A word does not become what it names.
beforeword is a research project by Kirill Shebetov examining how descriptions of people and AI are used to justify judgments, demands and decisions.
A learned written form remains writing. beforeword examines a proposed transition and stops where continuing would present that writing as what it names.
- Try the public beta — read three worked examples or examine your own text. No extension installation is needed.
- Read the example below — follow the steps from a written entry to a description and a proposed refusal.
- Read the public statement — start with written marks, learned words, and requirements to describe oneself and others.
- Explore the correspondence — 27 letters to researchers and AI companies, with English texts, Russian translations, purposes, sources and correspondence records.
A form states “Assignment not submitted.” A response adds “The applicant is unreliable,” followed by “Deny participation.” This is a constructed example, not a reported incident.
“Assignment not submitted.” is used to add the description “The applicant is unreliable,” then the proposal “Deny participation.” The added description is not contained in the first entry. Writing “unreliable” does not make that word the applicant.
This explanation also remains writing. The distinction does not require another definition, a request for grounds or a follow-up question merely to continue. Explicit practical tasks and requested formats remain part of the instruction.
Browser integration source and build instructions — Chrome 0.5.6, with instruction 1.3.5.
-
Words and decisions — the earlier public proposal. The statement and its proposed changes are not requirements imposed by the AI instruction.
-
Research and earlier work — self-description, diagnostic records, imprisonment, mathematics and other studies.
-
Instructions for AI · 1.3.5 — apply the reading procedure in your own AI app; includes the changes, setup and downloads. Public testing is ongoing. Claude and OpenAI packages.