Kojo Osei is a venture capital investor at Matrix Partners who coined the concept of "agent-driven commerce"—where AI agents become the primary buyers for organizations instead of humans—and developed the framework of "context gravity" as the new way companies build defensibility in the AI era. His concrete mental models for AI infrastructure have become reference points for other investors and operators, and he's known for bringing a global perspective that connects AI systems to real-world workflows beyond just Silicon Valley use cases.

• Deeply interested in the evolution of AI coding tools and how developers interact with them. A specific example is an appreciation for thoughtful design in utilitarian spaces, like noting the poetic loading states within the terminal interface of Claude Code and how artifacts have become a staple of AI interaction.
• Fascinated by the shift toward asynchronous and parallel AI workflows rather than linear chat interfaces. A great reference point is the excitement around the agent inbox paradigm, specifically pointing out how tools featuring isolated infrastructure per agent and parallel tasks represent the future of productivity.
• Focused on new distribution models for developer tools, particularly the idea that agent recommendations will replace traditional product led growth. An example to bring up is the belief that getting an AI coding agent to automatically suggest a startups tech stack to developers provides a massive competitive advantage.
• Believes that excellent code is merely the baseline for software companies and places a high value on go to market execution. This is highlighted by the observation that open source software alternatives often struggle because they underestimate the critical importance of proper packaging and distribution.
• Attentive to the technical nuances of how builders interact with AI, suggesting that prompts should be treated with the exact same rigor as traditional software. An interesting angle to discuss during a pitch is the idea of context engineering becoming the next big software category, along with the importance of writing robust prompts today so future models can execute them even more effectively.