Peter (Creo)

Peter is presented in E238｜聊聊Harness时代AI-First的组织架构：从信任人到信任AI as Creo’s CTO, with prior work on Meta’s Llama base-model team and Apple multimodal/model work. The episode uses his engineering account to define Harness Engineering as the system layer around agents rather than only prompt or context optimization.

Peter argues that AI can now handle more planning and coding, but architecture remains a human bottleneck because security, sandbox-host interaction, latency, and system tradeoffs require experienced judgment. He also describes Creo’s agent-driven CICD, bug triage, and low-risk autofixing workflows as examples of AI Coding Verification becoming part of the harness.

Connections

• Creo - company context.
• Harness Engineering, Agent Harness, and Context Engineering - technical concepts he explains.
• AI-First Organization, AI Organization Design, and Human Judgment Under AI - organization and human-role implications.
• AI Coding Verification and Agent Permission Boundaries - production quality and safety boundaries in his examples.