The Craft of Software and the AI Revolution: A Conversation with DHH

David Heinemeier Hansson, famously known as DHH, the creator of Ruby on Rails and co-founder of 37signals, has undergone a notable shift in his perspective on Artificial Intelligence. While previously expressing skepticism, DHH has recently embraced an "AI-first" approach to software development. This article delves into how DHH and his team at 37signals are now building software, the impact of AI agents on their ambition, why Ruby on Rails and Omachi are well-suited for AI workflows, the increasing importance of taste and beautiful software, and why skilled designers and engineers who value craft are poised to become even more in demand.

Omachi and Ruby on Rails: Building from the Ground Up

DHH's journey in building software spans three decades, beginning in 1994. His recent endeavors include the creation of Omachi, a new Linux distribution. This project, born from a summer project between endurance races, highlights DHH's philosophy of "scratching his own itch." He found that when he builds something that perfectly suits his needs, thousands of others with similar preferences discover and appreciate it. This has been a recurring theme throughout his career, from Ruby on Rails and Kamal to his recent push to move away from cloud infrastructure.

Ruby on Rails, DHH's most famous creation, emerged from a similar need. In the early 2000s, while building Basecamp, DHH chose Ruby for its flexibility. However, the lack of robust web application tooling for Ruby at the time necessitated building his own components, which eventually evolved into Ruby on Rails. Today, Rails is experiencing a renaissance, partly due to its "token efficiency" and suitability for AI agent workflows. DHH notes that the ability for agents to produce human-readable and verifiable code remains crucial, at least for now.

The success of Omachi, with hundreds of contributors and tens of thousands of users in just over six months, mirrors the impact of Ruby on Rails. DHH emphasizes that while many Linux distributions exist, his unique spin on it has resonated with a large community. He sees his role as flattening the curve for new Linux users, making it easier to adopt and enjoy the system.

37signals: A Legacy of Innovation

Founded in 1999 as a web design firm, 37signals transitioned into a software company in 2004 with the launch of Basecamp. This product, DHH believes, was their best idea objectively, and it has sustained the company for over two decades. Basecamp remains their flagship product, a testament to the power of a strong initial concept.

More recently, 37signals launched HEY, an email service, in 2020. This was a bold move into a market dominated by Gmail. HEY was built on DHH's accumulated opinions about what was lacking in existing email clients. The launch faced significant challenges with Apple's app store policies, leading to a high-stakes battle. Ultimately, HEY was approved, and the ensuing publicity provided a massive, albeit unintentional, marketing campaign, leading to tens of thousands of initial sign-ups. DHH uses HEY daily, finding it a more pleasurable and sacred experience than Gmail, which he feels allows too much intrusion. HEY's core philosophy is to make email lovable again by giving users control over their inbox through features like the "screener," which filters incoming messages before they reach the inbox.

The Craft of Design and Development at 37signals

DHH's approach to building products, including HEY, emphasizes a small, focused team in the initial stages. He believes that pouring many people into an uncertain direction is counterproductive. The process involves a core team, often just one programmer and a couple of designers, to shape the product's architecture and direction. This contrasts with the typical VC-funded startup model where product managers define specs and developers are brought in later.

At 37signals, designers are not merely tasked with making specs look pretty; they are integral to defining what should be built and how it should work. They often wear multiple hats, acting as product managers, intuiting customer needs, and even contributing to the implementation through HTML and CSS. This holistic approach, where designers understand the "fabric of the internet," leads to more cohesive and authentic products. DHH likens this to a jewelry designer understanding the properties of gold or an architect having an engineering understanding of structures.

This philosophy extends to programming as well. DHH argues that the acceleration provided by better tools, like Ruby on Rails and now AI agents, allows smaller teams to build significant products. The industry, he observes, is slowly converging towards this stance, recognizing the efficiency and effectiveness of smaller, more focused teams.

The Aesthetics of Truth: The Craft of Software

DHH holds a profound belief in the intrinsic value of aesthetics in software engineering. He posits that "aesthetics is truth," meaning that when something is beautiful, it is likely to be correct. This principle, he believes, applies across mathematics, physics, and various other domains. The pursuit of beauty, for DHH, is not just about visual appeal but also about creating well-functioning, enjoyable experiences that contribute to human happiness. Conversely, poorly designed and frustrating systems are a significant source of malaise.

He finds Ruby to be a particularly beautiful programming language, capable of producing elegant code without being rigidly ideological. This blend of broad scope and intense focus is rare. DHH advocates for working with beautiful tools and producing beautiful, fluid interactions, viewing this as the hallmark of a true craftsperson who polishes their work until it is perfect.

The AI Inflection Point: From Skepticism to Agent-First Workflows

DHH's initial skepticism towards AI, particularly early autocomplete tools like GitHub Copilot, stemmed from their intrusive and often incorrect nature. He found these tools to be a nuisance rather than an aid. However, the launch of models like GPT-4 and the development of agent harnesses marked a significant shift.

The advent of AI agents, capable of using tools like bash, accessing the internet, and producing high-quality, mergeable code, has been a game-changer. DHH describes this as an "inflection point," particularly with the release of models like Opus 45. This has led him to adopt an "agent-first" workflow, where he initiates projects with AI agents.

This new paradigm allows for unprecedented productivity. DHH recounts processing 100 pull requests in 90 minutes, a task that would have previously taken days. He also highlights the ability to tackle projects that were previously unthinkable, such as optimizing the "floor" performance (P1) of requests, reducing latency from milliseconds to sub-half-milliseconds. This explosion of possibility, he argues, is not just about efficiency but about exploring new frontiers and tackling problems that were previously out of reach.

What AI Means for Developers: A Bifurcation

The rise of AI presents a complex future for software developers. DHH suggests that we may have reached "peak programmer" in terms of the traditional learned skill of coding. While more software will undoubtedly be produced, the demand for developers solely based on their coding ability might decrease.

He foresees a bifurcation:

DHH believes that the most valuable developers in the future will be those who possess strong judgment, taste, and business acumen. They will be able to oversee complex systems, understand customer needs, and guide the development of profitable software, rather than simply being implementers. This aligns with 37signals' emphasis on designers who contribute to product strategy and implementation.

The Future of Craftsmanship in an AI-Dominated World

Despite the rapid advancements in AI, DHH emphasizes that taste, judgment, and craft remain paramount. The goal is not just to produce code, but to produce code that is beautiful, correct, and something to be proud of. AI agents can accelerate the process, but human oversight and a discerning eye are crucial for ensuring quality.

He also highlights the importance of designers who can not only create aesthetically pleasing interfaces but also contribute to product strategy and implementation. This integrated approach, where designers and developers work closely together, is a hallmark of 37signals' success.

The increasing capabilities of AI mean that the cost of exploring new ideas and tackling ambitious projects has plummeted. This allows for a greater focus on innovation and the pursuit of projects that were previously considered too risky or time-consuming. DHH's own work with Omachi and the development of CLIs for 37signals' products exemplifies this new era of accelerated development and exploration.

Key Takeaways