EXP-P1
Minimum sufficient invariant substrate
Minimum sufficient invariant substrate under controlled informational budgets.
Research candidate
Programme scope only; no experimental outcome is claimed here.
Invariant Research Lab
The science of reliable intelligence.
The Invariant Research Lab investigates whether intelligent systems can expand their effective capability by discovering, testing and preserving reusable invariant structure while maintaining legitimate constitutional boundaries.
Research process
Falsification over assertion.
Candidate invariants must survive comparison with baselines, adversarial testing and explicit falsification criteria. This is the research method, not a claim that every programme has completed it.
- 01
Discover
Identify candidate invariant structure and the conditions under which it might be reusable.
- 02
Hypothesize
State a testable prediction and explicit criteria that would falsify it.
- 03
Control
Define informational budgets, comparison baselines and controlled conditions.
- 04
Test
Probe predictions through controlled and adversarial experiments.
- 05
Evidence
Record provenance, methods, limitations and negative results for reproducibility.
- 06
Falsify / Refine
Reject unsupported claims or refine the hypothesis and repeat the investigation.
Experimental programmes
Three lines of inquiry.
Programme A
Constitutional computing
How rights, policies and delegated authority can be represented and enforced within computational systems.
Programme B
Intelligent agent systems
Reliability, coordination and accountability in single and multi-agent systems undertaking consequential tasks.
Programme C
Financial & economic computation
The relationship between computation, cost and value — and the reliability required for financial operations.
EXP-P1–P4
Named programmes. Bounded claims.
The programme descriptions define areas of inquiry. They do not stand in for experimental records, disclosed results or reproducible evidence.
EXP-P2
Condition-directed gated verification
Condition-directed gated verification, correctness, effort and repair.
Research candidate
Programme scope only; no correctness, effort or repair result is asserted here.
EXP-P3
Representation-focused investigation
Representation-focused investigation.
Research candidate
No additional methods or findings are inferred from this programme description.
EXP-P4
Reserved future hypothesis
Reserved future research hypothesis.
Reserved · not experimental evidence
No present experimental evidence should be inferred from inclusion in the programme.
Research maturity
Different evidence. Different conclusions.
The Invariant Intelligence scaling thesis remains under investigation. A deployed feature is not scientific validation; a passing experiment is not proof of commercial viability.
- Doctrine
- A stated theory, principle or research thesis. Not an experimentally established law.
- Implementation
- Evidence that a particular tool or component has been built. Not scientific validation of the wider thesis.
- Research candidate
- A hypothesis or programme proposed for controlled investigation; inclusion is not evidence of a result.
- Controlled research evidence
- Findings tied to controlled conditions, baselines, falsification criteria and reproducible provenance. Conclusions remain limited to the experiment's scope.
- Operational evidence
- Observations from real operation, with their context and limitations. They do not automatically validate a scientific thesis or commercial viability.
Publications & research archive
The library.
1 public reference
Article · Published
Invariant Intelligence: A New Path to Trusted Super Intelligence(opens in new tab)Public exposition of the research programme, not a peer-reviewed experimental report or a demonstration of Trusted Superintelligence.
- Research area
- Invariant Intelligence
- Evidence classification
- Doctrine
- Maturity
- Active, falsifiable research thesis
Canonical Threshold 007 Research and Reading editions will be listed only when approved public versions can be linked. No findings are inferred from unavailable editions.
Academic collaboration
Open to rigorous partners.
We seek collaboration with universities and research institutions working on AI reliability, governance, formal methods and computational economics.