Browse · People
Person · Latent Space wiki

Sam Julien

Product leader and developer educator focused on AI platforms, agent interfaces, and the boundary between automated synthesis and human judgment.

Sam Julien is Director of Product at CopilotKit, which builds open-source infrastructure connecting AI agents to user-facing applications. Previously a product manager at WRITER, he worked on developer experience and its AI platform. He is also coauthor, with Anthony Alcaraz, of Agentic GraphRAG, listed by O’Reilly with an August 2026 date.[2][3][5]

Choosing the agent interface layer

Julien announced his move to CopilotKit in “Why I’m Betting on Generative UI”, marked April 2026. Looking back on two years at WRITER, he described building its first developer go-to-market effort and developer experience before product-managing its AI platform. His work included the public API and LLM Gateway migration, SDK integrations with LangChain and AWS Strands, observability, and enterprise AI governance.[3]

His reason for choosing CopilotKit centered on giving people control over agent work: interfaces should let users review, approve, edit, and redirect an agent’s reasoning in real time. He called that interface layer “the most important unsolved problem in the agentic stack.” His stated remit was to make developer experience a first-class product and connect open-source adoption to the enterprise business.[3]

Julien distinguishes three forms of generative UI: static interfaces, where an agent selects components registered by the developer; declarative interfaces, where it emits structured descriptions of forms, tables, or flows for native rendering; and open-ended interfaces, where it generates HTML, CSS, and JavaScript inside a sandboxed iframe. His enterprise example is a procurement manager reviewing an agent’s supplier-risk analysis through a generated dashboard with relevant filters and drill-downs.[3]

He argues that better models increase the importance of governing consequential decisions: “The need for HITL doesn’t shrink as models improve.” He also sees approvals, rejections, and edits made in real workflows as valuable task-specific training signals, making the interface both a place for human oversight and a potential source of feedback for improving agents.[3]

Agents for synthesis, people for judgment

In his March 2026 account, “How I Automated the Worst Part of Product Management”, Julien described managing thirteen projects across three teams and building a workflow around Obsidian, Git, and AI agents. A WRITER Agent playbook collected Google Meet transcripts from calendar events and produced Markdown summaries with the full transcripts attached. Julien downloaded these into his vault and invoked workflows that routed notes, incorporated Jira activity, and drafted updates. Voice instructions helped him conserve his hands after RSI problems.[4]

Two design choices preserved useful context. When a meeting already had a WRITER Agent summary, the workflow appended its Google Meet transcript rather than creating a duplicate record. For weekly reports, it examined actual Git file diffs alongside meeting records and Jira changes, then compared the results with the previous week’s report to distinguish new work, progress, and stalled items.[4]

Julien recounts an agent identifying a shared engineering-bandwidth dependency between infrastructure and API-key migrations, allowing him to adjust their sequencing. He still edited reports for organizational context, unrecorded conversations, and the appropriate tone for escalating risks. Stakeholder conversations, prioritization, political judgment, and customer empathy remained his responsibility: “The agent handles information. I handle judgment.”[4]

Earlier foundations: frontend engineering and developer education

Before his AI-platform work, Julien developed C#/.NET and JavaScript applications at a renewable-energy nonprofit, designed APIs, and led an AngularJS-to-Angular migration. In his career retrospective, he describes running Angular Portland for five years, creating a 200-lesson migration course and courses for egghead.io, then moving into developer advocacy at Auth0. That frontend background supplies context for his interest in keeping agent state and user interfaces connected, and in bringing agent capabilities to Angular developers.[3]

At AngularConnect in London on November 6–7, 2018, Julien presented “How to Save Time and Money by Planning Your ngUpgrade.” His Auth0 account describes teaching four migration strategies through two fictional companies based on his experience and students’ questions. He also joined Elana Olson and Erin Coughlan for an upgrade panel and office hours.[6]

Timeline

  1. project

    Shares an agent workflow for product-management synthesis
    In his March 2026 account, “How I Automated the Worst Part of Product Management”, Julien described managing thirteen projects across three teams and building a workflow around Obsidian, Git, and AI agents.[4]

  2. role

    Announces move to CopilotKit to work on agent interfaces
    Julien announced his move to CopilotKit in “Why I’m Betting on Generative UI”, marked April 2026.[3]

  3. milestone

    O’Reilly lists Agentic GraphRAG, coauthored with Anthony Alcaraz
    He is also coauthor, with Anthony Alcaraz, of Agentic GraphRAG, listed by O’Reilly with an August 2026 date.[5]

Sources

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
    Agentic GraphRAG [Book]Accessed 2026-09-20
  6. 6

Model source and teaching reviews passed. Human editorial review pending. This profile describes the sources as of 2026-09-20.

Latent Space

Join Latent Space.

Sign in with Google, or get the newsletter by email.

Loading Google sign-in…

or subscribe by email

Free. Unsubscribe any time.