Field notes, not thought leadership.
8 published
- Content Platforms3 min read
Your CMS belongs in the production incident timeline
A content change can alter a live experience without a deployment. CMS activity should be visible beside application, cloud, and release events while the incident is active.
- AI Engineering3 min read
Treat a model upgrade like a permission change
A new model can use familiar tools differently. Replay an agent's permitted work and approval boundaries before changing the default.
- AI Engineering2 min read
Stopping an AI agent means checking what can still run
A cancelled agent can leave behind queued work or usable credentials. Rehearse cancellation with a harmless task and check the systems that could continue acting.
- AI Engineering2 min read
Your fallback model needs the same data rules
When a primary AI model is unavailable, its replacement can change how sensitive data is retained. Check the fallback before it receives the request.
- GEO2 min read
AI search starts with a question your page can answer
Take one question a buyer asks and check whether your page actually answers it. A worked example shows what useful detail looks like, without promising an AI citation.
- AI Rescue2 min read
Your AI-built app passes login. Can it keep accounts apart?
A successful sign-in does not prove that private records stay private. Use two test accounts to check a failure that a one-person demo can miss.
- Website performance3 min read
Use Core Web Vitals to investigate customer friction
Loading, response time, and page movement can interrupt a visit. Read these measurements alongside customer behavior and recent releases to decide what to investigate.
- Content Platforms2 min read
Before you reuse a CMS entry, decide who can change it
Two pages can contain the same words without sharing the same fact. A small content-model example shows when reuse helps and when it ties unrelated edits together.