By playing, you agree to uListen's Terms and Privacy Policy and YouTube's Terms of Service.
a16zWhy AI Has DHH Working 16-Hour Days
Summaries, chapters, transcripts, and FAQs on this page are made by uListen, not YouTube.
Great conversations run long. Their best ideas don't have to.
uListen is free on iPhone, iPad, Android, Chrome, and Claude.
See howCHAPTERS
Jump to publisher timestampsChapter summaries
Linux as the “OS for the age of agents” and why macOS/Windows don’t fit
David Heinemeier Hansson argues that macOS and Windows are poorly suited to an agent-driven workflow, while Linux’s historically “hard” traits are exactly what agents thrive on. The hosts react to the new feeling of being able to ask an agent to modify the OS and have it work.
- •David Heinemeier Hansson says macOS and Windows are “unfit for the age of agents.”
- •Erik Torenberg describes telling an agent to modify the OS and being surprised it works.
- •David Heinemeier Hansson claims Linux’s terminal-centric workflow, config files, and detailed errors are strengths for agents.
- •He frames Apple’s warnings about agent access as counterproductive for agent-based use.
DHH’s background: Basecamp, Rails, and shifting focus to a Linux distribution
Martin Casado asks for a quick professional overview, and David Heinemeier Hansson recounts decades building software, including Basecamp and Ruby on Rails. He describes his current work: creating Omachi as a native OS aimed at agent-driven computing.
- •David Heinemeier Hansson summarizes his history building Basecamp and extracting Ruby on Rails from it.
- •He mentions Rails being used for major apps and that Basecamp and Rails have continued to evolve.
- •He notes launching Hey.com and writing several books.
- •He introduces Omachi as “the first native OS for the age of agents.”
From Mac devotee to Linux desktop: why he switched and what changed
David Heinemeier Hansson explains he used Linux on servers since the ’90s but only moved to Linux on the desktop recently. He describes leaving the Mac ecosystem after growing frustrated with Apple’s approach to developers and platform economics.
- •He says his first Linux desktop distro was Ubuntu, about two and a half years ago.
- •He describes being a long-time macOS user starting with early OS X.
- •He criticizes Apple’s business incentives around taking 30% of others’ businesses.
- •He says developer relationships and increasing restrictions pushed him to switch.
Building Omachi with agents: a malleable OS users can reshape instantly
The conversation turns to how agents change what an OS can be: David Heinemeier Hansson says Omachi is being developed “all in” with agents and designed to let users modify core UI and behavior via prompts. He contrasts this with macOS’s settings workflows that don’t translate well to agentic control.
- •He says he hasn’t written Omachi code by hand for months, but has produced large amounts through agents.
- •He emphasizes rapid iteration and giving users the ability to change OS elements by prompting.
- •He gives an example of macOS hotkey remapping being trapped inside manual settings clicks.
- •He argues Linux’s CLI and config approach is naturally “agent-friendly.”
A window for a third OS—and what that implies for mobile platforms
David Heinemeier Hansson claims the agent era reopens a long-closed opportunity for new operating systems to gain adoption. He extends the logic to mobile, arguing a new platform becomes feasible when app availability and development cost stop being the barrier.
- •He says the first real opening for a new desktop OS in a long time is happening now.
- •He describes mobile as a long-standing duopoly tied to a 30% platform cut.
- •He argues past mobile challengers failed largely due to missing key apps.
- •He suggests agents reduce or eliminate the application-availability lock-in.
Apps rewritten fast, apps made irrelevant: why agents reset ecosystem lock-in
To illustrate the shift, David Heinemeier Hansson recounts an example from the Omachi community where someone rebuilt an app he relied on. He then argues many traditional applications may be bypassed altogether when agents can produce documents, edit images, and perform tasks without specialized software.
- •He says someone in the Omachi community replicated Adobe Lightroom’s functionality quickly and even improved it.
- •He argues ecosystem lock-in has historically depended on application availability.
- •He describes using agents to draft contracts, convert to PDFs, and apply signatures without a word processor.
- •He claims agents can handle tasks that once required learning complex tools like Photoshop, reducing the need for entire app categories.
Live-mutating software and token-speed acceleration: a preview of what’s next
David Heinemeier Hansson describes a future where software changes in real time as you prompt it, tied to faster generation speeds. He references a demonstration he saw as a turning point in imagining how quickly functional software can be created and iterated.
- •He contrasts today’s typical token speeds with faster modes that still don’t feel fully “live.”
- •He describes a demo where games were generated and modified rapidly through prompts.
- •He says the key shift is that “everything is just gonna mutate live.”
- •He speculates that higher token speeds will further collapse time-to-build and iteration cycles.
No moat in source code, compressed build cycles, and shifting economics of software
The discussion broadens to competition and how fast small teams can now build systems that compete with giants. David Heinemeier Hansson argues that the traditional defensibility of large codebases erodes when agents compress development time and reduce coordination costs.
- •He describes Omachi’s rapid development with a small team competing with far larger organizations.
- •He predicts development timelines compress from years to days, weeks, or months.
- •He says model capability jumps over short periods have been significant during Omachi’s development.
- •He argues anyone believing source code alone is a defensible moat is “delusional.”
Taste, craftsmanship, and the 16-hour days: humans steering agents to perfection
Martin Casado challenges the idea that code isn’t a moat by pointing to design nuance and long-tail polish. David Heinemeier Hansson agrees taste matters and says agents amplify his ability to iterate toward perfection—so much so that he’s working extremely long days.
- •He says agents collapse the barrier between noticing a flaw and fixing it.
- •He describes being energized by polishing the final small percentage of quality rather than the initial build.
- •He explains why many “vibe coding” projects stall before the last, difficult polish phase.
- •He contrasts managing thousands of humans with directing a “fleet of agents” without coordination overhead.
Superintelligence: software feels there already, while robotics may lag
Martin Casado asks about a personal test for superintelligence and shares his own robotics-based benchmark. David Heinemeier Hansson says he doesn’t have a single test, but believes the breadth and speed of AI already looks like superintelligence for software development, while physical-world capability may take longer.
- •Martin Casado describes repeatedly asking models to order parts to build a robot as an “ASI test.”
- •David Heinemeier Hansson says for software development, the AI already surpasses any programmer he’s known in breadth and pace.
- •He notes narrow areas where top specialists can still outperform agents.
- •He contrasts Silicon Valley timelines with perspectives from people working on actuators and mechanics.
Frontier labs vs. open weights: avoiding duopolies and keeping the ecosystem honest
Asked directly about frontier labs versus open source, David Heinemeier Hansson argues open weights are essential for competition and freedom to build without permission. He warns against government-enforced consolidation and compares platform gatekeeping to earlier eras of permissioned software distribution.
- •He calls open source and open weights essential to “keep the whole game honest.”
- •He contrasts an open web era with mobile platform permissioning and economic gatekeeping.
- •He says a duopoly (especially government-backed) would be a “nightmare scenario.”
- •He welcomes frontier advances while insisting competition prevents price and power abuse.
AI culture, “battle of the vibes,” and who he follows for signal
The discussion turns to social dynamics around AI adoption, including fear, tribalism, and public discourse. David Heinemeier Hansson describes a divide in attitudes toward AI, names people he respects who are “leaned in,” and points to the resurgence of pseudo-anonymous hacker culture as a key force.
- •He says respected programmers he knows are fully committed to using these tools.
- •He expresses sympathy for people anxious about professional disruption while insisting the technology won’t be abandoned.
- •He discusses a “battle of the vibes” around AI and the social dynamics he sees driving positions.
- •He cites respected builders and highlights pseudo-anonymous hackers as a source of innovation and momentum.
Anonymous hackers, X as connective tissue, and the return of ‘90s demo culture aesthetics
David Heinemeier Hansson celebrates the comeback of pseudo-anonymous creators and connects it to renewed individual impact in computing. The conversation ends with nostalgia for demo/BBS culture and how Omachi intentionally embraces a ‘90s terminal-and-ASCII aesthetic to bring back “computers with character.”
- •He says his team includes pseudo-anonymous contributors and frames it as a throwback to earlier hacker culture.
- •He argues AI re-democratizes consequential software work that had become dominated by large companies.
- •He calls the vibe “anti-LinkedIn culture” and credits X with connecting communities.
- •He describes Omachi’s ASCII-art screensaver, demo effects, and desire for computers to feel “mysterious” and less polished.
Publisher timestamps
- 0:00Intro
- 0:48From Rails and Basecamp to Linux
- 3:58Why AI agents make Linux the better OSRead this chapter
- 8:17Can a third mobile platform happen?Read this chapter
- 16:02Does taste still matter when AI writes code?Read this chapter
- 22:57What counts as superintelligence?
- 25:58Big AI labs vs. open sourceRead this chapter
- 31:50DHH’s plan to be bigger than macOS
- 38:24Who to follow, and the AI culture war
- 44:35The return of the anonymous hacker
- 47:3390s demo culture and computers with characterRead this chapter