A place to test explanations
Psyche Lab is a deterministic research environment for synthetic histories. It is designed to support execution, replay, ablation, and counterfactual branching so that a proposed mechanism can be examined rather than merely described.
Its core simulator is separate from Luna. It does not require private Luna identity data or a language model to execute its synthetic histories.
Replay
Begin with declared starting conditions, inputs, and a recorded evolution. Replay asks whether the same process reproduces the same history. A mismatch can expose an incomplete record or an undeclared dependency before interpretation runs ahead of the instrument.
Exact replay is a property of an execution under specified conditions. It is not, by itself, a measure of intelligence or subjective experience.
Intervention
An intervention changes a declared part of the system or its conditions. A useful comparison preserves what should stay the same, identifies what changed, and examines unintended effects.
Removing a component can change many things at once. A convincing interpretation needs appropriate controls and simpler alternatives, not just a before-and-after difference.
Counterfactual histories
A counterfactual branch explores what follows from a different declared condition. The common history and point of divergence must remain clear.
This creates a way to investigate causal explanations in a bounded setting. It does not make an artificial history equivalent to a lived human history, or turn a synthetic mechanism into a demonstrated conscious being.
Keep the result that disappoints
An invalid endpoint, a null difference, and a negative finding tell different stories. A defective instrument does not support the same conclusion as a valid experiment that contradicts a prediction.
The journal includes a limited framing pilot in which a prompt-parser mismatch invalidated an endpoint. The correction was made for future use; the original result was not rewritten.
Public explanation, protected internals
The public Observatory is a separately authored educational exhibit. It explains replay and perturbation through a small model whose behavior you can inspect in your browser.
It is not the Psyche Lab engine. Proprietary implementation details and private research records remain outside the website. Contact Neurasoft to discuss a specific method or review request.
Method descriptions are based on the maintained Psyche Lab documentation. No lab experiment was rerun as part of producing this website.