Agentic Coding at AirBnb
summary
Airbnb reports 97% weekly PR authors using agentic AI with many engineers saying it saved more than eight hours a week; PR throughput rises after two flat years measured by hours in active agent sessions and a true throughput metric that weights PR count by size. Internal SDK spawns 40+ team-built agentic apps (example: Draco for data analysis), CLI use splits technical and nontechnical users, and code quality shows no measurable drop. Remote managed agent infrastructure and orchestration experiments are in progress.
AI-written summary · about
chapters (19)
- 0:00Introduction to agentic coding
- 1:06AI adoption at Airbnb
- 3:13AI myths and truths
- 5:01Evolution of agentic tools
- 7:07Measuring productivity gains
- 9:56AI beyond engineering roles
- 12:26Future of prototyping
- 13:51Vibe coding quality insights
- 15:56Async agentic workflows
- 17:25Building agentic platforms
- 20:13CLI as universal interface
- 21:38Agentic migrations
- 22:21SDK for custom agents
- 24:07Data analysis agent
- 24:51Remote agentic environments
- 27:57Next big product: Air Remote
- 28:38Agent orchestration concepts
- 30:02Context graph and AI harnesses
- 30:44Q&A session
related talks
35mAgentic coding
26mAI Agents as Your Co-Developer
16mCoding Reinvented How AI Agents Are Shaping the Future of Software Development
27mAgentic Engineering: Working With AI, Not Just Using It
45mAgentic Coding, the New Frontier of Platform Engineering
01mAgentic Coding
23mScaling AI Adoption Across 3000+ Developers at Booking
21mThe Agent Native Company