Cognition
Research into cognitive organisation, regulation, behaviour and variation, including work on ADHD and broader cognitive architecture.
Explore Cognition →Riff Systems · Founded by Tristan Radford
Cross-Domain Cognitive Systems Architect
I develop large-scale conceptual architectures, research frameworks and practical tools for investigating complex systems across domains.
My work spans cognition, human–AI behaviour, systems architecture, adaptive organisation, empirical research and applied tool development. Only a selected portion of the wider research programme is public.
Public Research Surface
A curated view of a much larger programme
This site presents selected frameworks, papers, empirical investigations, domain work and tools from the broader Riff Systems research programme that are ready for public examination.
Much of the underlying body of work remains private by design, including unfinished research and IP-sensitive systems intended for later development or release.
Research Topics
Start with the question, not the framework name.
These gateways connect broader questions in established research areas to the relevant Riff Systems papers and empirical results.
01
Adaptation, Resilience & Complex Systems
When does persistence, recovery or ecological memory support a stronger claim of adaptive modification?
02AI Representations & Interpretability
What can decodability, activation intervention and prospective interpretation each establish?
03Cognition, Stress & Behavioural Regulation
How should cognitive load, offloading and stress-like instability be described without collapsing distinct mechanisms?
04Cross-Domain Causal Methods
Declared boundaries, frozen criteria, strong nulls, causal tests and preservation of negative results.
Selected Work
Public entry points into the programme
These areas are representative, not exhaustive. They show how the work moves between cognition, systems, interaction, empirical research and practical implementation.
Human–AI Systems
Research into behavioural, cognitive and relational structures that emerge through sustained interaction between humans and artificial intelligence.
Explore Human–AI Systems →Systems Architecture
Cross-domain architectures, ontologies, structural models and relational methods for representing and comparing complex systems.
Explore Systems Architecture →Adaptive & Coherent Systems
Selected work on coherence, regulation, persistence, retained modification and adaptation across different substrates.
Explore Adaptive Systems →Empirical Research
Tests designed to expose formal ideas to real data, conventional explanations, uncertainty, falsification and explicit failure criteria.
View Research →Tools & Applied Systems
Practical systems and research instruments derived from parts of the wider architecture. Some remain private while development and IP strategy are ongoing.
Explore Tools →How I Work
Architecture is only useful if distinctions remain meaningful when the domain changes.
Find the structure
Start beneath the disciplinary label: boundaries, relationships, organisation, behaviour, coherence, identity, pressure and change.
Make it explicit
Convert intuition into definitions, frameworks, models, diagnostics or formal decision rules that can be inspected.
Test what can be tested
Freeze important choices, use strong conventional explanations, report negative results and distinguish prediction from causation.
Build what can be used
Translate useful architecture into methods, software, research instruments and applied tools where the work supports implementation.
Empirical Research
Ideas should encounter something capable of disagreeing with them.
The empirical strand is not the whole research programme. It is one way I test whether selected components remain coherent when applied to real systems.
NEW · Empirical publication · Forest F1
Testing Adaptive Modification in Forest Systems
Forest F1 applies one frozen ACF classifier across six heterogeneous forest systems to test whether apparent adaptation can be distinguished from resilience, ecological memory, retained damage, filtering, ordinary recovery and continuing external forcing.
The cases reached different evidential depths under the same six-gate sequence. No completed case currently establishes forest-level adaptive modification, while one preregistered case remains unexecuted because the decisive raw data are not publicly accessible.
Current result
No completed case establishes forest-level adaptive modification.
Publication
Forest F1 · v1.0 · 4 October 2026 · interim results
The result is discriminative rather than simply negative: different forest systems stop at different causal gates without changing the ACF ontology or evidential rules.
Recent empirical publication · NBS-001 / NBS-002
Testing Structural, Causal, and Interpretive Claims in Non-Biological Semantics
A controlled empirical NBS study tested whether a TRAIN-derived evaluative-valence direction could survive structural controls, exert a selective causal effect, and support a prospectively specified human-readable interpretation on fresh data.
NBS-001 was rejected after a tokenisation confound was detected. NBS-002 then passed the controlled structural and causal stages, while the frozen fixed-coordinate interpretation criterion failed on independently generated descriptors and templates.
Confirmatory endpoint
STRUCTURAL_PASS → CAUSAL_PASS → INTERPRETATION_FAIL
Publication
NBS-001 / NBS-002 · v1.0 · 26 September 2026
The result preserves the separation between structural decodability, causal efficacy of an imposed displacement, and transfer of a fixed human-readable coordinate. The failed interpretation criterion remains the frozen confirmatory endpoint.
Completed empirical test · Titan T1
Testing Adaptive Modification in Titan's Atmosphere
The first completed empirical test applied one substrate-independent diagnostic from the wider systems work to a real planetary dataset using Cassini/CIRS observations.
An initially promising HCN prior-state signal survived several intermediate analyses before failing the strongest seasonal, regional and circulation controls under strict out-of-sample validation.
Result
Adaptive modification was not established.
Classification
History-dependent regulatory planetary atmosphere.
The purpose was not to prove Titan adaptive or alive. The value of the study was that the diagnostic produced a non-trivial result while preserving the distinction between complex physical memory and adaptive modification.
Future independent test
Titan T2
Atmosphere + surface methane/hydrocarbon boundary. T2 will be independently specified and cannot retrospectively alter the Titan T1 result.
Future independent test
Titan T3
Atmosphere + surface + relevant subsurface boundary, independently preregistered and tested when the required evidence and methods are ready.
Research Library
Formal work, with status and provenance
Public documents are separated by type and status so that completed results, developing frameworks and exploratory work are not confused.
NEW · Empirical ACF study
Testing Adaptive Modification in Forest Systems
Forest F1 · six-case empirical stress test · interim results.
Behavioural framework
The Dual-Channel Cognitive Offloading System (DCCOS)
A behavioural framework for presleep cognitive offloading and dream-state hyperprocessing in ADHD.
Foundational ACF framework
The Adaptive Coherence Framework
Definitive consolidated conceptual/formal paper.
Empirical test
Testing Adaptive Modification in Titan's Atmosphere
Titan T1 · atmosphere-only boundary.
Conceptual paper
Stress Without Stress
Why AI Behaviour Under Load Can Resemble Human Cognitive Instability.
Empirical NBS study
Testing Structural, Causal, and Interpretive Claims in Non-Biological Semantics
NBS-001 / NBS-002 · controlled structural, causal and interpretation sequence.
Foundational NBS paper
Non-Biological Semantics (NBS)
A proposed field-level research programme for the study of semantic organisation in artificial systems.
Research Principles
Public work should be inspectable without pretending certainty.
Separate proposal from evidence
What I propose, what the data support and what remains unknown are not the same thing.
Freeze important choices
Where practical, boundaries, variables and failure conditions are specified before decisive tests.
Use strong conventional nulls
Interesting results should survive simpler explanations before stronger interpretations are accepted.
Keep negative results
A method that can only confirm its creator is not being tested seriously.
Preserve provenance
Corrections, versions and superseded work should remain traceable rather than silently overwritten.
Curate the public surface
Publication is selective. Some work remains private because it is unfinished, sensitive or IP-relevant.
Tools & Applied Systems
Not every useful result should end as a paper.
Parts of the research programme are intended to become practical systems for analysis, reasoning, interaction, research and applied decision support.
Some tool development remains private and IP-sensitive. Selected tools will be released publicly when they are technically mature and strategically ready.
Future Tools →01Problem
→
02Model
→
03Method
→
04Implementation
→
05Evidence
→
06Limitations
Collaboration & Research Support
Serious questions are easier to answer with the right expertise.
I am open to conversations with researchers, technical specialists, universities, laboratories, funders and organisations where there is a clear fit with selected public work.
Areas of potential collaboration include independent replication, domain-specialist review, data access, computational modelling, research tooling and tightly scoped empirical studies.
I am particularly interested in partnerships that contribute specialist expertise, technical capability or infrastructure through clearly defined areas of collaboration, while allowing the underlying private research architecture to remain protected.
About
Riff Systems
Independent Cross-Domain Research & Systems Development
Tristan Radford is a Cross-Domain Cognitive Systems Architect and independent researcher. He develops behavioural frameworks, causal methods and empirical tests across cognition, adaptive systems, artificial representations and complex systems, using cross-domain comparison to separate superficially similar behaviours by their evidence and underlying mechanism.
Read the Full About Page →About the Programme
What this public hub does — and does not — represent
Is this the full research programme?
No. Riff Systems is the public surface of a broader body of work. Some research remains private because it is unfinished, sensitive or IP-relevant.
What empirical work is currently public?
Public empirical work now includes Titan T1, the NBS-001 / NBS-002 controlled valence study, and Forest F1, a six-case empirical stress test of the Adaptive Coherence Framework. Titan T2 remains a future independently specified test.
Is the Adaptive Coherence Framework the main focus?
No. It is one formal component within a much larger cross-domain systems architecture. Its life/adaptation diagnostic is a small application inside that work.
Public Research Record
Selected milestones
04 OCT 2026
Forest F1 empirical stress test published
Six heterogeneous forest systems were evaluated under one frozen ACF gate sequence. No completed case currently establishes forest-level adaptive modification; different cases stop at different evidential gates and one preregistered case remains unexecuted because decisive raw data are not publicly accessible.
PUBLISHED · INTERIM RESULTS · FOREST F1 v1.0
26 SEP 2026
NBS-001 / NBS-002 empirical study published
Controlled structural, causal and interpretation sequence completed in Pythia-70M. NBS-001 failed its tokenisation control; NBS-002 passed structural and causal criteria before the frozen fixed-coordinate interpretation criterion failed on fresh data.
PUBLISHED · STRUCTURAL_PASS · CAUSAL_PASS · INTERPRETATION_FAIL
17 AUG 2026
Titan T1 completed
First completed empirical test. Candidate HCN prior-state signal identified, challenged with progressively stronger controls and rejected as evidence of adaptive modification.
COMPLETED · RESULT RECORDED · T1 FROZEN
ONGOING
Private and public programme development
Multiple research, architectural and tool-building programmes continue beyond the work currently released here.
ACTIVE