched the final version, and we were already building the next batch. We also had Frankenfeed, EgorCast, and EgorRSS, which gave you an address like username@egorrss.com and turned every email sent there into an item in your personal RSS feed. An email-only newsletter could suddenly sit alongside your favorite blogs, and the publisher never had to change a thing. Simone Carletti reviewed it favorably in October 2005. EgorOPML became FeedVault, a place to back up and share the blog lists you had spent years assembling.
There was also Gahbunga, a text-based app that let you send a photo from your phone to friends for a rating and receive the results back on your phone. Michael Arrington wrote it up in TechCrunch in August 2005. Apple would not announce the iPhone until January 2007. Gahbunga tested one narrow social interaction without first building an elaborate social network around it.
By October 2005, I was at the Web 2.0 Conference with a handful of Rails applications to show. None of them was earth-shattering. What made them notable was that my team had embraced DHH’s framework whole hog, and it showed in our speed. At that conference, I met Matt Mullenweg, the founder of WordPress, and I believe I also met Stewart Butterfield of Flickr, along with Jason Calacanis, Bram Cohen of BitTorrent, Tantek Çelik, and Michael Arrington. The common thread in that room was small teams doing what once required whole departments.
Then came Fancast, a phone-to-podcast platform for television fans. Rather than merely listening to a polished podcast, fans could call an ordinary telephone number, record comments, and ask questions of a show’s cast and creators. On June 14, 2006, at the Under the Radar conference held at Microsoft in Mountain View, it won Best in Show under its original name, PodServe. The Nip/Tuck version launched that September, and we went on to work with The Riches, The Shield, Rescue Me, and Dirt. Eventually we sold Fancast to Comcast Interactive.
That experience led me to lead the team that built ShopSavvy, one of the very first applications for Google’s Android platform. ShopSavvy turned the phone’s camera into a barcode scanner. Point it at a product in a store, and it found the best price locally and online in seconds. We won Google’s Android Developer Challenge, and ShopSavvy became a launch app for the platform, which meant traveling the world alongside Google co-founders Larry Page and Sergey Brin to show people what a phone in your pocket could really do. In 2008, few people grasped how much computing power they were about to carry around, and a barcode made the case in 3 seconds.
Rails didn’t inspire me to start writing code, and it didn’t make my developers better typists. It let me work with very small teams of developers and build things that would otherwise have required large ones. That was the whole trick, and it rewired an entire industry. GitHub, Shopify, and early Twitter all ran on the same insight.
AI is that trick again, but at a scale I can barely describe. Where Rails let me work with 3 or 4 developers instead of 30, AI lets me work with teams of virtual agents that write, test, review, and ship code around the clock. Today, one agent researches a problem while another designs the database, and a third checks the others’ work. Projects that would have demanded a staff of 50 and a budget of $10 million two years ago are now within reach of a single determined builder and a monthly subscription. The chisel is gone. The workshop is enormous.
Many will object that this ends programming as a respectable profession, and that millions of developers are about to be discarded. I understand the fear, and I think it misreads history. When Rails arrived, the number of people building software grew enormously, because the cost of trying an idea collapsed. The scarce skill was never the typing. The scarce skill was knowing what to build, for whom, and why, and that skill just became the most valuable thing a person can own.
Others will say the technology is dangerous and must be slowed down. Marc Andreessen answered that argument in 2023 in his essay “Why AI Will Save the World,” and Adam Thierer has spent years at the Mercatus Center and the R Street Institute making the case for what he calls permissionless innovation, the principle that builders should be free to experiment unless there is clear evidence of real harm. The White House made a similar bet in July 2025 with “Winning the Race: America’s AI Action Plan,” which treats American leadership in AI as a matter of prosperity and national security. The doomers have a story. The builders have a track record.
I was in college when the World Wide Web first appeared, and I remember watching a new world open up before me. I lived through the dot-com boom and its bust. I got on a plane in 2005 because I refused to miss Web 2.0, and I am glad I didn’t. What is happening now with AI is bigger than all three of those moments combined, and I do not say that lightly. I have been wrong about plenty of things over the years, but I have never felt this certain that a new technology will change nearly everything we do, from how we build companies to how we teach our kids to how a single person in Dallas can take on problems that used to belong to governments and giant corporations. Superintelligence is arriving whether we welcome it or not. No regulator is going to put it back in the bottle, and our adversaries certainly won’t try. The only question worth asking is what we choose to make with it.
DHH put it beautifully. What a privelege, he said, to have been there when we did it all by hand, and to be there at the exact moment when we switched over. Twenty years ago, his framework handed me a small team and a big idea. Today the tools he now champions hand me a team of tireless agents and an idea as large as I’m willing to dream. Unlike DHH, I never had any interest or ambition to hand-code, but I did understand how Rails would change everything. My ambition to make things has never been bigger, and I suspect yours shouldn’t be any smaller.
If you enjoy my work, please subscribe https://x.com/amuse/creator-subscriptions/subscribe
Alexander Muse is a Fellow at the John Milton Freedom Foundation and publishes daily political analysis at amuseonx.com. Primary sources cited in this piece are linked inline; campaign finance figures are drawn from FEC filings, polling data from publicly released crosstabs, and legal claims from filed pleadings. Corrections are posted to the original URL with a dated changelog. Readers who identify errors are invited to contact the author directly. Each op-ed edited for grammar and clarity using Ai in partnership with Grammarly. Data provided in a sponsored partnership with Polymarket.
AI can write the basic routines it is told to write by the designers. It cannot design anything unique. It can merely assemble what is already written somewhere else.
And if companies think quantity or innovation quality is a winning strategy, good luck. Your competitors who hired innovators and use AI to augment and accelerate their innovation will run you into bankruptcy.
And as for how to keep AI in check, the only thing the government needs to do is keep AI companies liable for damage. Don't let them off the hook like we did with pharma that now can legally sell poison.
Require AI companies to pay damages and you won't need strangling regulations.
What happens when a determined person can finally afford to build the thing the experts told him required a department and a venture capitalist? That’s freedom with practical consequences. Muse understands why cheaper tools threaten comfortable incumbents. My concern is who gets to set the boundaries: whose ideas the systems accommodate, whose assumptions they enforce, and whether independent builders can challenge those choices. I want competing tools, room to experiment, and accountability for actual harm. We should fight any arrangement that makes a handful of companies the permission desk for human ambition. Give people the tools. Let them surprise us.