The pivot to an “AI-first” development philosophy at 37signals isn’t a sudden abandonment of craft; it is a recalibration of what it means to be a builder. David Heinemeier Hansson (DHH), once a vocal skeptic of AI’s utility in the software lifecycle, has spent the last year integrating agentic workflows into his daily routine. The result is a shift from manual coding to a high-leverage supervisory model that turns engineering teams into force-multiplied units of one.
The End of the “Autocomplete” Era
DHH’s initial skepticism was rooted in the ergonomics of early AI tools. The “autocomplete” paradigm—where a model constantly interrupts the flow to guess the next character—was a nuisance, not an accelerator. It forced the developer into a reactive state, constantly verifying and dismissing suggestions rather than maintaining the creative momentum of a coherent thought.
The inflection point arrived with the emergence of agent harnesses and frontier models like Claude 3.5 Opus. These tools moved beyond simple text prediction. By providing agents with access to the terminal, the file system, and the internet, the workflow transitioned from “tab-completion” to “delegated execution.” For DHH, this meant he could treat an AI agent as a junior developer who never sleeps: someone capable of handling boilerplate, refactoring, or even complex debugging, provided the supervisor has the taste and technical depth to verify the output.
Agent Acceleration and the Seniority Gap
The most profound impact of this shift is the widening gap between junior and senior engineering roles. In the current landscape, AI agents are only as good as the instructions they receive and the verification they undergo.
Senior engineers are seeing a 5x to 10x productivity boost because they possess the architectural intuition to steer agents effectively and the critical eye to catch subtle errors before they hit production. Conversely, junior developers—who historically learned the craft by wrestling with implementation details—are finding their traditional entry-level tasks being automated. This creates a precarious situation: if a junior developer cannot yet distinguish between “working code” and “correct, maintainable code,” they cannot safely leverage AI to ship to production.
Ambition in the Age of Automation
Perhaps the most compelling outcome of this pivot is the explosion of the “ambition pie.” Small teams at 37signals are now tackling projects that were previously deemed too costly or too low-priority to justify.
Take the “P1” performance project: optimizing the fastest 1% of requests. In a traditional resource-constrained environment, this is a vanity project. With agent acceleration, it becomes a weekend task. By reducing the cost of exploring a hunch to near zero, DHH and his team are no longer limited by the time it takes to write code; they are limited only by the time it takes to identify what is worth building.
The Future of the Craft
The industry is currently obsessed with measuring developer efficiency, but the real metric of the AI era will be the quality of the “why.” As implementation becomes a commodity, the value of a developer shifts toward product management, design intuition, and the ability to maintain a coherent system architecture.
We are likely witnessing the end of “peak programmer”—the era where simply knowing how to write code was a sufficient bottleneck to command a premium. The future belongs to the “product engineer”: the individual who understands the medium of the web, possesses the empathy to solve real user problems, and has the technical judgment to ensure that the code produced by agents is as beautiful and correct as the code written by hand.
The tools have changed, but the fundamental truth remains: aesthetics is truth. Whether the code is written by a human or an agent, if it isn’t beautiful, it isn’t correct. The challenge for the next generation of engineers is not to learn how to type faster, but to learn how to curate, verify, and shape the output of an increasingly automated world.
Sources
- https://www.youtube.com/watch?v=JiWgKRgdgpI
- https://en.wikipedia.org/wiki/David_Heinemeier_Hansson
- https://en.wikipedia.org/wiki/37signals
- https://da.wikipedia.org/wiki/David_Heinemeier_Hansson