Agent Factory Recap: Agentic engineering workflows in Google Antigravity 2.0

Short summary

Rody Davis, a Google agentic engineer, demonstrates how Antigravity 2.0's unbundled agent-first platform—comprising a desktop Agent Manager, CLI, SDK, and specialized IDE—enables developers to scale impact by offloading toil like test generation and marketing-site builds. Key strategies include using 'skills' as compressed context cheat sheets for agents, maintaining flat Bonsai-style architectures for easier agent steering, and providing pattern examples so agents can replicate consistent code. Davis predicts non-technical founders will ship vibe-coded companies by 2026, but poor codebase health—not context windows—will be the real bottleneck, creating demand for engineers who can fix AI-generated production failures.

• •Antigravity 2.0 unbundles into Agent Manager, CLI, SDK, and IDE for composable agentic workflows
• •Skills act as compressed-context cheat sheets that make agents faster and more accurate
• •Flat Bonsai-style architecture and pattern examples are key to steering agents effectively

Generated with AI, which can make mistakes.