Writing on content, context and provenance

The arguments

Your codebase is a CMS nobody chose

Content-as-data in a React codebase is good architecture that quietly makes a company's most important words editable only by its engineers. The fix is not to abandon the pattern — it is to add the layer that was always missing from it.

The prompt was never the problem

Teams spend enormous effort on prompt wording and almost none on what the model was allowed to know. Retrieval from a real content graph changes output more than any phrasing does, and it is the part nobody wants to build because it requires knowing what content you have.

"Last edited by" answers no useful question

Who wrote it, what generated it, which facts it relied on, who reviewed it and who published it are five separate questions. One field answers none of them, and the moment a claim is disputed is the moment that becomes expensive.

Availability is a fact, not an adjective

The most common content defect in software marketing is a roadmap capability written in the present tense. It is mechanically detectable — but only once availability is a record with an owner rather than a word somebody chose.

Structure beats volume for AI search

The work that makes content usable by answer engines is mostly the work of having a coherent content model: clear entities, consistent claims, answerable questions. There is no separate optimization product to buy.

Migration-first adoption is why content platforms stall

A platform that requires replatforming before it demonstrates value gets evaluated once. Connecting to what exists inverts the order and makes the migration decision one the customer can make later, with evidence.