One history · many sources
The Knowledge Project
Turning years of proceedings, emails, registries, decisions and economic records into an auditable history: what we know, why we know it, what is disputed and what is still missing.
The corpus
The problem is not only finding documents. It is preserving continuity.
The relevant archive is distributed across email, Google Drive, OneDrive, judicial and administrative records, registries, accounts, valuations, professional correspondence and research material. It contains duplicates, versions, repeated annexes and derivative documents.
Preliminary planning estimate: the unique relevant textual corpus may represent roughly 35–120 million token-equivalents, with a central working estimate near 65 million. This is not an audited census and will be revised as deduplication progresses.
AI transparency
We have not developed our own AI model.
We rely predominantly on ChatGPT and other available tools to search, compare, classify and analyse material. ChatGPT is not a witness, judge or source. An AI-generated conclusion is not evidence: the underlying document is the evidence.
Our contribution is the accumulated corpus, human knowledge, evidential methodology, chronology, classification, questions and the progressive construction of a traceable knowledge base.
Method
Source → proposition → event → decision → consequence.
Every material proposition should be traceable back to its source. We distinguish documented facts, party positions, inferences, disputed matters, items under verification and document gaps. Adverse evidence and alternative explanations are preserved rather than hidden.
The objective is not for AI to decide who is right. It is to reduce the loss of context that occurs when one continuous history is divided among many proceedings and institutions.
Publishing project · Fundación Por Derecho
Reason to Believe · Razón para creer
One Asset. One History. The evidence repository reconstructs the documentary history. The book tells the human and institutional story behind it, and asks what one exceptionally complex case may teach us about effective judicial protection, legal memory, the rule of law and responsible use of AI.
The book is being developed while the underlying history continues to evolve. Material factual propositions are intended to remain traceable to the project's evidential record. It is an associated publishing and educational project of Fundación Por Derecho; that description does not by itself assign copyright, royalties, translation, audio, film or other commercial rights to the Foundation.
Read the story. Examine the evidence. Test the conclusions.
Publication rights, royalties and other commercial rights are separately governed. A book purchase and a donation to Fundación Por Derecho are distinct transactions.
Resource programme
The 100 Million Token Challenge
100 million is a prudent planning ceiling, not a consumption target. The proposed programme is staged: a 10M proof phase, expansion to 25M, a usable system at 50M, and movement toward 100M only if measured knowledge gaps justify it.
The primary metric is verified reusable knowledge per million tokens: canonical sources, verified propositions, resolved contradictions and avoided reprocessing.
Public interest and commercial pathway
Fundación Por Derecho and Havidia LLC have different intended roles.
Fundación Por Derecho is the public-interest pathway: research, education, methodology, access to justice and academic collaboration. Donations and grants should support its public-interest purposes and are not investments in Havidia or interests in private recoveries.
Havidia LLC is the intended US pathway for exploring commercial applications, partnerships, services and possible future technology development. We do not claim that Havidia has developed its own AI model or owns intellectual property that has not yet been formally assigned.
The relationship between the two should be governed by financial separation, documented data and IP rights, publication independence and explicit conflict policies.
Participate
We are not asking people to believe us. We are asking for help to test, preserve and structure the record.
We welcome academic collaboration, legal review, engineering, AI/compute support, research, public-interest funding and clearly separated commercial conversations.