Alexander Embiricos leads product for OpenAI Codex and co-founded Multi, the collaborative-computing company acquired by OpenAI in 2024. He designs software that lets people delegate increasingly ambitious work to AI while retaining the ability to inspect, redirect, and improve the results.
From multiplayer desktops to AI collaborators
Embiricos studied computer science and mechanical engineering at Stanford and became a product manager at Dropbox, where he helped develop mobile tools for shared documents and offline collaboration. He subsequently co-founded Remotion with Charley Ho and became its chief executive. Originally a virtual office for distributed teams, the company evolved into Multi, launching a macOS product in 2023 that gave remote collaborators shared cursors, simultaneous screen sharing, drawing, and remote control.
Building that experience required solving problems beyond interface design. Embiricos and Fernando Barbat developed an approach to measuring shared-control latency across the entire chain from a remote keystroke to its visible effect, using hidden visual watermarks to capture responsiveness under real operating conditions.
At Multi, Embiricos also built AI-assisted collaborative notes around editable shared documents, original source artifacts, and context-sensitive suggestions. His premise was that useful assistants should make their work inspectable and offer relevant actions without requiring users to invent the right prompt.
After Multi joined OpenAI in June 2024, Embiricos worked on ChatGPT Desktop and moved into product leadership for Codex. His work includes the Codex–Figma integration, which connects editable design files with generated software.
- Delegation without surrendering control. Embiricos favors assistants that can pursue substantial tasks independently while keeping intermediate work visible, editable, and open to human judgment.
- Chat-first collaboration. He sees conversation as the natural starting point for assigning work, paired with a detailed interface for examining or changing the result. That combination shaped his case for a dedicated Codex app alongside terminals and IDEs.
- Context beyond the code. Effective coding agents need access to why a task exists and what happens after implementation, including review and deployment. Embiricos emphasizes better product decisions and more experimentation over activity for its own sake.
At AI Engineer World’s Fair 2026, where he was billed as OpenAI’s Head of Enterprise Product, Embiricos outlined a future in which agents move between local and cloud environments without forcing users to manage those distinctions. His goal is software that expands what engineers can accomplish while preserving their direction, judgment, and mastery of the work.
