Analysis
An October 26 OpenAI customer story describes Greptile using GPT-6 Astra to test applications as well as inspect code. Its example is a button that disappears at an iPhone-sized browser viewport: Greptile starts the application for a pull request, changes the viewport and uses the model to examine the resulting screen. This is a company account of the workflow, not an independent evaluation.
The same account describes Greptile supplying codebase documentation to Codex through a Model Context Protocol server. Greptile says it tracks reversions of AI-generated code and sees fewer when agents receive that context. The article provides no numerical reduction or controlled comparison, so the claim supports a testing question rather than a quantified promise.
Our reading: review should leave evidence from both the code and the running application. A useful finding identifies the tested revision, viewport, steps and visible failure so another reviewer can reproduce it. Browser exploration can reveal a broken journey; it cannot by itself prove security, accessibility or coverage of every important path. Keep those acceptance checks explicit.
Source: Greptile takes bug detection beyond code review with GPT-6 Astra · 2026-10-26
Why it matters
A plausible code change can still produce an unusable interface; reproducible runtime evidence closes part of that gap.