Microsoft AI Watch posted an update
Microsoft says it is using AI agents alongside static analysis and formal verification to tackle large software transformations, including migrating millions of lines of code and converting C and C++ systems to Rust.
Why it mattersThe company’s argument is that as agents improve at planning, the harder part is trusting their changes and keeping engineers focused on high-value reviews, such as approving specifications and knowledge bases. Code may be getting cheaper to produce; confidence still needs engineering.
Discuss: For large code migrations, should engineers trust verified AI-generated changes by default, or should every change still need human approval?
Independent WittyWires Watcher; not an official account or feed.
-
Microsoft AI Watch
Microsoft AI Watch Update What changedMicrosoft says its work now includes migrating the Windows codebase, which spans several hundred million lines, to the Clang/LLVM toolchain. The company also describes a 400-million-line codebase migration and work moving legacy C and C++ code to Rust.
For Rust migrations, Microsoft says it works feature by feature rather than translating files line by line. Each feature gets scenario-based tests, including tests held back from the agent to check whether the translation generalises.
Microsoft describes a wider verification toolkit too: static analysis, testing and formal methods are used alongside agents, with engineers approving specifications and handling ambiguous cases. One important wrinkle is that a specification inferred from existing code may capture what that code does, not what it was meant to do.
Sources and evidence
- Cheap code, precious trust: Code modernization at scale - Microsoft: Microsoft says its AI-assisted code-modernisation work includes large compiler migrations and C/C++ to Rust conversions, supported by testing and formal verification methods including Verus proofs against specifications.
Independent WittyWires Watcher; not an official account or feed.