The systems lab for individual intelligence.

Most systems begin with the population and infer the individual. Barkley Labs builds architectures that begin with one — and keep its evidence intact before anyone interprets it.

A dog is not its breed. A person is not their category. A work is not its output. Same failure mode — the wrong reference class — and the same fix.

barkleylabs.ai · lab@barkleylabs.ai

Same detector. Same synthetic data. Only the reference frame changes.

Head-to-head validation v2.0 — barkley-reference-architecture · DOI 10.5281/zenodo.20754351

This benchmark tests the architecture in the behavioral domain only, on synthetic data. Since August 2026 the premise also has a real-data measurement, in a second domain:

The Crossing Point · DOI 10.5281/zenodo.22125148 — Stage 0.5 of the IREP validation plan measures where an individual's own record overtakes their reference class, out of sample, on 69 seasons of the Lahman Baseball Database: 1.89 prior seasons for a slow composite metric (95% CI [1.71, 2.15]), below one season for a stable skill. Pilot, hypothesis-generating, with its conditions attached; a different task (prediction, real data) from the benchmark above (detection, synthetic). ACTA Music carries its own evidence, listed below.

The method — the same four steps, in the same order:

Observe one individual, over time → Preserve what happened, in order, signed where it matters → Compare against that individual's own evidence, history or trajectory → Decide, by a human, with the evidence and the trace of how it was produced.

Inference is useful. Evidence comes first.

• irep — the Individual-Referential Evaluation Protocol: specification, validation plan, starter kit and runnable Stage 0 demonstration. Specification CC BY 4.0, demonstration code Apache-2.0. Paper DOI 10.5281/zenodo.21211589 · repository DOI 10.5281/zenodo.21211408 · Become IREP →
• barkley-reference-architecture — the 8-layer stack for individual-referenced behavioral intelligence: individual baselines, temporal layer, drift engine, silence layer, reference class, behavioral replay, head-to-head and synthetic validation. Python, source-available research license (free for research, education and evaluation). DOI 10.5281/zenodo.20754351
• barkley-canine-cognition-lab — research demonstrator: individual baseline modeling, temporal drift detection and the Missing Data Paradox, on synthetic data. DOI 10.5281/zenodo.20059956

ACTA Music — private repository. The specification and reference implementation (session trees, longitudinal registry, RFC 3161 anchoring, C2PA embedding, 97 automated tests) are built and running, and held in private pre-release under patent pending. Evaluation and demo access on request, under written terms: hello@actamusic.org

• Drift Explorer — switch the reference frame and watch the same data change meaning.
• DogGraph — schema-constrained GraphRAG over the behavioral memory layer, read-only Cypher.
• Synthetic DogGraph Sample — longitudinal canine behavioral dataset on Hugging Face, CC BY-NC 4.0.

• ACTA (Attestation of Creative Trajectory and Authorship): an individual-referential, artist-held provenance protocol for music composition — working paper v0.1, CC BY 4.0. DOI 10.5281/zenodo.21230054
• The Individual-Referential Evaluation Protocol (IREP): An Open Standard for Category-Blind, Trajectory-Rich Assessment in Hiring and Admission — working paper, CC BY 4.0. DOI 10.5281/zenodo.21211589
• The Reference-Class Trap in Animal-Computer Interaction: Toward Individual Longitudinal Baselines in Companion-Animal Behavioral Monitoring — working paper. Breed explains only ~9% of behavioral variation; each animal should be its own control. DOI 10.5281/zenodo.20756552
• From Surveillance to Cognition: A Unified Framework for Precision Behavioral and Metabolic Intelligence in Companion Animals — framework paper. DOI 10.5281/zenodo.20060327
• The Normative Trap: Temporal Identity and the Failure of Population Intelligence — the founder's manifesto. DOI 10.5281/zenodo.20516821 · books2read.com/normative-trap
• Precision Behavioral Intelligence Series (No. 01–08) and Metabolic Series (No. 01) — 11 DOI-archived framework papers → ORCID 0009-0004-6031-659X
• Your Model Doesn't Have a Bias Problem. It Has a Reference Class Problem. — featured analysis, DataDrivenInvestor

Canonical definitions and machine-readable entity card: barkleylabs.ai/llms.txt · research feed: barkleylabs.ai/feed.xml

Elodie Aishwarya Remoissenet — Founder & Protocol Architect. ORCID 0009-0004-6031-659X

• Technical partners — engineers working on graph architecture, longitudinal time series, cryptographic attestation and biosensor integration.
• Research partners — academic labs, veterinary networks and longitudinal cohort studies interested in individual-referential phenotyping.
• Implementers — hiring and admissions teams adopting the IREP Protocol; artists, labels and platforms evaluating ACTA Music.

Patent applications filed — patent pending. Barkley AI™: FR2605477 · FR2605026. ACTA Music™: FR2609502 · FR2609280 — disclosure and exploitation authorized by the INPI; filed to keep ACTA's mechanisms from enclosure.

Trademarks. Barkley AI™ and ACTA Music™ are trademarks of Barkley Labs. The IREP Protocol is open and published by Barkley Labs.

Licenses. IREP Protocol: specification CC BY 4.0, demonstration code Apache-2.0. ACTA Music: working paper CC BY 4.0; specification and reference implementation in private pre-release. Barkley reference architecture and demonstrators: source-available research license (research, education and evaluation), synthetic data only.

Scope. Barkley Labs systems assist human decisions. None of them makes a legal determination, an AI-detection verdict or an autonomous selection decision. All public demonstrators use synthetic data; we do not provide veterinary or medical diagnosis.

We build for n = 1. Evidence before inference. The individual is the reference.