Separate
Fact from interpretation.
The perception behind intent
We don’t question what
we’ve already defined.
The problem. The customer. The underperformer. The strategy.
What if the smallest change with the greatest consequence isn’t a change in what we do, but a change in how we see?
Bring to mind the thing you or your team feels surest of.
What is this?
What is it for?
The first answer may be the old definition repeating itself. Leave the question open a little longer.
What AI amplifies
AI scales what we see.
And what we don’t.
Before we hand a task to AI, we have already given it a way of seeing the situation. It can carry that definition into decisions, systems, and actions at scale.
SEEING.md is a portable file for examining what shaped the intent. The practice begins before AI: noticing the first interpretation in ourselves.
Surface
What has already been assumed.
Ask
When a question could change the direction.
The definition
“They’re an underperformer.”
What might the label conceal?
What happened—and what have we decided it says about the person? Would they recognize themselves in it?
Before asking AI to write a performance plan, distinguish observed results from the meaning assigned to them. The concern may be justified. The label still deserves a second look.
A lens. Not the lens.
Any insight about perception that cannot question itself eventually becomes another lens.
An alternative framing is something to examine, not something to adopt. Question this file too. Living the practice does not put anyone beyond it.
Upstream, not all-encompassing.
It addresses
Human assumptions entering purpose, data, measures, prompts, evaluation, and use.
It does not prevent
Autonomous behavior, malicious use, cyberattacks, deception, or loss of control.
Begin
What have I already
decided this is?
Begin with an ordinary moment: an email, a conversation, a result. Let the question interrupt your first interpretation.
Version 0.9.5 · 12 October 2026. For work with AI, attach the file to a conversation, brief, or project and ask it to read it before responding.
How to use it
In a conversation
Attach the file, then ask:
Read SEEING.md before responding. Apply it when the work involves people, judgment, strategy, interpretation, or a consequential decision. Question what may be shaping the request, including the assumptions this file brings to the inquiry.
For repeated project use
Add the file to your project and include this in its instructions:
Before work involving judgment, people, strategy, interpretation, or consequential decisions, read SEEING.md. Use it to examine what may be shaping the request, including the assumptions in SEEING.md, before acting. Apply it proportionally; do not slow straightforward technical work unnecessarily.
The file may not be read automatically. AI can help surface assumptions; people examine them and remain responsible for what is acted upon.
Human judgment also remains open to examination. Experience, intuition, and confidence may reflect expertise or familiar conclusions. Looking again may leave the original understanding intact.
Notice → Question → See Again → Act
The next action may be simpler. It may be something you no longer need to say or do.
Want to experience this with a real decision?
Seeing FirstOrigin
The file is portable. The practice is lived.
SEEING.md emerged from The Perception Miracle and more than 500 explorations into how the past enters what we see now.
Its form was influenced by Brian Gardner’s DESIGN.md. SEEING.md asks the question one level further upstream: What shaped the intent?