GuideIncident response
From a postmortem finding to a working improvement
A backup-failure alert illustrates the gap between shipping a postmortem action and giving its recipient the access needed to respond.
Articles
Guides, news and explanations about AI and reliable systems.
CommentaryEngineering leadership
Making time off workable for an SRE team
A vacation backup who can receive pages but cannot perform recovery shows why coverage must transfer access, knowledge and workload together.
CommentaryEngineering leadership
When on-call work no longer fits the delivery plan
Repeated night work affects the next day’s capacity. Immediate relief and repairs to recurring demand have different jobs in a sustainable rotation.
CommentaryEngineering leadership
Where SRE leadership changes the reliability plan
SRE leaders resolve collisions between reliability work and delivery commitments, remove response obstacles and decide what the team can actually fund.
GuideObservability
What an AIOps assistant adds to an investigation
A deployment and a simultaneous dependency outage illustrate how AI-assisted observability can locate useful evidence and distinguish competing explanations.
ExplainerAI in production
From an anomaly score to a useful service signal
Baselines, event-level precision and recall, and the cost of waiting determine whether an unusual pattern belongs in investigation or paging.
GuideObservability
A first distributed trace from request to background job
A queued background job gives an initial tracing project a concrete question: did the delay occur before the worker started or during execution?
GuideAI in production
Routing support tickets with Python and NLP
An executable keyword router provides a baseline for NLP classification. An ambiguous login ticket shows how handoffs and human triage affect the result.
ExplainerObservability
Why a refreshing dashboard can still show stale data
Observation, collection and evaluation timestamps explain why the last zero-error result may no longer describe the service.
ExplainerSLOs & reliability
Choosing an SLO for the work users need completed
A queued upload illustrates the choice of SLI, target and window, including why endpoint success can miss the user’s unfinished task.
ExplainerObservability
Reading time, concurrency and gaps in a distributed trace
Concurrent spans, retries and selected samples explain what a trace reveals about one request and where an investigation needs another observation.
ExplainerSLOs & reliability
How SRE feedback loops reveal work that moved out of view
Alert cleanup can reduce interruption or conceal a recurring failure. Customer reports and correction effort help distinguish the two results.