O’Reilly, Komoroske Outline 5 Principles for Human-Centric AI as Common Fabric Stays Unreleased
Updated
Updated · O'Reilly Media · Nov 4
O’Reilly, Komoroske Outline 5 Principles for Human-Centric AI as Common Fabric Stays Unreleased
1 articles · Updated · O'Reilly Media · Nov 4
Summary
An hour-long October 20 conversation centered on the Resonant Computing Manifesto, which Alex Komoroske said sets five principles for AI systems: private, dedicated, plural, adaptable and prosocial.
Common Fabric’s answer is infrastructure where each piece of data carries cryptographic provenance and policy rules, enforced by an open-source runtime that can block unsafe computation before it runs.
Komoroske argued that model lets untrusted, TypeScript-based “patterns” use sensitive data without exposing it to their authors, while trusted execution environments and taint-checking keep confidential information from leaking.
Tim O’Reilly said the still-unreleased consumer product left him unusually optimistic, framing AI as a chance to reset software away from engagement-maximizing silos and toward user-controlled data and collaboration.
The broader pitch is economic as well as technical: users or AI service providers would pay for compute directly, while open infrastructure and AI-written software could support new business models beyond surveillance capitalism.
If privacy-preserving AI requires users to pay for their own compute, will this create a digital divide where only the wealthy afford true privacy?
Can AI agents rapidly rewrite code without developers seeing the data, or will strict metadata policies accidentally break complex software ecosystems?
Will relying on trusted organizations for data policy defaults actually decentralize the web, or simply shift power to a new group of tech gatekeepers?