ReverseX

ReverseX Engineer — AI-Powered Software Investigation, Reconstruction & Verification** Understanding and recreating software functionality traditionally requires developers to manually explore interfaces, document behavior, write code, and test the results. ReverseX Engineer explores how much of that process AI can automate. Powered by **Meta Muse Spark**, ReverseX autonomously investigates authorized web applications through Playwright, observes user interactions, and records evidence of how features behave. Using those observations, ReverseX can generate an independent implementation without providing the original application's source code to the reconstruction model. Unlike systems that simply generate code, ReverseX includes an independent behavioral testing pipeline that compares reconstructed functionality against the original application. When differences are detected, its bounded repair system can attempt corrections and revalidate the results. An evidence-verification layer checks whether reported conclusions are supported by recorded observations. Results are presented through two interfaces: a Simple Report for nontechnical users and a Developer Report containing technical evidence, test results, and implementation details. The working prototype demonstrates autonomous investigation, feature reconstruction, behavioral validation, and controlled AI-assisted repair. It also supports investigation of approved local applications, with clear reporting of unverified behavior and incomplete results. Our vision: Move toward AI systems that can understand software through interaction, reconstruct observable functionality, and verify their own engineering outputs rather than relying on generated code alone.