Airbnb’s custom AI agent now handles ~33% of North American support issues, global rollout planned. Their custom agent resolves about a third of U.S. and Canadian support tickets and Airbnb is preparing a global rollout. This is a large-scale example of production agentization — expect investments in escalation handoffs, monitoring, and artifactization as teams treat agents as delivery lanes (Principle 04, Principle 09).

The o16g Manifesto Validates What We’ve Been Building Since July. Cory Ondrejka’s manifesto codifies outcome-engineering practices and signals industry convergence on outcome-first, agent-centered workflows. Treat it as a practical blueprint for aligning human intent with orchestration and outcome validation (Principle 01, Principle 16).

Peter Steinberger (OpenClaw) joins OpenAI to drive next-generation personal agents; OpenClaw remains open source. OpenAI brings the OpenClaw founder onboard while keeping the project open source, accelerating personal-agent primitives and ecosystems. That movement will expand the set of reusable agent components you can plug into orchestration layers and personal-agent flows (Principle 09).

I built RetrieveIT.ai in 6 days with Claude Code — proof Context Engineering works at speed. RetrieveIT.ai combines permission-aware semantic search across GitHub, Confluence, Slack, Gmail, and Drive into a single product built in six days with Claude Code. It’s a concise example of rapid context engineering and how deployable knowledge artifacts shorten the path from intent to outcome (Principle 06).

Just Talk To It — the no-bs Way of Agentic Engineering. Peter Steinberger outlines pragmatic agentic patterns — parallel agents, human checkpoints, and blast-radius controls — for shipping agent-driven work fast and safely. Use these patterns to preserve human oversight while scaling agent productivity inside your orchestration and delivery systems (Principle 03, Principle 09).