How-To
How to evaluate incident-response agents
Build repeatable cases and check recovery independently, with a runnable Python grader.
Practical guidance for SRE and platform teams operating reliable services and production AI. Independent analysis, field guides, and working resources.
Selected reading
Prompt caching: measure the cost of useful work
Build a prompt-caching pilot that counts retries, failed work and unknown charges, with a local calculator and sample ledger.
A practical pilot for separating cache savings from the cost of completed work.
Read the article
Original writing / Newest first
More from AIOpsSRE
Technology Overviews
Supervising coding agents with Herdr
Check agent status, terminal persistence, and recovery before adopting Herdr.
AIOps
What MTTD misses
Separate detection from response and find delays that an improving average can conceal.
AIOps
When agents can change production
Define the action, contain its scope, and verify what changed at the destination.
How-To
Connect the signals with OpenTelemetry
Use instrumentation and context propagation to connect the evidence in an investigation.
How-To
The SRE and platform handoff
Clarify ownership across shared platforms, service reliability, and incident response.
Releases, research & operational impact
Latest AI news
Inference infrastructureNews analysis
TensorRT’s multi-GPU serving results put capacity in focus
NVIDIA’s September 21 demonstration accelerates one video-generation workload across eight GPUs. A service decision still needs the same offered load and completed-work count.
Read the analysis
Agent evaluationNews analysis
SWE-Serve tests coding agents through live inference
NVIDIA’s new SGLang benchmark finds patches that pass local checks but fail a running server. Its result is a reason to examine what a green test actually covers.
Local AI agentsNews analysis
Antigravity brings local models to its agent SDK
Google adds local Gemma and LiteRT workflows. Running the model on a workstation still leaves a separate decision about what its tools may read, change or send.
From reading to doing
Work through it, step by step.
Incident response
Build a usable runbook
Work through a database connection example, from the first symptom to a verified recovery.
Worked example + Markdown templateRead the guide
Reliability planning
Put your error budget to work
Define the service objective, calculate the allowance, and agree when a release needs to wait.
Calculations + policy fieldsRead the guide
Observability
Trace a request across services
Start with one important request path. Check propagation, sampling, and the gaps in your evidence.
Implementation guideRead the guide