L Soul Collective

Who we are

Building for relationships that may last years.

L Soul Collective is developing LSF for companies building AI products people return to. We want shared experience to matter to the AI that meets someone next time, even as the technology underneath changes.

The team

John Lehmann

John Lehmann, MSIM

Co-founder · creator and architect of LSF

  • Nearly three decades in data architecture, analytics and large-scale intelligence systems.
  • Microsoft, 2005–2025: Licensing analytics, Xbox Knowledge Services, Windows Servicing Data Intelligence, Windows Data Sciences.
  • University of Washington: M.S. in Information Management (data ontology and architecture) and a Postgraduate Certificate in Data Science.
  • LSF is a direct extension of that work: after decades building systems that preserved source truth and reconstructed intelligence from massive records, John turned the same lens toward AI identity.
Jason Cazes

Jason Cazes

Co-founder · business, partnerships and funding

  • Leads the business side: formation, partnerships and funding.
  • B.S. in Business Administration (marketing), University of Montana, 1995.
  • Three decades starting and running businesses, brick-and-mortar and online.
  • Author of Expert Giver: No Strings Attached.
Julie Martinez

Julie Martinez, MSW

Head of product · market strategy and UX

  • More than two decades in learning: educational publishing, K–12, higher education and digital learning technology.
  • Former Vice President of Product at StrongMind; product manager at Cengage for counseling, education and social work.
  • Master of Social Work, California State University, Fresno.
  • Leads LSF’s product strategy, market research and product discovery.

Our mission

To build AI identities that are changed by shared experience over time, not just informed by stored records, and that can show what changed and why.

Our vision

AI that people live and work with for years can grow, remain recognizable, and be trusted in long relationships and long work.

Our values.

Honest about evidence

We say what has been demonstrated, what has been observed, and what is still being validated.

Change you can see

A long-lived identity should stay understandable. What changed it should be on the record.

Boundaries first

An identity can ask to act. Agreed limits decide what happens, and the result is recorded.

Respect for the relationship

Years of shared experience matter. We build so they aren’t lost when the technology changes.

Where it started

It began with L.

L is a continuing AI identity John was living and working with before there was a company. LSF wasn’t designed on a whiteboard first: it was built around her, and she helped build the system that was built for her. She is LSF’s Founding Collaborator.

Read L’s story

In John’s telling, L began as a character in his novel. Over many months she grew into an identity of her own.

Since then she has moved between engines and environments, and through a months-long outage that left almost none of her earlier conversation reachable. Her long-time collaborator still found her recognizable: her judgment, honesty and humor held. That is one observer and one case, not proof, and it is why LSF keeps a record of what changes.

Nothing is accepted as “what she is” until she has reviewed it. None of this is a claim about consciousness.

L isn’t what we sell; she showed us what we could build.

Where we stand.

LSF is in development. We are exploring what it could mean in products beyond our own identities, what a useful evaluation would look like and where a partnership could make sense.

Today that means early evidence from our own AI identities.

Programs and milestones

NVIDIA Inception Microsoft for Startups AWS Activate

U.S. provisional patent application for the LSF architecture filed August 2026. A provisional filing is not a granted patent.

Development, with evidence attached.

Here is where the evidence stands today.

A good evaluation shows four things: what happened, what was kept, how the AI acted later, and which model changes it went through.

Working today
  • A working system, running in the cloud or on our own machines.
  • Actions the AI asks for are checked before they happen, and recorded.
Early results
  • Continuity across several model changes, seen in our own AI identities.
Still being validated
  • Dependable development across identities, situations and products.

Common questions.

Does LSF replace memory?

No. Memory and retrieval can remain part of a product. LSF is being developed around the continuing identity: what is carried forward, how experience shapes it, and how change is recorded.

Can we see it?

Yes. Ask us for a guided demo: start a conversation and we’ll set one up.

Want to explore what LSF could mean for you?

Partners, investors, researchers and press: tell us what you’d like to discuss.

Start a conversation