Articles
Operator notes.
Problem-first writing on AI, automation, and the process underneath. Each piece names a real bottleneck, exposes the broken layer nobody owns, and shows what the working system actually looks like.
Short-lived tokens aren't AI agent lifecycle management
AI agent lifecycle management isn't a token expiry. Okta shipped Agent SSO on 24 August 2026. The four transitions it can't decide for you are still yours.
7 min read → Enterprise AILong-running AI agent tasks break on the wait, not the work
Long-running AI agent tasks fail in the gaps between steps. MCP's 22 August 2026 roadmap and AWS's session defaults show exactly where the state goes missing.
8 min read → Enterprise AIThree AI vendor shutdowns in five weeks. Every migration path is: rebuild it.
An AI vendor shutdown takes the workflow with it. Flowise, Relay.app and OpenAI's Assistants API all say the same thing: rebuild it from a description.
8 min read → AviationAI crew scheduling: the redundancy is at the wrong layer
AI crew scheduling now gets dual-cloud failover. The worst crew collapse in Ryanair's history involved nothing going down — a calendar definition changed and the plans underneath it didn't.
8 min read → Enterprise AIAI model routing is a business rule nobody wrote down
AI model routing decides which model does your work. Stripe reportedly paid over $7B for that switch — and in most stacks a vendor default is setting the policy.
8 min read → Enterprise AIAI agent memory is a second database nobody wrote a retention policy for
AI agent memory is a second copy of your data with no owner, no expiry and no reconciliation. Three 2026 measurements show what that costs.
9 min read →