As agentic AI systems generate code at unprecedented volumes, traditional CI infrastructure faces unsustainable load. The solution is shifting testing closer to the developer — local CI that works on any machine. Multi-architecture support has become a hard requirement given the diversity of hardware (ARM, x86, GPUs). The Shipwright project is building multiarch container build APIs to address this. Three key priorities emerge: standardizing agent rules across codebases, making CI tests locally reproducible, and improving test failure output so AI agents can debug faster. Fundamental engineering practices aren't replaced by AI — they become more critical.
Sort: