I Built a Skill That Learns My Voice by Arguing With Me
A Claude Code skill called /unslopify rewrites AI text against a brand guide, then learns from my edits until the guide gets good enough to write without me.
Thoughts on building specifically for the AI age. Process over code.
A Claude Code skill called /unslopify rewrites AI text against a brand guide, then learns from my edits until the guide gets good enough to write without me.
A sprint board reading 100% completion told me nothing about whether the work actually held up or mattered at all. The gap between motion and verification is where AI adoption breaks.
This is a reaction to Dan Koe's challenge of writing in a more authentic way in the age of AI.
We talk AI is saving our time on one hand, while we praise 16h/d grinds on the other..
My Monday planning went from four hours to two with AI. Then the agent broke, I did it manually in 90 minutes, and realised the savings were never AI's. The subjective win is real anyway.
A story about how buying a laser printer, which was heavily overpriced at that moment, didn't make me a graphic designer.
So imagine you want to have something like a digital twin.
The highest leverage use of AI is often not shipping code. It is turning personal chaos into a small set of solvable problems, measured inputs, and calm protocols.
AI isn’t green. But you can reduce waste by avoiding unnecessary inference.
A daily investment committee that trains taste by forcing one decision at a time.
A flight recorder for cognition: capture state without turning it into a performance.
A repeatable, operator-first blueprint for building a Slack-native executive board: roles, constraints, context design, and the key architectural pivot.
A working system that gives me structured disagreement, not a single polite assistant.
A retrospective on attention, systems, and the year the Operator mindset quietly formed.
Sometimes the best result arrives as an unintended byproduct, if your context is rich and your judgment is awake.
Vague prompts feel fast. They’re often the most expensive way to use AI.
AI didn’t delete anything by itself. It did it because I approved the warning while half asleep.
A calm way to stop AI from guessing.
How to evaluate AI output without relying on vibes.
Why long AI chats decay and how Operators keep context clean.
Why mid-stream prompting smears good work.