Writing
Notes on AI, automation, and marketing that actually ships.
Field notes from building and running AI systems for marketing teams.
Get new pieces in your inbox. Subscribe to the newsletter.
ai agents
Brand DNA for AI Marketing Agents: Extract, Lock, Inject
A brand PDF cannot steer ten AI agents. Here is the four-step loop we use: extract Brand DNA, approve it, load it into drafting tools, and check each draft.
SEO
How to Stop AI Content Speed From Creating Keyword Cannibalization
Fast AI publishing makes it easy to ship pages that compete with each other. Here is a five-step check before each new brief to keep one clear page per search intent.
AI marketing
JEV for Marketing Ops: Score and Choose First, Write Second
A practical guide to using TypeSafe's JEV as a low-cost decision layer before drafting, spending, routing leads, or publishing.
AI Visibility Trackers: Buy by Job, Not by Engine Count
A practical matrix for choosing Otterly, Peec, Scrunch, Profound, or first-party checks, plus the workflow to run after measurement.
n8n
n8n vs Claude Code vs Claude Skills: What to Use for Marketing Ops
The choice depends on the job: same method every week points to a Skill, multi-tool loop with judgment points to Code, and reliable event-driven plumbing points to n8n. All three support approval gates before anything touches a live system.
AEO
What is llms.txt for Marketers (And What It Is Not)
What llms.txt can help agents find, what it cannot control, and when it is worth publishing.
Claude
5 Rules From the Claude Fable 5.1 System Prompt for Marketing Agents
Five paste-ready rules from the Claude Fable 5.1 system prompt for handling missing evidence, missing files, tool results, tagged text, and writing.
AI marketing
Claude Got Marketing Write Access. Keep the Measurement Desk on READ / PROPOSE.
When Claude connectors can touch ads, CRM, and CMS, run a weekly measurement desk: anomaly flags, dashboard narrative, GTM event specs — READ or PROPOSE, then stop for approval.