Edge Delta said it has introduced a telemetry-native AI SRE it calls the prod immune system, which works directly on a company's telemetry to understand, investigate and operate production. It also launched AI SRE Arena, an open benchmark that measures how well AI SREs catch and diagnose production incidents, and published the first results.
In those results, the company said, Edge Delta detected 18 of the 21 incidents on its own, while the competing product tested alongside it detected 12. No final recommendation from Edge Delta was judged incorrect or unsafe.
The first version of the benchmark adds 21 faults to a Kubernetes deployment of the OpenTelemetry demo application. An AI model acts as judge, reading each product's final report and grading its root cause and proposed fix against a fixed answer key, using the same rubric for every product. The scenarios, answer keys and scoring code are open source, and the company said any team can run the benchmark itself.
Edge Delta said its AI Teammates run on the telemetry pipeline itself, where logs, metrics, traces and events already flow, rather than pulling data from other monitoring tools after an alert fires. The agents spot abnormal behavior early, investigate with the full context and propose fixes that pass through approval gates the team controls. Recent additions include Loops, standing goals for agents between incidents, and Guardrails, which lets teams set how much the agents may do on their own.
Ozan Unlu, co-founder and chief executive, said customers had used Edge Delta to collect and route their telemetry and that agents working on that data now do much of the investigating. Fatih Yildiz, co-founder and chief technology officer, said most agents see production only through someone else's dashboards, while Edge Delta's read telemetry as it streams through its pipelines. The company said the product is available now with a free 14-day trial and paid plans from $20 a month, and that it will provide more product comparisons in the coming weeks.
