Read metrics and calculate the right rate
A reassuring fleet average can coexist with serious failures in one region. Follow the counts, units and time window to see which question a metric answers.
Articles
Guides, news and explanations about AI and reliable systems.
Not sure where to begin? Choose a short reading path, or browse at your own pace.
Reduce toil without automating the confusion
A retry script can save clicks while losing the judgment that kept reports from duplicating. Compare repair, simplification and automation around the whole task.
Bring a specialist into an incident without losing the thread
An unfamiliar storage account changes the questions. Preserve the observation, involve the right responder and keep sensitive details out of the general update.
Read a trace across concurrent work
A request takes 2.4 seconds, but its recorded operations add up to 2.8. Draw the overlap once, then investigate the half-second the trace does not explain.
Make the first on-call shift workable
Put access, ownership, escalation and recovery checks into a usable runbook. Consider coverage and workload alongside the technical procedure.
Set severity, roles and decision authority
Organize a Dispatch incident with impact-based severity, accepted roles and a concrete route to deciding whether a rollback can proceed.
Why Linux sees two files when you typed one name
Learn what Bash passes to a command, then use quotes, patterns, variables and help to predict the result in a disposable Ubuntu lab.
Find where the export waited, not just how fast status answered
Follow job-42 from acceptance to ready, separate the quick status request, and identify the missing evidence needed to explain processing and download.
Trace the journey from request to result
Follow an export from acceptance to authorized retrieval, mapping handoffs, dependencies, owners and the checks that distinguish failures.
How AI got from rules to reasoning models
Follow the changes in learning methods, data and compute that led to today’s models, and see why newer does not always mean better for a task.
Linux history explains why one machine has several identities
Unix, GNU, the Linux kernel and distributions contribute different pieces. Use that history to find which layer owns a missing command or system behavior.
Set up a safe Linux lab
Create a disposable Ubuntu lab, identify the machine and user, and verify that a small practice file survives stopping and restarting the VM.