373 saved links — your professional home turf. The corpus documents SRE’s third age arriving in real time: from Google’s canon, through the observability 2.0 debate, to agents holding the pager. Read alongside AI Agents — this page is effectively the R&D feed for your incident-management platform.
Related: DevOps · AI Agents · Kubernetes · Platform Engineering
Start here
- Google’s free SRE books — still the canon; plus 11 lessons learned from 20 years of SREing and the “prodverbs”.
- “Hired an SRE. Renewed Datadog. Still having the same incidents.” — SRE ≠ observability ≠ tools; the best short corrective in the tag.
- the future of software engineering is SRE — when code gets cheap, operational excellence wins. The thesis behind your pivot.
- SRE Roadmap 2026 and SRE vs. platform engineer.
- howtheysre — how top tech companies structure SRE teams.
The AI-SRE wave
The densest 2025–26 cluster — the market building what you’re building.
- where and how Google deploys agentic AI in SRE · Google’s VP of SRE on how AI is changing SRE — and how it isn’t · Gemini CLI for outage response: from paging to postmortem.
- Anthropic’s cookbook: an incident-response agent with read-write MCP tools — diagnosis, remediation, postmortem — the closest published blueprint to your platform.
- Azure SRE Agent: end-to-end agentic operations (video) and connecting the Azure SRE agent to Azure MCP.
- OpenSRE — an open-source framework (and training environment) for AI SRE agents · 60+ tools in the AI-SRE space, collected.
- The Third Age of SRE: AI Reliability Engineering · agentic observability: AI on top of your existing tools.
- Case studies: Uber’s Genie on-call copilot — 70k answers, 13k hours saved · Intuit: GenAI on-call across 325+ clusters · Google: GenAI cut incident-summary time 51% · accelerating incident response with GenAI.
- DIY: building an AI SRE agent to analyze production · intelligent alert handling with Prometheus + n8n + OpenAI · an SRE extension for the Gemini CLI.
The observability 2.0 debate
- the 1.0 vs. 2.0 distinction — “it’s not marketing speak” and Charity Majors explaining the shift (video).
- the technical core: one source of truth, arbitrarily-wide structured log events.
- “Is it time to version observability?” with the companion essay.
- the observability cost crisis, part one — whether your o11y should cost what it costs.
- the Observability CAP theorem · every SDLC stage is collapsing except monitoring — and monitoring must evolve into the feedback loop for agents.
- observability’s past, present, and future · the weekly tombstone industry, for perspective.
OpenTelemetry in practice
- the “Demystifying OpenTelemetry” guide and OTel Blueprints — cutting the deployment complexity.
- OpenTelemetry in 180 seconds (video) · a curated OTel resource list.
- OpenTelemetry tracing implemented in under 200 lines of code — the best way to actually understand it.
- Prometheus & OTel, better together · OTel best practices: naming · Elastic Distributions of OpenTelemetry (EDOT).
- the Aspire dashboard — best local OTel visualizer.
Observing agents and LLMs
Where your two worlds fuse — telemetry for systems that reason.
- GenAI observability with OpenTelemetry and Aspire · tracing agent sessions (video).
- AI agent observability, explained · top 5 agent-observability practices.
- the LLM stack’s new golden signals · the complete LLM observability guide.
- OTel observability for AI agents · agentevals — evals computed from OTel traces — evals and observability as one discipline.
- closing the verification loop: observability-driven harnesses for building with agents.
- LangSmith agent observability (video) · Galileo on multi-agent observability · AI observability pricing compared — 8x to 100x spreads.
The Prometheus/Grafana stack
- a Prometheus basic tutorial · Grafana for Beginners and the Grafana stack explained (videos).
- modern Grafana dashboards for Kubernetes · distributed tracing with Tempo.
- Scaling war stories: why TiDB moved from Prometheus to VictoriaMetrics · Zomato’s journey from Thanos to VictoriaMetrics · a Prometheus core dev on why classic metrics collection still wins.
- eBPF frontier: Coroot — eBPF-powered open-source APM · Netflix’s bpftop.
- OneUptime — the open-source Datadog alternative.
Incident craft
- how Netflix empowers engineers with incident management.
- a complete open-source incident-response program pack — process, from the ground up.
- Airbnb: your status page shouldn’t depend on the system it monitors — circular dependencies in monitoring.
- debugging event-driven systems: the five recurring problems — directly relevant to your Service Bus architecture.
- Culture: Google SRE applying control theory (STPA/CAST) to complex systems · the free systems-engineering syllabus · Ben Treynor Sloss on the evolution of SRE · the SRE deep dive into the Linux page cache · positive affirmations for SREs — because the pager is heavy.