A catalogue raisonné without waiting ten years.

Q: What is CultureOS (also known as CulturalOS)?

CultureOS (also referred to as CulturalOS or CultureOS Labs) is an AI-native technology and research practice for cultural institutions, smaller museums, artist foundations, and archives. It helps institutions modernize their digital infrastructure, workflows, and catalogue raisonné processes.

Q: What services does CultureOS provide for museums and foundations?

CultureOS provides three core modernization offerings: Diagnostic (a thorough audit of systems and workflows with an executable roadmap), Digital Transformation (connecting CMS, collections, donor tools, and websites into an integrated modern ecosystem), and Infrastructure Upgrade (cloud migrations, server hardening, automated backups, and institutional knowledge preservation).

Q: What is CultureOS Labs and its pilot with the Tom of Finland Foundation?

CultureOS Labs is the research arm conducting deep technical investigations into catalogue raisonné research and open evidence frameworks. Its founding laboratory pilot is with the Tom of Finland Foundation in Los Angeles, proving AI-assisted archival ingestion and scholarly evidence graphs at scale.

Q: How does CultureOS ensure accuracy and prevent hallucination in cultural archives?

CultureOS implements an Open Cultural Evidence Framework where every AI observation, transcription, and entity match is permanently linked to physical provenance records, source images, and uncertainty metrics. Scholarly authority is never automated; expert human scholars retain sole authority to accept, reject, or verify claims.

CultureOS Labs

Archival material related to this practice

CULTUREOS LABS

pilot

A catalogue raisonné without waiting ten years.

The research arm of the practice. We are building a methodology for the most expensive kind of cultural scholarship, so that smaller institutions can afford it.

Follow the pilot Start a Research Conversation

/labsCultureOS Labs · 2026

THE OLD MODEL

Large research teams. Millions of dollars. Years of manual evidence gathering.

THE DIVISION OF LABOR

Machines prepare. Scholars decide.

Machines prepare

Searching · transcription · matching · classification · bibliography · initial provenance reconstruction · contradiction detection · continuous monitoring

Scholars decide

Authentication · attribution · physical examination · interpretation · ethical judgment · final inclusion · publication

IN TRANSLATION

Two languages. We speak both.

The sentence is for the director. The line beneath it is for the engineer. Standing between the state of the art and the institution, translating in both directions, is most of what we do.

It reads the archive.

Born-digital records, scans, handwriting, artwork photography, catalogues, invoices, conservation records, auction data, and bibliography are read into one searchable body of evidence.

Multimodal ingestion, OCR and handwriting recognition, layout analysis.

It figures out what belongs together.

People, artworks, galleries, institutions, places, exhibitions, publications, and ownership events are reconciled across historical records.

Entity resolution, temporal knowledge graph, Linked Art compatibility.

It recognizes the same artwork wherever it appears.

Duplicate scans, alternate photography, recto and verso, studies, variants, reproductions, and potentially related works are identified across the corpus.

Cross-modal embeddings, image retrieval, perceptual hashing.

It reconstructs where the work has been.

Ownership, exhibition, consignment, acquisition, and transfer events become temporally ordered, evidence-backed provenance graphs.

Event extraction, uncertainty intervals, source-weighted evidence.

Many investigations run at once.

Research agents independently pursue chronology, provenance, bibliography, exhibitions, archive references, visual relationships, and conflicting evidence.

Parallel agent orchestration, recursive investigation, model routing.

It tries to prove itself wrong.

A dedicated agent challenges every high-consequence assertion by searching for counter-evidence, alternative readings, and source inconsistencies.

Adversarial verification, contradiction mining, multi-model checks.

Every answer can explain why it believes something.

Claims stay connected to supporting and contradictory evidence, confidence, provenance, version history, and human review status.

Assertion-level evidence architecture, source traceability.

Scholars make the decision.

Experts review high-consequence conclusions with sources, contradictions, confidence levels, and reasoning visible in one place.

Confidence-gated escalation, expert adjudication, review provenance.

TWO CATALOGUES

A living catalogue and an authoritative one.

The working catalogue holds uncertainty, incomplete provenance, possible duplicates, hypotheses, and contradictions. It changes every day. The scholarly catalogue contains only what has crossed the institution’s approval threshold. Only that is published as authoritative.

TRADITIONAL

Before

A scholar spends a day or more per work: searching folders, reading correspondence, comparing photographs, checking publications, assembling the case by hand before any judgment can begin.

WITH THE LABS

Prepared by the Labs

Candidate CR-1974-0182 · proposed date 1974 · confidence 0.96 · 17 supporting sources · 1 contradiction · 4 related works · provenance 1974–1988 complete, 1988–1992 unresolved · estimated human review: 6 minutes.

Illustrative. Identifiers and values are not real. The day-or-more figure is the traditional average used in our model, not a measurement.

TARGETS

Measure the work before making the claim.

50–80%

Target reduction in repetitive research-processing cost

5–10×

Target scholarly throughput

100%

Human authority over consequential decisions

Targets until validated. We will publish measured results, including where automation failed.

WHERE THE MACHINE STOPS

Final inclusion. Authentication. Interpretation. Legal conclusions. Unobserved physical facts. Ethical and access decisions.

If no person or instrument has examined the object, the system may not assert its paper, dimensions, watermark, or condition. The machine prepares the evidence. The decisions remain human.

THE INVISIBLE COLLECTION

A catalogue reveals the collection nobody knew existed.

When a serious catalogue is announced, works in private hands surface, with invoices, photographs, correspondence, and exhibition history. Collectors, museums, dealers, and scholars all have reasons to contribute. Money can support research. Money can never buy inclusion or attribution.

THE PILOT

Begin with one hundred works.

Eight to twelve weeks on a defined body of work and its archive. Labor measured before and after, accuracy against known records, correction rates, review time. Our founding laboratory is Tom of Finland Foundation.

OPEN FRAMEWORK

Some infrastructure should belong to the field.

The evidence model, the human-review protocol, and the evaluation method will be published openly. Production systems and institution-specific implementation remain professional work. Open the protocol; commercialize the execution.

Open framework → /labs/open-framework

START HERE

Working on the same problem?

Provenance, digital humanities, evaluation of AI systems. We would like to compare notes.

Start a Research Conversation Book a Diagnostic