#501 – DHH: Future of Programming, AI, Agentic Engineering, Vibe Coding & Linux
with David Heinemeier Hansson
27 Aug 20265 min read45m 44s
TL;DR
DHH marks November 24, 2025 — the release of Claude Opus 4.5 — as the exact moment AI agents crossed from useful assistant to genuine collaborator, delivering code quality indistinguishable from expert human output. He built Omarchy Quattro, an entire Linux distribution, without writing a single line of code by hand, while still reviewing architecture. The real bottleneck to AI-accelerated development, he argues, is not coding ability but human organizational bureaucracy, lack of vision, and the inability to work directly with agents without managerial intermediaries.
Key Moments
DHH
“I have not written any of the code that's shipped in Quatro by hand. I've reviewed the shape of all of it.”
DHH describing his experience building the Omarchy Quattro Linux distribution entirely through AI agents over three months.
“To get that magical 10X, 100X, in a few rare cases, 1000X productivity boost, you have to interact with the agents directly, and you cannot intermediate that bandwidth with another human because it's simply too slow.”
DHH explaining why large companies with many approval layers cannot replicate the productivity gains he sees working solo with agents on Omarchy.
“Most organizations don't know what they want. They don't know how to make it better. They're not bottlenecked on implementation. They're bottlenecked on ideas. They're bottlenecked on vision. They're bottlenecked on taste.”
DHH explaining why the AI coding revolution hasn't visibly accelerated software at large established companies like Adobe.
“I would rather get an agent-written pull request to one of my projects than I'd get one written by a human, and it's not just because the quality's better. It's also because I feel a lot less bad if I just reject it.”
DHH arguing that AI-generated open source contributions are superior to most human ones, both in quality and in the social dynamics of rejection.
“I don't think that's true anymore at all. I have seen things you people wouldn't believe, ideas coming out of models so great that it makes me humble as a person who otherwise prides himself on having good ideas.”
DHH rejecting his earlier belief that all good ideas must originate with humans, with agents only implementing them.
David Heinemeier Hansson (DHH) is the creator of Ruby on Rails, CTO of 37signals, and the creator of Omarchy, a Linux-based operating system. He is a best-selling author and professional race car driver who has been an influential voice in programming and software development for over 25 years. Since late 2025, DHH has publicly transformed his approach to programming, fully embracing agentic AI engineering and becoming one of its most outspoken advocates.
Takeaways
1
You need programming intuition to vibe code safely When 37signals let non-programmer designers vibe code on Basecamp, individual PRs looked fine but collectively destroyed the architecture. DHH concludes that vibe coding on existing substantial codebases requires someone with architectural intuition to govern the overall shape, even if they write zero lines themselves. Starting fresh is safer for non-programmers than modifying a legacy system.
2
One person can now build Premiere or Photoshop equivalents DHH argues that the combination of agents and a clear personal use case makes it completely feasible for a single developer to rebuild major proprietary applications like Adobe Premiere for Linux. He built a full Markdown editor in C++ — a language he'd never used — in 20 minutes, then abandoned the commercial alternative within two days. The 'I only need 5%' pattern means personal versions can be better than the original for any given user.
3
November 2025 was AI's Wright Brothers moment DHH identifies Claude Opus 4.5 on November 24, 2025 as the precise inflection point when AI agents went from useful sidekick to genuine collaborator producing expert-level code. He draws a direct parallel to the Wright Brothers' first flight — obvious in retrospect, but world-changing in the moment. Anyone who hasn't updated their priors since then is operating on outdated information.
4
Bureaucracy, not AI, is the real bottleneck DHH's core thesis is that the 10X–100X productivity gains from agentic AI are only accessible when a single person works directly with agents — every additional layer of human approval degrades the gain. Large companies with PMs, VPs, and multi-stage review processes will see modest acceleration at best, not transformation. The real unlock is organizational, not technological.
5
Agents now generate better open-source PRs than most humans After reviewing tens of thousands of human pull requests over 25 years, DHH asserts that the median human contributor produces lower-quality PRs than a well-instructed agent — missing context, skipping tests, omitting comments. On Omarchy, he now has agents triage the 400+ open PRs for him, surfacing only the decisions worth his attention. The social friction of rejecting a human PR is also eliminated.
6
AI agents are now originating ideas, not just implementing them DHH's earlier model held that humans supply creative direction and agents execute — he now says this is obsolete. Working on Omarchy Quattro, he repeatedly encountered agent-generated ideas he would not have thought of himself, which he describes as humbling. He calls anyone still describing agents as 'parrots just regurgitating existing ideas' delusional about progress made in the last six to nine months.