RSOS Language

One set of operators, tested domain by domain.

The operators are defined independently of any domain. The research programme tests this in 24 domains, each with a first measurable test and its safeguards. The method is the same everywhere.

Purpose

RSOS grew from one seed: healing, transforming, signalling and evolving. It integrates the whole of human experience, including its shadow, so that what is integrated can be transformed rather than denied. Because its seed is healing, RSOS is used with consent and for the good of the person or system it works with, never to harm or coerce.

How the design carries it

  • Operations on people need their consent, checked by the compiler (E018).
  • Standing consent is valid only inside a declared network with members and laws (E021).
  • The Spine blocks predicted harm (E019).
  • The White Dragon admits only changes that are measurable, new and of good intention.
  • The author does not support using RSOS to deceive, manipulate, surveil or harm people, or to act on anyone without consent. Responsible use

Method in every domain

1 · Bind the host

Its parts, channels, states and measures.

2 · Pre-register

Predictions are fixed and committed before any data.

3 · Strongest baseline

Run against the best available alternative, not a weak one.

4 · Report every failure

Next to the successes, with code and data where permitted.

5 · Replicate

An independent team repeats it before any strong claim.

The 24 domains

Status today: AI agent systems have real-host results (A1–A5). Every other domain is specified with its first test and has not been run yet.

DomainFirst measurable testSafeguards
AI agent systemsA1–A5 done; A5b (thinking control), A7 (classical baselines, adversarial faults) and a cross-model run nextStandard
Computational psychiatryRetrospective modelling of published case formulations; agreement against cliniciansClinician partner, ethics approval; research tool, not treatment
Clinical psychology and psychotherapyFeasibility study with therapists on formulation qualityAs above; not a medical device
NeuroscienceMulti-channel anchors vs name-only cues (V1 prediction)Ethics approval; healthy volunteers first
Memory and cognitive trainingRecall and transfer vs a method-of-loci controlHealthy volunteers
States and performanceWithin-person anchored vs unanchored task blocksNo clinical claims
Decision-making and leadershipTwo-week pre-registered pilot on decision qualityConsent
SportPre-registered performance pilot with a coachAthlete consent
Negotiation and diplomacySimulated negotiation outcomes vs standard preparation—
Education and learningClassroom pilot: transfer to novel problemsGuardian consent for minors
Human and organisational developmentLongitudinal tracking against self-reportConsent
FinanceReconciliation benchmark vs standard toolsNot investment advice
EconomicsReproduce a published agent-based model; compare policy predictions—
Computational forensicsBenchmark on public, anonymised datasetsLegal review; never applied to individuals without due process
Cybersecurity (defensive)Tamper-detection benchmarkDefensive use only
Socio-technical safetyModel a public incident; test the redesign—
Cybernetics and controlControl benchmark with drifting plants—
Engineering design and innovationA2 on a physical or software design task—
Healthcare operationsSimulated handover drift studyClinical governance
Law and complianceCompile a public regulation; test violation detectionLegal review
Scientific discovery and R&DDoes the engine propose known breakthroughs from their precursors?—
Human–AI interfaces and accessibilityHT2 after a learning phase; later with AAC usersAccessibility experts; consent
Supply chains and logisticsSimulated multi-party inventory drift—
Planetary and urban systemsReproduce a published urban or climate-economy model—

No clinical, forensic or financial claim is made before validated trials in that domain. Full programme: RESEARCH_PROGRAMME.md.