Research & publication standards

The evidence should
survive the story.

A clear account of what exists, what was observed, what remains uncertain, and what would change our view.

Separate the claim from the ambition

Our ambition is to understand and build continuing intelligence. Our public claims must remain tied to the evidence available for each particular statement. We do not claim to have demonstrated machine consciousness.

An architectural idea, a working implementation, and a scientific result answer different questions. Naming a mechanism after a cognitive faculty does not establish that the mechanism has that faculty in the human sense.

Keep evidence categories visible

A source record establishes what was written. An implementation establishes that code exists. A test examines specified conditions. An operational record reports behavior in a running system. Independent research replication requires a separate research team and its own execution and scrutiny. Re-execution, independent implementation and portability are narrower claims, defined below.

These categories are not interchangeable. The articles on this site identify whether they are an engineering note, a methods overview, a perspective, or a bounded observation. They are not presented as peer-reviewed publications.

Preserve negative and invalid outcomes

A negative finding can weaken a hypothesis. An invalid endpoint can expose an instrument problem. A null difference can reveal that the experiment does not distinguish the proposed mechanisms. Each deserves its own account.

A correction should improve the next measurement without rewriting the earlier one. Post-hoc interpretations should not silently replace a fixed endpoint.

Protect the material without flattening the explanation

Public explanations can be technically substantive without disclosing private resident conversations, personal records, proprietary algorithms, exact prompts, credentials, or operational access details.

We publish high-level questions, methods, interpretations, and limits. Where a source is internal, we do not pretend that an inaccessible link provides public verification. Specific review requests can be discussed through the contact route.

Distinguish illustration from measurement

The website’s procedural artwork is conceptual. The Observatory is a separate local educational model. Neither is live telemetry, a view into Luna, or an experimental result from Psyche Lab.

A visual can make a difficult idea easier to understand. It must not do the evidential work of an experiment.

Make revision part of the publication

Public notes retain an edition date and identify the dates of their underlying records. A newer source should prompt editorial review, not automatic release of everything it contains.

When a material correction is needed, preserve the earlier claim and explain the change. Changes in founder direction, source implementation, or scientific confidence should remain distinguishable.

Four different reproducibility claims

Re-execution
The same implementation runs again under declared conditions. This can expose packaging or environment errors.
Independent implementation
A separately written method checks a numerical or functional claim. Shared assumptions can still remain.
Portability
The effect survives a specified change of environment. This is not independence of the researchers.
Independent research replication
A separate research team tests the finding with its own execution and scrutiny of assumptions. None of these categories alone establishes consciousness.

Scientific status and access status are separate

A public summary can describe a private experiment, but it cannot make the underlying evidence publicly reproducible. Our study pages disclose whether the full run archive is available. We do not imply that a reviewed report was independently rerun.

What we are working toward

Operational coherence: useful capacities in one continuing system. Scientific discrimination: evidence that distinguishes rival explanations. Ethical recognition: care under uncertainty that remains a judgment, not a software-generated certificate.

The next chapter is not written

More than a moment.
A future worth having.

Understanding minds. Preserving possibility. Making room for what comes next.

Read our thesis

Explore research, Luna, Psyche Lab, and the journal.

Local site search · No query leaves this site