Michael Richman is Bitly’s vice president of engineering for platform, infrastructure, and security, and the founder of Cmd+Ctrl, a system for managing AI coding agents across devices and development environments. He builds tools and organizational practices that make increasingly autonomous software development easier to supervise without chaining engineers to their desks.
From internet infrastructure to coding agents
Richman worked on AIM, MapQuest, and AOL.com at AOL before joining Bitly in 2010 to help build its first paid product. His responsibilities expanded across application engineering, distributed systems, infrastructure, security, reliability, and trust and safety. His account of Bitly’s adoption of Go and projects including git-cleanup reflect a longstanding interest in making production systems and development workflows more manageable.
In 2025, he helped launch Bitly’s AI-assisted engineering adoption initiative for approximately 60 engineers. His account of the rollout describes colleagues sharing real prompts, mistakes, and corrections during over-the-shoulder learning sessions. He emphasizes privacy safeguards, human review of generated code, and assessments that consider accuracy, confidence, cognitive load, and satisfaction instead of relying on commit counts or lines of code.
Richman also co-developed the Bitly MCP server, which connects AI assistants to Bitly’s links, QR codes, and analytics while preserving existing authentication and account permissions. He announced the integration with fellow engineering vice president Raphael Syed.
Building for the limits of autonomy
An agent-agnostic control plane.Cmd+Ctrl uses agent-specific daemons to monitor sessions across local machines and cloud environments, reporting changes to a central service. Its architecture lets developers resume conversations, launch tasks, receive notifications, and organize sessions from a browser or phone. Supported integrations include Claude Code, Codex, Cursor, and Gemini; Richman created the orchestration layer, not those underlying agents.
Fear of missing agent time. Richman uses this phrase for the frustration of discovering that an ostensibly autonomous agent stalled moments after its operator stepped away. Selective notifications and remote access reduce that wasted time without requiring continuous supervision.
Agent choreography. His AI Engineer Europe appearance reframed developer flow as coordinating concurrent sessions: unblocking one agent, redirecting another, and deciding which work merits attention. He treats breaks as essential to handling the cognitive demands of parallel agentic work, not as an interruption to productivity.
