Microsoft Project Strong ARMed plans AI code‑porting agents

Project Strong ARMed is one of the clearest examples yet of using AI to work on code at infrastructure scale rather than as a developer sidekick. The goal isn’t just Copilot-style suggestions; it’s to build agents that understand large, messy x64 codebases and systematically port them to ARM64 and cross‑platform targets, generating pull requests and evaluating dependencies along the way. If it works even moderately well, it could slash the friction of migrating workloads onto Microsoft’s own Arm-based Cobalt 100 servers and future custom silicon.([windowslatest.com](https://www.windowslatest.com/2025/12/30/microsofts-project-strong-armed-wants-ai-agents-that-auto-port-x64-codebases-to-arm64-on-windows/))

In the AGI context, this is important because it moves AI deeper into software engineering itself as an automated, persistent actor. A system that can reason over millions of lines of legacy code, plan a migration strategy and iteratively refactor until tests pass is edging toward the kind of generalized tool-using competence many associate with early AGI capabilities. It also hints at a world where AI agents continuously reshape the substrate they run on — porting, optimizing and hardening code to fit new hardware and platforms without human supervision.

For rivals, this raises the bar: simply offering code-completion isn’t enough. Hyperscalers with custom chips will need credible migration agents to reduce switching costs for enterprise customers, or risk leaving stranded workloads on the wrong architecture.