Stripe CEO Patrick Collison says a friend used prompts to generate a custom OS compatible with the Linux ABI
Stripe CEO Patrick Collison wrote on X on November 12 that a friend recently used prompts to generate an operating system compatible with the Linux ABI for his cloud applications. He questioned why people would continue using off-the-shelf libc, device drivers, or the TCP stack, and said they bring bloat and an attack surface.
Collison said that this PromptOS is faster than standard Linux in production, and speculated that everyone will soon do the same. The statement is a personal observation and prediction, and did not provide code, benchmark data, or independent verification results.
Original source:
Title: A friend recently prompted his own Linux ABI-compatible OS for his cloud applications...
Publisher: Patrick Collison (@patrickc)
Publication date: 2026-11-12
URL: x.com/patrickc/status/2120907937627684940
Source type: Personal social media post
Verification status: The full original post was actually read for this report; verification date: 2026-11-12
Supporting content: Supports Collison’s full original wording, including the expressions “prompted his own Linux ABI-compatible OS”, “PromptOS is also much faster than stock Linux in production” and “everyone will be doing this very soon”
ABAB AI Insight
This post describes the practice of using large-model prompts to generate custom operating system components, rather than a public product launch or an update to a standard distribution. ABI compatibility means it still relies on the Linux system call interface, so it is not completely detached from the existing ecosystem.
The claim of being “faster in production” and having a “smaller attack surface” is based on observations from a single party. There is currently no public benchmark or security audit to support it. A custom stack reduces generic components while shifting maintenance and patch responsibilities to the generator or user.
Historically, minimal runtimes such as unikernels have embodied a similar idea, but debugging and compatibility costs have limited large-scale adoption. AI lowers the barrier to writing, but it does not automatically solve runtime observability or long-term maintenance issues.
ABAB News · Cognitive Law
After AI lowers the cost of generating system software, the main constraints on custom stacks shift from writing difficulty to debugging, compatibility, and long-term maintenance responsibility.
Without independent benchmarks and audits, a single production observation of being “faster and safer” does not amount to a general conclusion that it can replace the standard stack.