DHH and the Expanding Developer Toolbox

The expanding developer toolbox

DHH is excited about AI. He spent the last three months using it to perfect Omarchy, his new Linux distribution, including remarkably low-level features without learning any low-level languages. The web developer is now also an operating system designer and a desktop application product manager.

In his keynote, DHH announced that Hey is being re-imagined as a collection of six native applications plus a mail server written in Rust. In short, he's continuing to use a wider array of tools.

The web is amazing at low-friction / occasional interactions, but Hey is high-friction / high-frequency. After convincing new users to change how they use email, having them install an app is a small ask.

On the other hand, Basecamp's users are mixed: some may use the app every hour of the workday, others just once or twice in their lifetime. If Basecamp required installing an app, it would lose a meaningful number of project collaborators, and its value to the project would decrease.

DHH was telling Rails developers that not every problem requires a web application solution. AI allows developers to choose the tool that best fits the problem without balancing it against yesterday's learning curve.

The Rails "opportunity"

At the same time, non-developers are learning to be makers too, and many of the problems they want to solve are exactly the kind that benefit from the web's low-friction distribution.

The real problem for Rails isn't Rails programmers branching out into non-web technologies. It's missing this new market of makers because AI chooses Next.js for them. Rails has overcome shifts in attention before, in no small part because passionate advocates like DHH kept making the case for a better way to build web applications.

AI doesn't make the case for Rails irrelevant. Reading business logic as code builds trust in the system, and Ruby on Rails still produces exceptionally readable code. Rails' advantage isn't that AI needs readable code. It's that humans still build trust by reading it.

The Open Question

The open question is how long trust relies on reading code. Today, the code is still the authoritative description of what the software actually does. But if AI can eventually give humans a simpler representation of their software that is guaranteed to match its implementation, readable code and Rails may become tools humans can afford to leave behind.