Tom Moor is head of engineering at Linear and founder of Outline, a publicly developed knowledge base for teams. He builds the infrastructure that helps people share institutional knowledge, coordinate product development, and work alongside accountable AI agents.
From early startups to shared knowledge
An early engineer at Buffer, Moor worked on its platform, API, and web application, including real-time analytics infrastructure for monitoring API activity and product usage. He subsequently co-founded Sqwiggle, a communication product for distributed teams, worked on the remote-collaboration tool Speak, and became a principal engineer at the design-collaboration company Abstract.
Building team products sharpened his attention to cross-platform compatibility, reliable real-time communication, and the challenge of persuading entire groups to adopt new software. His writing on collaborative products treats product polish, dependable infrastructure, and flexible preferences as practical requirements for organization-wide adoption.
Moor founded Outline to give teams a shared home for documentation, specifications, onboarding materials, and other institutional knowledge. He develops the product publicly because inspectable code can increase customer confidence, encourage contributions, and improve engineering discipline. Outline’s source-available model permits inspection and self-hosting while restricting competing hosted services. By September 2023, its repository had reached 20,000 GitHub stars.
Building useful AI into team workflows
Embedded AI that solves existing problems. At Linear, Moor helped develop similar-issue detection, natural-language filters, and automatic conversion of Slack conversations into structured issues. His explanation of duplicate detection connects embeddings to a concrete organizational problem: engineers unknowingly duplicating work across crowded issue backlogs.
Pragmatic search infrastructure.Linear initially combined OpenAI embeddings, pgvector, and Google Cloud, favoring established components while testing early AI features. It subsequently adopted hybrid retrieval with Turbopuffer and Cohere embeddings, combining keyword and vector search with query rewriting, reranking, and deterministic rules. That foundation supports suggested labels and assignees, duplicate identification, customer-feedback analysis, and richer relationships between issues and projects.
Agents as accountable teammates. Moor’s approach to agent coordination gives agents workspace identities, activity histories, audit trails, OAuth installation, scoped permissions, GraphQL access, and event-specific webhooks. Integrations can investigate bugs, draft pull requests, manage feature flags, or coordinate customer-support responses inside existing team workflows.
Legible behavior before autonomous action. Moor expects agents to acknowledge requests quickly, communicate a concrete plan, clarify uncertainty, update issue status, and explain progress concisely. Their usefulness depends on behaving predictably within the same working conventions that human teammates follow.
