Microsoft CoreAI project targets AI translation of legacy Windows code

This story is less about Rust fan culture and more about how big platforms intend to use AI as a lever on decades of legacy code. Microsoft’s CoreAI team is explicitly targeting the problem of translating millions of lines of C/C++ into safer languages with significant automation, with a stated “North Star” of one engineer migrating a million lines per month. If that works even partially, it changes the economics of modernizing massive software estates like Windows, Office and Azure infrastructure.

From an AGI race perspective, this matters in two ways. First, it is AI being applied to the substrate the AI stack itself runs on—compilers, kernels, services. Being able to rapidly harden and refactor low-level systems code makes it easier to deploy more powerful models without drowning in security debt. Second, the project is itself a demanding testbed for code-understanding models: an AI that can faithfully rewrite complex OS components while preserving behavior is implicitly demonstrating deep semantic reasoning about programs. That’s exactly the sort of capability frontier labs are trying to push, and success here would feed back into better agentic coding systems for everyone.