# ControlTheory Launches Dstl8 2.0, the AI SRE That Doesn’t Need An Observability Platform - Link: https://thestackobserver.com/controltheory-launches-dstl8-2-0-the-ai-sre-that-doesnt-need-an-observability-platform/ - Published: 2026-11-09T12:49:05-05:00 - Author: Brian Ducharme _With Telemetry Distillation now spanning logs, metrics, traces, and events, Dstl8 follows AI workloads from agent to GPU, takes on APM, and predicts problems before they hit._ **SALT LAKE CITY, UT, November 9, 2026** – ControlTheory today launched Dstl8 2.0, its telemetry-native AI SRE, at KubeCon + CloudNativeCon North America. Dstl8 now extends Telemetry Distillation to metrics and traces in addition to logs, so it works from telemetry it distills itself rather than data borrowed from an observability platform. This release challenges a decade of conventional wisdom about why teams need observability at all. For years, teams have bought observability the way they buy insurance: because the alternative feels risky, and because it is what has always been done. They pay to ingest, index, and retain every byte, then sample and drop data to keep the bill in check. Companies are paying real-time prices for an archive; at ControlTheory, we call that the Telemetry Tax. AI SREs promised to change the equation, but most were built on top of that same platform. They investigate what it flags and reason over whatever survived the cost cuts. If nothing fires, they never look. That is what we call the Telemetry Trust Problem. “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. “For most teams, the honest answer is: less than they’re paying for. Instead of following the herd and adding an AI SRE to observability, we built observability into our AI SRE.” The approach is already in production. At an enterprise fintech, Dstl8 resolved 328 incidents across 13 Kubernetes clusters in its first two months, without the team writing a single alert rule. ![](https://thestackobserver.com/wp-content/uploads/dstl8-20-tr-1024x441.png) ![](https://thestackobserver.com/wp-content/uploads/dstl8-20-me-1024x893.png) ![](https://thestackobserver.com/wp-content/uploads/dstl8-20-kn-1024x630.png) **An AI SRE is only as good as the telemetry it’s handed** Two-thirds of organizations hosting generative AI models now run some or all of that inference on Kubernetes, according to the [2026 CNCF Annual Cloud Native Survey](https://tracking.us.nylas.com/l/114889fecece40548684b4b7028dd8ec/1/ea8f592a073e81cd5fa0d65229f9e68747364f6ac8a77138e3f6d29028fce7bc?cache_buster=1794242084). The same clusters now carry services, agents, LLM calls, and GPUs, and teams ship changes to all of them faster than ever. One bad deploy can ripple through every layer at once. A bigger blast radius needs a wider net. With Telemetry Distillation across logs, metrics, traces, and events, Dstl8 reasons over sentiment, patterns, anomalies, and severity to find and fix problems that used to take several tools. **Key benefits of Dstl8** * **AI workloads, from agent to GPU.** Dstl8 follows an AI request from the 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 shows which step is slow or failing, with the model, provider, and token counts on every call, while prompts and replies stay out. Plain-English readouts explain what each GPU and model server is doing and what to change, for example: “gpu-instance-0 holds 29.3 GiB, 66% of its memory, but was busy only 1% of the time. Something has it reserved without using it.” For teams running AI on GPU clouds, Dstl8 answers the first question in every incident: is it the GPU or the code? * **APM, without a separate APM tool.** Traces empower Dstl8 to detect application performance problems, slow endpoints, rising error rates, and new request paths, and tie each one to the logs, metrics, and the deploy behind it, down to the line of code. * **Predictive detection.** Metrics come with forecasts: when a disk will fill, when a CPU will hit its limit, when GPUs will max out. On traces, Dstl8 picks out the worst request paths on its own and points engineers straight to them. * **Cross-signal diagnosis.** Logs, metrics, traces, and sentiment are read together with Kubernetes events (deploys, restarts, OOMKills, a pod rescheduled off a GPU node), so every incident is explained by what changed and when, with evidence from every signal, not just the one a single tool happens to hold. For GPU and AI cloud providers and their customers, that means a faster answer to the question asked most, “Is it your infrastructure or my code?”, with the evidence to settle it, and a clear view of GPUs that are reserved but sitting idle. **Additional new features** * **Dynamic Canvas.** Create dashboards on demand, or have [Möbius](https://tracking.us.nylas.com/l/114889fecece40548684b4b7028dd8ec/2/5eca86264beb1add95676982b1d56d3e97e61e5e169880f91c49cbcfdea71cd7?cache_buster=1794242084), Dstl8’s AI agent, build them around each diagnosis, so engineers see the evidence instead of maintaining static dashboards nobody reads. * **Works with the tools you already have.** Prometheus- and Jaeger-compatible endpoints let Grafana and Jaeger read Dstl8 metrics and traces, so teams can keep familiar dashboards while Dstl8 does the work. * **Nothing to assemble, nothing metered.** One command, dstl8 setup, connects Kubernetes, AWS, and OpenTelemetry sources in minutes, with no integration project. There is no ingest meter, no investigation meter, and every resolved incident is remembered, so teams never solve the same one twice. “Most telemetry gets stored before anyone looks at it,” said Eric Anderson, co-founder and CTO of ControlTheory. “Telemetry Distillation flips that order. Every log, metric, and trace is analyzed first, close to the source, and what we keep is chosen based on what we found. That’s what lets Dstl8 show its work. And it looks ahead. ‘Your disk is full’ means you’re already in trouble. ‘Your disk will be full at 11 p.m.’ gives you time to fix it.”  **Availability** Dstl8 2.0 is available today, with metrics and traces in preview. See it live at ControlTheory booth 188 (Hall 1, Solutions Showcase, Salt Palace) at KubeCon + CloudNativeCon North America in Salt Lake City, November 9-12. Start a 14-day free trial at [dstl8.ai](http://dstl8.ai/). **Related** * [Dstl8 September 2026 Platform Launch](https://tracking.us.nylas.com/l/114889fecece40548684b4b7028dd8ec/3/72df431d0471ba81535cd3db53e8819fe42dec50947d5b000bb7a36ae055d95e?cache_buster=1794242084) * [ControlTheory at KubeCon North America](https://tracking.us.nylas.com/l/114889fecece40548684b4b7028dd8ec/4/a1ff42cf6eb09524ae46bc880b41e5828b5bab4ca5c6dc620145d01b2295cfa6?cache_buster=1794242084) **About ControlTheory** Founded in 2024 by observability veterans from StackEngine (acquired by Oracle Cloud) and CopperEgg (acquired by Idera), ControlTheory builds Dstl8, the telemetry-native AI SRE. Built on Telemetry Distillation, Dstl8 analyzes logs, metrics, and traces close to the source, finds root cause with zero alert rules, and delivers the fix to Slack, Claude Code, Cursor, and Codex, so teams don’t need a separate observability platform. ControlTheory also created [Gonzo](https://tracking.us.nylas.com/l/114889fecece40548684b4b7028dd8ec/5/30c108da39b84c3471206ce1ee47fcb7c06266481f36a44ca3608047bc26b83c?cache_buster=1794242084), the open-source OTLP terminal UI with 2,700+ GitHub stars. The company is based in Austin, Texas. Learn more at [controltheory.com](http://controltheory.com/).