https://jaredl.dev/tags/incidents/ Recent content in Incidents on Agent-Native Ops https://jaredl.dev/posts/slack-native-ai-oncall-agent/thumbnail.png https://jaredl.dev/posts/slack-native-ai-oncall-agent/thumbnail.png Hugo -- 0.147.7 en-us Sun, 06 Sep 2026 11:33:56 +0800 https://jaredl.dev/posts/five-phases-of-ai-agency/ Wed, 03 Jun 2026 00:00:00 +0000 https://jaredl.dev/posts/five-phases-of-ai-agency/ Agent maturity isn't how autonomous a system sounds — it's which decision boundary you moved from human or code into the agent. Five phases, five boundaries. https://jaredl.dev/posts/what-customers-want-when-things-break/ Wed, 13 May 2026 00:00:00 +0000 https://jaredl.dev/posts/what-customers-want-when-things-break/ Customers want four things when things break — service works, fix fast, informed, no repeats. Every on-call layer reorganizes when the responder isn't human. https://jaredl.dev/posts/self-evolving-triage-agent/ Fri, 01 May 2026 00:00:00 +0000 https://jaredl.dev/posts/self-evolving-triage-agent/ Two models, two jobs — one drafts triage responses, a different model audits every word and tool call. Patches self-validate, then a human cherry-picks. https://jaredl.dev/posts/two-boundaries-redact-and-sanitize/ Thu, 30 Apr 2026 00:00:00 +0000 https://jaredl.dev/posts/two-boundaries-redact-and-sanitize/ The LLM sits between two boundaries — redact sensitive strings before you prompt, and treat output as attacker-controlled once prompt injection is on the table. https://jaredl.dev/posts/prefilter-classify-triage-funnel/ Wed, 29 Apr 2026 00:00:00 +0000 https://jaredl.dev/posts/prefilter-classify-triage-funnel/ A cheap classifier should pay for the expensive investigator. Three stages — pre-filter, classify, triage — match cost to value on a mixed-value alert queue. https://jaredl.dev/posts/investigate-locally-triage-server-side/ Wed, 15 Apr 2026 00:00:00 +0000 https://jaredl.dev/posts/investigate-locally-triage-server-side/ Same triage agent, two deployment shapes — wide-open MCP locally for exploration, narrow direct-API client server-side for fast first-answer. https://jaredl.dev/posts/context-engineering-operational-ai-agents/ Wed, 01 Apr 2026 00:00:00 +0000 https://jaredl.dev/posts/context-engineering-operational-ai-agents/ Four techniques for context engineering operational AI agents — public MCPs, custom MCPs, CLI skills, bundling — turning a setup into a team asset. https://jaredl.dev/posts/slack-native-ai-oncall-agent/ Sun, 01 Mar 2026 00:00:00 +0000 https://jaredl.dev/posts/slack-native-ai-oncall-agent/ An AI co-oncall pattern — the agent investigates in parallel and writes findings to a Slack thread, so the engineer arrives at a running log, not a blank slate. https://jaredl.dev/posts/time-to-customer-awareness/ Sun, 01 Feb 2026 00:00:00 +0000 https://jaredl.dev/posts/time-to-customer-awareness/ TTC-Aware is the window between customer impact and customer communication — where trust is earned or burned, and almost nobody measures it. https://jaredl.dev/posts/measuring-pagerduty-alert-noise/ Thu, 01 Jan 2026 00:00:00 +0000 https://jaredl.dev/posts/measuring-pagerduty-alert-noise/ Diagnose why an alert is noisy using three temporal axes — frequency, inter-event time, and resolution time — so the fix matches the cause.