ControlTheory Launches Dstl8 2.0 AI SRE With Metrics, Traces and GPU Visibility

ControlTheory launched Dstl8 2.0, its telemetry-native AI site reliability engineering platform, at KubeCon + CloudNativeCon North America 2026 in Salt Lake City, extending its Telemetry Distillation approach from logs to metrics, traces and events. Dstl8 2.0 tracks AI workloads from an agent's first step through model calls, including those routed via LiteLLM gateways, down to the vLLM model server and the NVIDIA or AMD GPU, reporting model, provider and token counts per call while excluding prompts and replies, and flagging reserved-but-idle GPUs. Metrics and traces are in preview, and CEO and co-founder Bob Quillin said, "Instead of following the herd and adding an AI SRE to observability, we built observability into our AI SRE.

ControlTheory Launches Dstl8 2.0 AI SRE With Metrics, Traces and GPU Visibility SALT LAKE CITY -- ControlTheory launched Dstl8 2.0, the latest version of its telemetry-native AI site reliability engineering platform, at KubeCon + CloudNativeCon North America 2026. The release extends the company’s Telemetry Distillation approach from logs to metrics, traces and events, so Dstl8 works from telemetry it distills itself rather than data pulled from a separate observability platform. The company said Dstl8 now follows AI workloads from an agent’s first step through every model call, including calls routed through gateways such as LiteLLM, down to the vLLM model server and the NVIDIA or AMD GPU doing the work. It reports the model, provider and token counts on each call while keeping prompts and replies out, and flags GPUs that are reserved but sitting idle. “If your AI SRE finds an issue, diagnoses it, and fixes it, all on telemetry it brings itself, it’s fair to ask what your observability platform is for,” said Bob Quillin, CEO and co-founder of ControlTheory. “Instead of following the herd and adding an AI SRE to observability, we built observability into our AI SRE.” APM and Predictive Detection Dstl8 2.0 uses traces to detect application performance problems such as slow endpoints, rising error rates and new request paths, and ties each one to related logs, metrics and the deployment behind it. Metrics now come with forecasts for when disks, CPUs or GPUs will hit their limits, and the platform reads all signals together with Kubernetes events such as deploys, restarts and OOMKills. Other additions include on-demand dashboards, Prometheus- and Jaeger-compatible endpoints for existing Grafana and Jaeger setups, and a single setup command for Kubernetes, AWS and OpenTelemetry sources. Dstl8 2.0 is available now, with metrics and traces in preview.