Infer is capable of quickly analyzing programs and changes in massive codebases in large part because of its scalable and compositional design. Although compositionality also makes it naturally amenable to incremental analysis, numerous technical challenges – reliably determining which procedures are affected by a code change, efficiently caching/delivering summaries at scale, and avoiding flakiness or loss of precision – have in the past rendered these benefits more theoretical than practical. In this talk, I will discuss the history of Infer’s incremental analysis infrastructure, present recent strides we have made towards addressing each of the aforementioned technical challenges, and describe some future work aiming to deliver Infer analysis results to users even earlier in their development process.