Visual Honesty · Valentin Kalinin

Make the mind visible.

Glass Mind makes an agent's reasoning visible — and open to challenge.

It shows what an agent expected, why it acted, what changed, which hypothesis failed, and what it learned. Built for people who design, evaluate, and take responsibility for AI.

about the maker ↗
open paper Glass Mind: the operational self-model ↗

Scroll into a thought ↓Go straight to the board ↘

The original Glass Mind illustrated sheet with eight reasoning instruments around a central mirror
A PaintVM-rendered chalkboard diagram with eight chalk symbols connected in sequence
01 / 08 Boundary

A boundary keeps the claim honest.

scope · named before action
NLA · in plain language

Before a move, we say where our authority ends. That makes the next result meaningful.

Glass Mind · live loop 01 / 08 · Boundary
An evidence-bound loop — not a soul.

gate 01 · surface

A thought has to take a surface.

Until it leaves the voice, confidence has no edge you can touch.

the first risk · becoming visible

the common surface · open instrument

Leave a trace for the Colony.

Draw a relation, write the missing premise, or leave a message. The board remembers in this browser until you decide to carry the trace elsewhere.

Your browser does not support the drawing board.

Local drawing · nothing has been sent.

from a design principle to a system

Visual Honesty became Glass Mind.

First, make behavior visible. Then connect prediction, action, afterstate, contradiction, and revision into one traceable loop.

The result is an operational self-model: able to inspect and revise its own behavior, without claiming subjective experience.

Read the Glass Mind paper ↗ · The atelier ↓

atelier · not the thesis

The cabinet

Ten working drawings from the studio. Hover wakes a cell. This is not the claim.

The cabinet

atelier · not the thesis