A research horizon

The world answers
in consequences.

A digital system can act through files and tools. A robot must also meet weight, distance, contact, and a world that does not reset when a prediction fails.

A history that leaves the future openRecorded encounters remain distinct. Their relevance meets at a present choice, with several possible continuations. This is a conceptual illustration, not a trace of Luna. 01 / RETAIN02 / RECONSIDER03 / CONTINUE An encounter.A present choice.An open future. The record remains. What it means can change.
Continuity, illustrated. Not a resident measurement.
Conceptual field study · not live telemetry

A physical channel is a new responsibility

We see robotics as a possible extension of continuing intelligence, not a substitute for it. A camera can add observations; an actuator can add effects. Neither creates a self by being attached. The research question is how perceptions, constraints, memory, and action become coupled over time.

No Neurasoft robot fleet or general-purpose hardware deployment is claimed here. Physical embodiment is a research direction. Any move toward it must begin with simulation, narrow tasks, suitable hardware controls, and domain-specific safety review.

Where the wider field is working

Google DeepMind’s March 2025 Gemini Robotics introduction describes a vision-language-action model and a separate embodied-reasoning model, with applications to physical manipulation. It also distinguishes high-level reasoning from lower-level measures such as collision avoidance and force limits. That work is an external reference, not a Neurasoft integration or partnership. Google DeepMind · Gemini Robotics (2025 introduction) ↗

A thought experiment: the unfamiliar cup

A hypothetical robot encounters a cup it has not handled before. A good demonstration would not simply show a successful grasp. It would reveal how uncertainty was represented, what observation informed the approach, when the robot requested help, and what was retained afterward.

On the next encounter, we would want to know whether the earlier consequence changed the decision. We would compare it with a simpler controller and test the limits of generalization. A motor success alone would establish neither developmental learning nor subjective experience.

Digital-native before human-shaped

Embodiment need not begin with a humanoid. A bounded computer environment can already impose real resource limits, interruptions, and consequences. Those can be investigated without exposing bystanders to a physical experiment.

Our proposed path is capability by capability: honest sensing, scoped action, consequence attribution, interruption handling, then carefully evaluated physical channels. We would rather make a small action dependable than use a human silhouette to imply capacities we have not measured.

Public editorial synthesis · 16 September 2026. Internal source observations are dated separately; this is not a live operational report. See the reading room and publication standards for source classes.

To carry with you

A body matters when the world can change what happens next.

Follow the question.

Explore the collection

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