article --summary
The summary
This interview piece looks at how developers can keep up with fast-moving AI agent tools without relying on workflows that go stale with each model release. Angie Jones separates two skills: learning to work alongside an agent, and learning to build one that others can depend on. She points to shared protocols such as Model Context Protocol and Agent2Agent as stable footing for builders, and notes that the foundation's certification and working groups cover identity, reliability, security, and governance.
Jones also describes lessons from Block and the open source Goose project. At Block, a small group made the development environment easier for agents, so other engineers benefited without learning every prompting technique. At Goose, maintainers moved to an issue-first process so contributors and reviewers agree on the problem and design before a patch is written, since agents make proposing changes cheap but review still costs human time. The piece closes with her view that junior engineers bring speed with agents while experienced reviewers decide which changes are worth shipping.
Editorial note: This is a summary of an article first published by The New Stack. The full piece is on their site.