DHH Declares End of Hand-Coded Software

The era of writing code by hand for professional software development is over.

What changed

• David Heinemeier Hansson (DHH), creator of Ruby on Rails, stated that 37signals will no longer write code by hand for professional work.
• This declaration comes amidst ongoing debates about the impact of AI coding assistants on the software development lifecycle.
• The company is embracing AI tools to augment and, in some cases, replace manual coding efforts.

Why it matters

This is a significant statement from a prominent figure in the software development community, particularly within the Ruby on Rails ecosystem. The honest version: DHH is signaling a strategic shift for his company, 37signals, towards heavily integrating AI coding assistants. This isn't just about using AI to write boilerplate; it's a declaration that the core act of manual code construction is becoming secondary. For organizations operating at scale, this could represent a paradigm shift in how development teams are structured and how productivity is measured, potentially leading to faster iteration cycles but also raising questions about code quality and maintainability.

The catch

The catch: This pronouncement is specific to 37signals' context and their current scale. While they operate at a significant size, they are not necessarily in the same league as hyperscalers like Meta or Google, who have different constraints and existing infrastructure. What this replaces: The traditional model of senior engineers meticulously crafting every line of code. Pairs with: GitHub Copilot, or any other AI code generation service. The trade-off here is clear: potential gains in speed and reduced manual effort versus the risk of increased technical debt, reliance on opaque AI models, and a potential decline in deep, first-principles understanding among engineers.

Ship it

For teams already heavily invested in AI coding assistants and facing pressure to accelerate delivery timelines, this is a signal to double down on integrating these tools. In practice: Evaluate current AI assistant usage for tasks beyond simple autocompletion. Consider pilot programs for AI-assisted refactoring or even initial code generation for well-defined components. The key is to do this with rigorous testing and code review processes in place.

Bottom line: Expect more companies to follow 37signals' lead in embracing AI for core development tasks, but proceed with caution regarding code quality and engineer skill development.

— Filed to /engineering