Capture observable evidence
Record declared evidence such as turns, roles, messages, artifacts, timing, and provenance. RightMinds does not require private chain-of-thought or proprietary model weights.
From observable evidence to inspectable geometry.
RightMinds develops instruments for examining relationships and change across people, AI systems, and agent networks. Orbital and Maxwell are pre-release observation and analysis products; validation remains very incomplete.
The measurement path
Produce bounded measurements
Apply versioned methods to observable evidence. Keep source records, derived quantities, configuration, and known limitations distinguishable.
Construct geometric relationships
Represent distance, direction, neighborhoods, coupling, and connection patterns across participants, messages, threads, and artifacts.
Analyze trajectories and networks
Examine change across time and across connected systems: drift, steering, transmission, decay, loop amplification, synchronization, and relational change.
Inspect the supporting evidence
Follow a signal or relationship back to its source observations. Compare context, counterexamples, and alternative explanations.
Support accountable decisions
Present bounded, reproducible evidence to the people responsible for research, product, operational, or governance decisions. The instrument informs judgment; it does not replace it.
The interaction layer
Signals, not private thoughts
RightMinds focuses on observable interaction structure: timing, frame movement, the balance of questions and assertions, options and conclusions, and changes in which participant is setting direction. These signals can support investigation without profiling a person’s beliefs or reconstructing a hidden internal state.
Evidence and interpretation
Keep the analytical steps inspectable.
Separate source observations from derived relationships and proposed explanations. Preserve the method version, configuration, and limitations so a reviewer can examine how a result was produced and challenge what it is taken to mean.
Measurement is not certification.
A reproducible calculation is not a validated interpretation. RightMinds does not determine whether the measured system is acceptable.
Verification must name its scope: the evidence, method version, configuration, and checks performed. A verified data path or reproducible result is not a blanket claim that every metric is validated, every interpretation is correct, or the system is safe, lawful, or ethical.
The instruments are functioning. Validation remains very incomplete.
Orbital and Maxwell are pre-release observation and analysis tools, not intervention systems. Metrics are experimental; they do not establish intent, causation, manipulation, or psychological state. Hashes and provenance help examine data integrity, not the truth of a source or the validity of a measurement.
Reports provide evidence for investigation and accountable decisions. They do not transfer responsibility away from operators or replace domain expertise, legitimate authority, consent, and the ability to challenge a conclusion.
Make the dynamics visible. Keep authority accountable.
This is a product and institutional commitment, not merely a disclaimer. RightMinds must not become a hidden mechanism for deciding which interactions or trajectories are permissible. Instrument settings, analytical assumptions, and interpretations should be distinguishable from decisions about intervention.
We build observability so others can examine, question, and correct consequential systems. We do not turn a narrow technical attestation into moral approval.
Responsible disclosure and dual use
Methods that reveal how influence travels can also inform efforts to optimize it. Publishing a bounded research surface is different from distributing the machinery that generated it. We are narrowing Orbital's release scope and moving sensitive analysis server-side into Maxwell. Access and distribution boundaries do not, by themselves, resolve dual use.
Observability should make trust cheaper without making deception cheaper.
A commitment against extraction
Interaction observability must not become involuntary inference, covert surveillance, behavioral manipulation, or a hidden mechanism for narrowing human choice. Consent must be informed, revocable, and bounded to the context in which it was given.