Interview with Conor White-Sullivan, Founder of Roam

Roam Research’s founder, Conor White-Sullivan, frames Roam as a knowledge system built around how ideas actually connect in a person’s mind—using backlinks, tags, and overlapping collections—rather than around the file-and-folder logic that dominates most note-taking tools. The core motivation traces back to a long-running problem with online discourse and learning: arguments often get lost in ad hominem, weak structure, and hard-to-audit summaries. Roam aims to make intellectual work more “traceable,” so claims can be tied back to their sources and stitched into coherent syntheses over time.

White-Sullivan’s path to that approach runs through earlier work on crowd-sourced debate and user-generated content at AOL, where he concluded that the web’s page-centric structure is the wrong organizing unit for knowledge. Around 2013, he shifted toward building a personal tool first—one that could map multiple authors’ ideas together, preserve the relationships between concepts, and let users trace summaries back to the original material. He also describes trying to do similar mapping with tools like Workflowy and Evernote, only to hit a wall: file-based systems force arbitrary decisions about where each note “belongs,” encourage duplication, and make cross-document relationships hard to track.

A major design principle in Roam is supporting the “big picture” while the structure is still unknown. Instead of relying on linear presentations or graph visualizations as the primary onboarding path, Roam emphasizes collecting ideas into overlapping “buckets” and intermediate packets—groupings that can later be assembled into a final deliverable like a blog post, tweet storm, or video. This matches how people often write: first they gather and cluster related fragments, then they discover the thesis structure as connections emerge.

White-Sullivan also distinguishes Roam from specialized productivity tools. Users may apply Roam to task management, spaced repetition, or class notes, but the underlying abstraction is general-purpose knowledge management: bullets can be embedded in multiple places, linked bidirectionally, and organized through tags and expandable/collapsible structures. He argues that Roam’s category shift matters because most competing systems still assume hierarchical storage—forcing users to “burn” or archive notes to maintain coherence—whereas Roam treats older notes as continuously reusable building blocks.

On privacy and trust, Roam cites standard encryption practices (encrypted at rest and in transit) and a privacy posture that emphasizes user ownership of data, with no advertising or data selling. End-to-end encryption isn’t fully implemented yet, largely because password loss would permanently lock users out; the company is exploring ways to allow selective, stronger protection for highly sensitive notes.

Finally, White-Sullivan positions Roam as a utility for millions, not a niche tool for productivity obsessives. He predicts that switching costs will fall as Roam improves importers and exports (including a planned JSON export) and as an API enables others to build on Roam’s underlying graph data. The long-term vision extends beyond personal notes toward publishing and remixing knowledge graphs—potentially replacing Twitter as the place where evergreen intellectual threads live, and enabling new educational formats and even detailed argument maps for fringe or conspiratorial claims.