Selection Lab — Soul Engine™: Human-Grounded AI Reliability Infrastructure

I have spent 22 years documenting people.

Photography, video, interviews, documents, conversations, institutions, families, vulnerability, achievement, conflict and change over time.

I did not create this archive for AI. I created it because I believed these human stories were important and should not disappear.

But now I think its purpose became bigger.

AI is developing extremely fast. At the same time, more and more of what exists online is synthetic, repeated, copied or generated from something that was already generated before.

I think now is the time to do something about it.

If we don't, synthetic AI content and influencers may eventually become stronger than the real human signal underneath them.

And I don't think the solution is simply more data.

The missing layer is human vision.

You can put every Van Gogh painting into AI and you still don't have Van Gogh.

You have the result of his vision. You don't have the person who could look at the world and understand what was important enough to select.

Human culture was built by people with vision, talent, judgment, achievement and meaningful public work. I believe people like this need to participate in building the human layer of AI.

This is where my archive becomes useful.

It contains not only 22 years of material. It contains 22 years of human selection.

What I photographed. What I did not. What I returned to years later. What I connected. What I recognized as important inside enormous amounts of ordinary human reality.

I don't believe we need to reduce this ability to some perfect formula first.

I want to test whether AI can learn from the record of human selection itself.

For me the process is:

raw material → selection → composition → meaning

And selection is the part I believe today's systems are missing.

At the same time, AI must remain connected to evidence.

If the system makes a claim, I want to know where it came from.

What supports it?

What contradicts it?

What happened first and what happened later?

What is fact?

What is interpretation?

What is missing?

What do we simply not know?

This is the foundation of Soul Engine™.

Selection Lab already has an initial technical proof of concept. The work includes archive retrieval, metadata, reference anchors, source attribution, grounded responses and a structure designed to preserve provenance rather than return an answer without its evidence.

We are also beginning to use the same logic in real institutional work.

In the first Selection Lab client case, information is not simply turned into a polished business story. Claims are classified as Confirmed, Client-provided, Projection, Proposed, Partially supported or Contradicted.

That process has already identified real inconsistencies involving investment classification, ownership language, partner language and the difference between existing and proposed governance.

So this is not only an archive project.

I want to build a trainable human reliability layer.

The system should be able to learn through human annotation, selection, continuity mapping and corrective feedback over time, while keeping the original evidence accessible.

I am not trying to build another foundation model.

I want to build something that can work with existing AI and help preserve the human source underneath its conclusions.

What I want to build next

The next research prototype will use a carefully prepared part of the 22-year Human Archive.

We will develop and test:

— archive ingestion and metadata
— retrieval and source attribution
— claim-to-evidence mapping
— provenance and chronology
— contradiction detection and preservation
— uncertainty and missing-evidence handling
— human annotation and corrective feedback
— selection-learning experiments
— evaluation of context preservation over repeated AI use

The goal is not to prove that AI can reproduce a human being.

The question is whether AI can learn from human vision without disconnecting the result from the human evidence and history that produced it.

Funding

I am requesting $150,000 for 12 months.

$55,000 — engineering and data infrastructure
$30,000 — research and evaluation
$25,000 — archive preparation and metadata
$20,000 — privacy, security and legal safeguards
$12,000 — testing and expert review
$8,000 — operations and documentation

The project can begin at a smaller level.

With $25,000, I can prepare a bounded archive sample and develop the first citation, provenance and selection-learning experiments.

With $75,000, I can expand this into evidence mapping, contradiction handling, human feedback and several real cases.

With $150,000, I can build and evaluate the complete research prototype and prepare the strongest components for institutional pilot testing.

Who I am

My name is Tatiana Ilina. I founded Selection Lab and built this archive over 22 years.

My professional background is journalism, documentary photography, filmmaking, interviews, archives and long-term work with people and institutions.

My formal education includes a Journalism degree and professional qualification as a journalist.

I did not come to this problem from a traditional AI laboratory.

I came to it after spending most of my professional life looking at human evidence and deciding what needed to be preserved.

Selection Lab is currently founder-led. Institutional Legacy LLC is the legal company behind the work.

I have funded the archive and early technical development myself.

English is not my first language. I use AI for translation, writing support, research and organizing my work. I have never hidden this.

In fact, using AI so extensively is part of how I began to understand the problem I am trying to solve.

AI gave me capabilities I did not have before.

I don't want less AI.

I want better AI.

I want the next generation of AI to have access not only to more information, but to the human vision, evidence and continuity behind it.

I think we still have time to build that layer.

That is why I am doing this now.

Selection Lab:
https://theselectionlab.ai/genesis

Human Archive / founder work:
https://tatianailina.com