ARTSSN Research Proof Desk

Catalyst Grant 2026 Application

Kilian Liptrot | Founder, Blinding Visuals Press LLC and ARTSSN

Mainz, Germany | kilian@blindingvisuals.com | https://artssn.com

Requested funding: £25,000 over six months. This document describes proposed work, not a funding award or completed research.

1. The problem

An editor needs to know whether a manuscript's visual evidence is ready for publication. The relevant material can be scattered across figure files, source records, credit lists, permission statements, email and successive manuscript versions. Checking that a figure still corresponds to its cited source, that its attribution survived revision, and that unresolved permissions remain visible is a repeated coordination task rather than a single text-generation problem.

Independent publishers and smaller research teams cannot simply add another specialist to every review. The risks include publishing the wrong version, losing attribution or accepting an unsupported provenance assertion. I know the operational fragmentation from producing a photography book and building the publishing infrastructure that followed. The proposed research would measure the burden in a defined editorial workflow; I am not claiming an established quantitative baseline.

2. Your workflow

Research Proof Desk would begin when an editor submits a manuscript and its authorized supporting files, or uploads a revision. The agent would identify figures and attributed claims, retrieve their available source and identifier records, compare attribution and version continuity, and assemble an evidence packet. Missing sources, inconsistent credits and unverified rights assertions would become explicit review items rather than silent passes.

An editor would inspect each finding beside its supporting record, correct it or request additional evidence, then approve an export tied to the exact manuscript revision. A changed revision would require a new decision. The agent could plan checks, retrieve authorized records and prepare the packet. It could not publish, certify copyright ownership or override the editor.

The prototype would extend ARTSSN's existing publishing and image-identity infrastructure. Import and export would initially use ordinary manuscript, image and structured-record files, avoiding a requirement that a publisher replace its entire editorial system.

3. Trust, audit and governance

The interface would distinguish a contributor's recorded assertion from independently corroborated evidence. An identifier or fingerprint can establish continuity of a recorded file; it does not prove that the contributor owns its rights. That distinction would remain visible in the exported packet.

Each finding would retain its source, retrieval time, relevant version and action outcome. Uncertainty would be named. Unsupported conclusions would become requests for human review. Access would be limited to material authorized for the project, original files would remain unchanged, and approval would bind to the reviewed revision. Editors would retain responsibility for publication decisions and could reject or correct any finding. Evaluation would specifically test false confidence, stale evidence and failures to invalidate an earlier approval.

4. Team

I am the sole founder of Blinding Visuals Press LLC, established in 2024, and the builder of ARTSSN. I am an attorney, U.S. Army officer and photographer. In 2025 I published Y(our) National Parks. The practical difficulty of coordinating that book's production and distribution led me to build a connected publishing platform with AI-assisted development.

My contribution is hands-on publishing experience, legal and operational judgment, and the ability to implement and test a workflow in a system I own. I am not presenting myself as an academic machine-learning researcher. Independent editorial reviewers would be recruited for the evaluation; none is represented here as an already committed partner or adviser.

5. Where you are today

ARTSSN is an existing platform, available at https://artssn.com. Its publishing workspaces and image-identity work provide a starting point. Research Proof Desk itself is a proposed module. There is not yet an academic pilot, validated time-saving result or research-publisher customer for this module.

Months one and two would define the workflow and build a rights-cleared, open or synthetic evaluation corpus. Months three and four would implement retrieval, checking, review and evidence-packet export. Months five and six would compare manual and assisted review using prospectively recruited reviewers.

The corpus would contain seeded missing records, wrong credits and stale versions. Measures would include review time, missed issues, false flags, overrides and recovery from incorrect findings. Results would determine whether to continue, narrow or redesign the approach. Non-sensitive methods and findings would be shared where rights and consent permit.

6. Alternatives and competitors

The immediate alternatives are manual folders and checklists, reference-management software, manuscript systems and general-purpose AI assistants. These are established categories, not absent competition. Manual review offers direct control but requires people to connect records; a general assistant may explain a concern without creating a durable, revision-bound evidence trail.

This proposal is narrower than a universal plagiarism or image-authenticity detector. Its differentiator would be the complete review process: authorized source retrieval, explicit unverified assertions, editable findings and an approval linked to the exact version. Evaluation must show that this combination improves review rather than merely producing a more polished report. The project is at concept stage; no market-share or competitor-performance claims are made.

7. Where this goes

The longer-term opportunity is a portable integrity packet that can accompany visual research from author to editor and publisher. Initial users would be independent publishers and research teams working with substantial visual material. Later applications could include institutional reports and other evidence-heavy publishing.

Editorial workspaces or institutional subscriptions are possible payment models, but pricing would be tested against demonstrated value and willingness to pay. The first objective is an evaluated workflow, not an assumed recurring-revenue forecast. ARTSSN's existing infrastructure makes it possible to test a focused extension without first building an entire publishing system.

8. Fit with Digital Science

This sits between completed research and publication, where evidence, attribution and editorial decisions must remain connected. It addresses the Catalyst call's publication and research-integrity workflows through a multistep agent with explicit human control.

Digital Science's research-workflow expertise would be valuable in defining a useful evaluation, identifying interoperability requirements and challenging assumptions about editorial practice. The desired outcome is a working, assessable prototype with an honest account of its limits, not an autonomous assurance claim.

9. Budget

Request: £25,000 over six months: £10,000 prototype implementation; £6,000 rights-cleared evaluation materials and independent editorial review; £5,000 interoperability and evidence-packet exports; £4,000 accessibility and documentation. This would move the proposed module from concept to an evaluated prototype. Costs would be ring-fenced from existing image-edit research support and any separately funded work.