OpenAI and Airbnb to share AI agent strategies at QCon

The upcoming QCon San Francisco 2026 conference, scheduled for November 16 to 20, will focus on the architectural challenges of integrating AI agents into production systems. As autonomous agents transition from writing code to executing live customer transactions and querying system data, senior engineers must establish new guardrails. The event, held at the Hyatt Regency San Francisco, features early-bird registration priced at $2,955 through October 13, offering a venue for practitioners to address these emerging operational risks.

Key sessions will highlight real-world deployments where agents handle significant scale and revenue. Weiping Peng, a distinguished engineer at Airbnb, will detail how the company guardrailed its AI customer support agent, which serves millions of users, using shadow testing, classifiers, and rapid-response mitigations. Meanwhile, Brian Yang from OpenAI will discuss how the company utilized coding agents to build and scale its OpenAI Ads platform to more than $100 million in annual recurring revenue in under six weeks, highlighting the balance between automated generation and human oversight.

Other industry leaders will address how production systems must adapt to be understandable to non-human users. Liz Fong-Jones of Honeycomb will share insights from developing a Model Context Protocol server, which is now used by more than 40% of Honeycomb's weekly active users to run production queries via agents. Additionally, Netflix engineers Joseph Lynch and Ayushi Singh will discuss adaptive compression techniques for distributed key-value storage, managing billions of daily requests and petabytes of data to balance latency and storage costs.

For practitioners, these developments shift the focus from basic prompt engineering to system-wide integration, token economics, and safety. The conference will also feature the InfoQ Certified Architect Program, a four-day cohort ending in a workshop led by Luca Mezzalira on November 19 to help attendees earn their ICSAET certification. Hands-on training sessions on November 19 and 20 will cover tools like Claude Code and Codex, helping engineers design robust systems that safely accommodate autonomous agents.