What Board Games Taught Me About Working with AI

By: Katie Parrott (Working Overtime column, Every) Source: Every Type: article

Summary

Fill after reading.

Key Ideas Extracted

Fill during processing.

Notes

Partial paywall — ~1,500 words accessible; full article requires Every subscription
Cross-reference: Parrott is adapting Kieran Klaassen’s compound engineering plugin to a writing context — this is a practitioner’s translation of Compound Engineering into a non-engineering domain
The TASTE.md concept parallels the knowledge-capture-as-side-effect pattern: writing preferences accumulate as the system is used

Raw Content

Source: Every / Working Overtime column, February 19, 2026. Partial — paywall after ~1,500 words.

Katie Parrott is a staff writer and AI editorial lead at Every. She writes Working Overtime, a column about how technology reshapes work, and builds AI-powered systems for the Every editorial team.

What Board Games Taught Me About Working with AI The skills I transferred to my writing agent from playing Settlers of Catan

I’d been stuck on trying to build my own writing agent for months when I found myself scanning my board game shelf. Suddenly, the problem wasn’t about AI anymore.

It was the end of Think Week, Every’s twice-yearly retreat where we break to explore possibilities outside the flow of our regular work. The team was in a beach house in Panama, decked out in shorts and sunglasses with palm trees swaying in the background. I was under 10 inches of snow in Ohio, locked in a battle of wills with my dog about going outside.

From my laptop, I watched Austin Tedesco, Every’s head of growth, demo a dashboard [content continues behind paywall]

The teach

[The board game tradition of explaining pieces before strategy — applied to understanding Kieran Klaassen’s compound engineering plugin. Plugin became model for a writing equivalent.]

The framework maps four stages:

Identifying components (pieces)
Understanding possible actions (moves)
Recognizing how moves interconnect (systems thinking)
Defining success conditions (victory)

The writing equivalent:

Created equivalent roles to engineering plugin’s agents and reviewers
Created TASTE.md file encoding writing preferences alongside CLAUDE.md analog
Writing loop: brainstorm, interview, outline, draft, edit (adapted from plan-work-review-compound)
“Each stage serves specific functions; skipping steps produces ‘polished garbage.’”

The two-layer system:

Key problem encountered: AI got confused — Parrott’s (Katie’s) preferences showed up in what was supposed to be the user’s personal profile. The system was supposed to learn your taste. Instead, it was handing you mine.

Solution: split the system into two layers:

Defaults file: opinionated baseline for good writing standards
Taste file: starts empty, fills up over time with individual user preferences

Key insight: “The engine only reveals its flaws when I actually play.”

Victory condition: Each piece of writing makes the next one easier. The system learns the writer’s voice, reducing the gap between intention and expression. The larger victory: learning how to work with AI systems thoughtfully.