An end to buggy web apps

Replay argues that coding agents can eventually detect and fix every bug users encounter, but only if they have reliable evidence about how an application behaves. Its proposed ingredients are capable agents, tools that reveal both bugs and their causes, and a system connecting discovery to repair. Replay’s time-travel debugger records the inputs to Chromium and Node.js with low overhead, then allows an application’s execution to be replayed deterministically. The resulting record exposes code execution, program state, and control and data dependencies, giving agents detailed context to trace symptoms back to root causes instead of guessing.

Replay QA uses this debugger to explore applications, reproduce sessions from partial or redacted user data, and identify issues ranging from broken features to performance and security problems. In a test of 40 apps promoted by app-building platforms, Replay says an independent system confirmed and deduplicated more than 900 bugs. Two examples illustrate its approach: for a nonworking download button, QA pinpointed a missing React click handler, while another report traced a shifting page layout through React commits, network requests, and component values. These reports are contrasted with surface-level testing that can flag a failure but leave developers to infer its cause. Replay QA is generally available, with integrations, scheduled testing, and an API; the company is also developing deeper backend recording and full-stack debugging.