Harness Engineering: How to Build Software When Humans Steer, Agents Execute — Ryan Lopopolo, OpenAI

By ai.engineer

Categories: AI, Tools

Summary

Code is now free and infinitely abundant—the scarce resources are human time, attention, and model context. Ryan Lopopolo from OpenAI reveals how banning engineers from touching editors and routing all work through AI agents shifts engineering from implementation to systems thinking, unlocking 5-5,000x concurrent capacity per engineer.

Key Takeaways

1. Shift team culture from code ownership to prompt engineering and guardrails. Document ADRs, persona-oriented guidelines, and historical code review logs—this breadcrumb trail trains agents to replicate your team's decision-making patterns and quality standards.
2. Implement P0/P2 stack ranking: in human-constrained teams, P3s never ship. But with free code, kick off P3s 4x in parallel, pick the winner, and ship. This parallelization strategy unlocks projects previously deemed too low-priority.
3. Build localization and internationalization by default into all internal tools at zero marginal cost. With code free, features like multi-language support become table stakes rather than trade-offs against other team capacity.
4. Reframe engineering leadership as resource delegation: each engineer now manages 5-5,000 concurrent agent team members constrained only by GPU capacity and token budgets. Plan 1-day, 1-week, and 6-month structures to productively harness this capacity.
5. Automate synchronous human time first. With agents handling implementation, redirect human attention to unblocking bottlenecks, reviewing outcomes, and steering long-horizon work—not writing code. This moves engineering toward staff-level systems thinking.

Topics

• AI Agent Delegation Patterns
• Harness Engineering Framework
• Prompt Engineering for Teams
• Agent-Driven Software Development
• Token Budget Resource Allocation

Transcript Excerpt

Our next speaker is here to speak about harness engineering. How to build software when humans steer and agents execute. Please join me in welcoming to the stage member of technical staff at OpenAI, Ryan Leopo. Good morning, London. I'm super excited to be here today. I'm Ryan Laapo and for the last nine months I have had the privilege of building software exclusively with agents. Uh I am a token billionaire and I believe that in order for us to get into our AGI future, we want everybody to be t...