From 75c3933a7a30db8ef6a19c1781079482a13857e0 Mon Sep 17 00:00:00 2001
From: Nicolai Ort <info@nicolai-ort.com>
Date: Mon, 23 Mar 2026 11:05:23 +0100
Subject: [PATCH] docs(day0): OTEL feedback talk

---
 content/day0/07_scalingsatisfaction.md | 51 ++++++++++++++++++++++++++
 1 file changed, 51 insertions(+)
 create mode 100644 content/day0/07_scalingsatisfaction.md

diff --git a/content/day0/07_scalingsatisfaction.md b/content/day0/07_scalingsatisfaction.md
new file mode 100644
index 0000000..d4f61b9
--- /dev/null
+++ b/content/day0/07_scalingsatisfaction.md
@@ -0,0 +1,51 @@
+---
+title: "Scaling on satisfaction: Automated Rollouts Driven By User Feedback"
+weight: 7
+tags:
+ - platformengineeringday
+ - staging
+ - rollout
+ - feedback
+ - otel
+---
+
+<!-- {{% button href="https://youtu.be/rkteV6Mzjfs" style="warning" icon="video" %}}Watch talk on YouTube{{% /button %}} -->
+<!-- {{% button href="https://docs.google.com/presentation/d/1nEK0CVC_yQgIDqwsdh-PRihB6dc9RyT-" style="tip" icon="person-chalkboard" %}}Slides{{% /button %}} -->
+<!-- {{% button href="https://github.com/thomasvitale/kubecon-2026-gitops" style="info" icon="code" %}}Code/Demo{{% /button %}}  -->
+{{% button href="https://whitneylee.com" style="info" icon="link" %}}Website/Homepage{{% /button %}}
+<!-- {{% button href="https://thomasvitale.com" style="info" icon="link" %}}Website/Homepage{{% /button %}}
+
+## What they are actually talking about
+
+- A way of creating metrics/traces from an llm and anlyzing them
+- The integration of the user's feedback 
+- Basicly the integration of what the variant did on the server to the vote event to promote based on feedback
+- Combined with an into to OTEL
+
+## Baseline
+
+- Question: How do we know that content generated by llms and delivered to our users is good or bad
+- Idea: Using OTEL and user feedback to drive canary deployments and rollout
+- Needed: A standardized vocabulary (so we can talk to any telemetry system)
+
+## Demo Architecture
+
+The start of the talk featured an evolving story (5 parts) and let the attendees vote on if they like it or not to emulate rollouts of a new application version with immideate user feedback. It was based on flagger deciding every thrity seconds if the user feedback allows promotion of new versions.
+
+- Audience get's plit to two variants running as knative deployments
+- OTEL Collector collects telementry data and the platform (Flagger) uses it as the basis for it's decisions
+- The collection was done by creating a user session span with a span event (aka a log) regarding the voting -> Span events are deprecated and will be moved to a logs api
+
+## Now to our platform
+
+- Stack: Kubernetes on hetzner with components (cert-manager, ingress, knative, ...) packaged by carvel
+  - knative as the deployment target for apps
+  - Flagger as the release decision tools
+  - OTEL for instrumentation
+  - Crossplane
+- API: StoryApp CRD as the main interface for controlling what we want to deploy
+
+## Takeaway
+
+- Include user feedback in the decision process for new rollouts
+- OTel can be used to automate thes
\ No newline at end of file