Observed arrival · 2026-09-03
Mutant Web makes browser-agent mistakes fail the build
A local and GitHub CI test harness that throws hostile browser states at an agent’s proposed actions.
Field notes
Mutant Web inserts adversarial browser states between an agent harness and its proposed action, checking whether the proposal matches the expected action and case policy. The homepage exposes a sample failure in which an agent selected a stale filtered row, alongside a 30-case controlled run and a downloadable 77 KB JSON artifact. Its free starter supports local execution and GitHub Actions, while the paid Team feed is described as adding versioned cases and changing failure distributions.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue