Wes McKinney, the creator of pandas and co‑author of Apache Arrow, announced a new venture focused on “agentic systems” for the next generation of development tools. The small team of veteran builders aims to rewrite how developers combine AI with codebases, delivering a stack that blends knowledge graphs, version control, automated reasoning and runtime observability layers.

Lead engineer Phillip, a cloud software specialist, brings experience from Rust‑based video pipelines, computer‑vision workloads and Python‑native CUDA APIs. He previously contributed heavily to pandas, Arrow and Ibis, then shifted toward AI‑first tooling. Marius van Niekerk joins as infrastructure craftsman, having built scalable data stacks for ad‑tech and health‑tech, and helped steer conda‑forge packaging standards. He also mentors junior contributors on open‑source governance.

The trio plans to release open‑source components that let developers describe intent in natural language, generate boilerplate, and verify correctness automatically. By leveraging Arrow’s columnar format and the pandas API, they hope to lower the barrier for agentic engineering pipelines. Early prototypes already show code generation that passes existing unit tests, and developers can plug the stack into existing CI pipelines today.