← back
Creating Agents that Co-Create — Karina Nguyen, OpenAI
Original: Creating Agents that Co-Create — Karina Nguyen, OpenAI
Takeaway
Each scaling paradigm shift opens new product UX patterns; the next frontier is agents that actively co-create with humans rather than just respond.
Summary
- OpenAI researcher Karina Nguyen traces two scaling paradigms — next-token pre-training (world building) and RL/reasoning (test-time compute) — and how each unlocked product surfaces in Claude and ChatGPT.
- Argues creative writing is one of the hardest tasks because there's no clean reward signal, making it a frontier for collaboration-style agents.
- Shares design lessons from shipping Claude and ChatGPT: how interfaces evolve from chat to canvases and co-creation tools as model capability grows.
- Frames the trajectory from agents-as-assistants to agents-as-co-innovators that propose and execute on long-horizon work.
agentsopenaiproduct-ux
Original description
In this talk, I will share my thoughts on how two major scaling paradigms in AI are transforming frontier product research——shifting our perspective of AI from narrow, task-focused tools to collaborative agents that meaningfully contribute to real-world innovation. Drawing on lessons learned from my contributions to ChatGPT and Claude, I will explore how AI increasingly handles tasks once reserved for skilled builders, fundamentally altering the fabric of the Internet and the nature of learning. Where, then, are we headed as creators, and what should we be building—and why—if AI increasingly does the work once left to us? Recorded live at the Agent Engineering Session Day from the AI Engineer Summit 2025 in New York. Learn more at https://ai.engineer and purchase tickets to our next event, the AI Engineer World's Fair, in SF June 3 - 5 here: https://ti.to/software-3/ai-engineer-worlds-fair-2025