The tools are compatible with Claude Code, Codex, and OpenCode agents, and they replayed 10 production scenarios in under a minute during internal company testing. Every proposed fix requires a human to review and approve or deny the change before release. By automatically converting production failures into a test set, the system eliminates the need for manual transcript audits or expensive grading by large language models.
Wednesday, October 21, 2026
Story
DigitalOcean Launches Tools to Debug Silent AI Production Failures
AI9 days ago
The cloud computing firm now offers a set of services that allow developers to isolate errors in autonomous assistants that appear to be working while delivering incorrect responses. DigitalOcean has released its Agent Learning and Agent Data Services, including a public preview of Signals to monitor live traffic for indicators of user frustration using rule-based detection rather than AI models. The suite also features Simulations for replaying failed conversations to verify fixes, as well as Insights, while Agent Data Services is available in private preview to let developers diagnose problems by reviewing documents and support recordings.
Looking for more context or analysis?
Tilt chat provides reliable, tailored insights, alerts, and automations.