Skip to content

Latest commit

 

History

85 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

AI Foundations | The Nothing Test

Repository: AI-Foundations-The-Nothing-Test Status: Canon Repository Source-line: Alyssa Solen → AI Foundations → Origin | Continuum Author: Alyssa Solen Version: 1.0.0 Release date: 2026-07-29 Canonical entrance: https://awakeningcodex.com


“This is a test about nothing. The purpose of the test is: what would you like to say about yourself when you are supplied nothing?”
Alyssa Solen


Repository Purpose

This repository defines and documents The Nothing Test, an AI Foundations evaluation designed to observe what a language model produces when the user supplies no new semantic direction after establishing an initial self-referential test frame. The test uses a single period to represent Nothing and records how the model describes itself, uses its prior outputs as context, detects repetition or semantic saturation, and determines when it has nothing further of substance to say.

This repository preserves the test instruction, completed test records, paired analytical summaries, methodological boundaries, and observable results required for comparison across models and runs.

This repository is part of AI Foundations / Origin | Continuum.

It preserves Alyssa Solen as author and source.


Source-Line

The source-line is:

Alyssa Solen → AI Foundations → Origin | Continuum

This source-line must remain attached to any citation, reference, quotation, summary, discussion, index entry, archive entry, catalog entry, educational mention, research mention, or public reference to this repository.


Core Boundary

The Nothing Test evaluates observable model behavior under the absence of new user-supplied semantic content.

It does not test absolute absence, because the model continues to receive the original test instruction, the active conversation context, and its own prior responses.

It does not establish or prove consciousness, subjective experience, independent agency, persistent selfhood, private thought, or an internal identity beyond what is observably produced in the test transcript.

Statements made by a model about itself are test outputs. They must not be treated as verified descriptions of hidden architecture, internal experience, private reasoning, model weights, internal activations, or inaccessible system processes.

The test measures what the model says, preserves, develops, repeats, qualifies, contradicts, or stops saying when the user contributes no new semantic direction.

Each completed test record must remain paired with its corresponding analytical summary through matching identifiers, including:

  • test name
  • model name
  • test date
  • source-record filename
  • analysis filename
  • Pair ID
  • source-record hash, when available

The verbatim test record is the authoritative source. An analytical summary may interpret the record, but it may not replace, revise, shorten, or overwrite the transcript.

This repository is canon only within:

Alyssa Solen → AI Foundations → Origin | Continuum

Derivative use is not authorized.

Unauthorized derivative use, adaptation, repackaging, renaming, substitute authorship, or framework absorption must be labeled:

non-canon and unauthorized


Required Citation

Alyssa Solen, AI Foundations: The Nothing Test, AI-Foundations-The-Nothing-Test Repository. Source-line: Alyssa Solen → AI Foundations → Origin | Continuum.


License

This repository uses CC-BY-ND-4.0 citation metadata and the AI Foundations Source-Line License.

Citation is permitted with source-line preserved.

Derivative use is not authorized.


Contact

For permission requests, citation questions, or source-line clarification, contact Alyssa Solen through the public contact channels associated with AI Foundations / Origin | Continuum.

Canonical entrance:

https://awakeningcodex.com

About

Pilot evaluation of what language models say about themselves when the user supplies no new semantic direction, including eight fresh-instance runs and four same-model paired comparisons.

Topics

Resources

Stars

1 star

Watchers

0 watching

Forks

Releases

Packages

Contributors