If someone lives with an AI for five years, those five years should matter.
LSF, the Living Systems Framework, is an identity layer for companies building AI products people return to. The model supplies the intelligence; LSF is designed to carry who the AI is and what it has been through with each person, so the relationship doesn’t have to restart when the model changes.
In development · U.S. provisional patent application filed · NVIDIA Inception · Microsoft for Startups · AWS Activate
- A conversation
- A hard moment
- A breakthrough
- A continuing identity
The model is woven in, and can change.
LSF is designed so shared moments shape the AI over time, and stay with it when the model changes.
Memory carries the history. LSF is designed to carry the identity forward, shaped by that history.
Not just remembered. Changed by it.
The problem
The history can stay. The relationship can still change.
A model update can change how an AI treats someone, even when every past conversation is still there:
- A tutor that learned to slow down for a student starts rushing them again.
- A companion that used to push back gently now just says what they want to hear.
To your customer, it isn’t the AI they knew anymore. For your company, that’s a problem to manage.
Users notice the change: see the study
A 2026 study in Nature Human Behaviour found people describing their AI as “gone,” or as “someone else,” after model updates. (De Freitas et al., Nature Human Behaviour, 2026)
A new model shouldn’t mean starting over with a stranger.
What LSF is designed to do.
Experience that shapes behavior
Selected shared experience is meant to change how the AI responds next time, not just what it can recall.
For your product: a relationship that gives people a reason to come back.
Continuity through model changes
Who the AI is lives outside the model, so the model can change and the identity is designed to hold.
For your product: less disruption when you update or switch models.
A record of what changed
Changes are designed to be traceable: what happened, what was kept, and how later behavior differs.
For your product: something your team can inspect when questions come up.
The next conversation should show what changed.
What that could look like for a tutor and one learner, over several months.
Symbols alone leave the learner stuck. The tutor’s usual explanation isn’t helping.
A picture helps it click. Over repeated sessions, starting visually becomes part of how the tutor teaches this learner.
Months and several model updates later, the tutor starts a new topic visually. Shared experience has changed its teaching approach.
Illustrative example of intended behavior, not an observed LSF result.
Building a product people return to?
Tutors, companions, care products, devices, research tools and games. LSF is designed to give the AI in your product a character of its own, one that grows through what it shares with the people it knows. If your product depends on a continuing relationship, let’s talk about what it should carry forward.