OpenAI unleashes agents on neglected scientific lab code
A new exploratory report tests coding agents on academic repositories and bioinformatics. It shows that models can clear technical debt and rewrite old tools in Rust, provided a human guides them like a product manager.