Working Surfacelink · 19 Nov 2026
04
Two figures seen from behind sit side by side at one potter's wheel with their hands on the same spinning clay, while a thick closed manual lies unopened on a stool beside them, the whole scene drawn with wide empty margins on every side.
The Beautiful Mess

TBM 452: Contexting vs. Context

John Cutler · 18 November 2026

Company story · collaboration · artifacts · Agent context files

Read the original
Takeaway

John Cutler argues that experts struggle to write down what they know, so teams racing to give AI agents more context may skip building shared understanding.

Summary

John Cutler is a product and organisation coach who writes the newsletter The Beautiful Mess. He describes a product leader who felt tested by their team's questions while finding their own manager hard to read, and traces both to the difficulty of sharing tacit knowledge. He applies this to the current effort to give AI agents more written context.

Key points
  • Experienced people often decide by pattern recognition and struggle to explain how they know what they know.
  • People asking for context often cannot say what they are trying to decide or what they already know, so both sides end up frustrated.
  • People who explain well can look more competent than people who judge well but explain poorly.
  • Cutler's own team abandoned an elaborate effort to give AI agents context after seeing that it was avoiding building shared context itself.
Implication

A team that writes context files for AI agents may need to check them with the people whose judgement they describe. That is an inference and is not the author's claim. Cutler suggests designing AI to take part in how people build shared context, through observation and doing the work together.

Plays
Product

When asking a colleague for context, say which decision it is for and what is already known.

Product

Pair with a colleague on one real problem to see how they judge it, instead of asking them to explain their reasoning in the abstract.

Derived by Working Surface from the article; more in the Playbook. Source line: We were trying to avoid doing the work of developing shared context ourselves.

Source issue

19 November 2026: as AI agents write most of a team's code, the human check moves from reading each pull request to owning the rules on which changes a person reads.