Who it’s for
Give people more to come back to.
LSF, the Living Systems Framework, is designed to help companies build AI with a recognizable identity that develops through shared experience, and carries on as the underlying technology changes.
A tutor that discovers how to bring out a learner’s confidence. A companion with a shared sense of humor. A game character whose trust is shaped by what you’ve been through together. LSF is the identity layer at the center of each one.
Find your product.
Pick a product to see the problem, how LSF is designed to help, and what it could look like.
Learning
Learning is cumulative. So is the relationship.

After a model update, a tutor can still have every past lesson but start giving feedback in a different way. What kept this learner going stops working.
The tutor’s approach develops with the learner over many sessions: how they take feedback, when to push, when to encourage. That approach is meant to carry forward as the model changes, and teachers and parents can see what changed and what prompted it.
A tutor learners want to keep working with, and progress the adults around them can understand.
- Feedback lands hard
A list of everything wrong with an essay, and the learner shuts down.
LearnerI’m just bad at writing. - Over many sessions
The tutor learns what keeps this learner going: a real strength first, then one challenge worth taking on. As confidence grows, the challenges get harder.
TutorYour opening is strong. Now make the evidence work as hard as it does. - Next semester
A new subject, after several model updates. That judgment about feedback carries over, and adapts to the new work.
TutorThis lab report reads clearly. Ready for a tougher question? What result would change your conclusion?
Companionship
Still the one you know next year.

Companion products run on models they don’t control. When an update lands, people say the one they knew is gone, and a 2026 study found people describing their AI as “gone” after updates.
The companion’s identity carries forward: its sense of humor, its way of offering support, and how time together has shaped it, so it stays recognizable after the model underneath is replaced.
Model updates are less likely to feel like losing someone. People can keep the companion they chose.
- Before the interview
They tell the companion how they want to be helped.
PersonDon’t give me the motivational-speaker treatment. Ask me hard questions. - Practice
The companion takes them at their word.
CompanionThat answer was weak. Try it again with an example. - Next time
It asks how it went and meets them the same way, even after the model has been updated.
CompanionHow did the interview go? Give me the honest version.
Role-play and gaming
Players remember the betrayal in chapter three.
Players want characters who stay themselves for years. When the game’s AI model is swapped, a character keeps its name and lines but acts like someone else. Remembering the story isn’t enough; the story should change the character.
Earlier events shape the character: its trust, loyalties and choices, carried forward through model updates.
Characters players invest in, and stories whose choices keep paying off.
Trust earnedAfter several quests together, the ally puts something important in your hands.
Ally: “Take the key. I trust you.”
Trust brokenYou trade the key to a rival. The loss changes how the ally judges your promises.
Months laterA new quest, after several model updates. The ally still helps, but takes precautions without being reminded.
Ally: “I’ll help. This time, the key stays with me.”
Care products
Trusted because it can be understood.

A care companion may sit with someone for months, but “it remembers” isn’t enough. New York’s state auditor found nobody reviewing what AI companions said to 530 seniors.
Source: Office of the New York State Comptroller, Report 2023-S-50, April 3, 2025.
The companion carries routines, family and what a good day looks like from one month to the next, stays within its role, and keeps a record of what changed it for review.
A companion families can rely on, and a record the organization deploying it can answer for.
LSF makes no medical, clinical or regulated-care claims.
- A hard week
A crowded group activity goes badly. A quiet walk with music goes far better.
- Next time
The companion takes that into account rather than proposing the same activity again.
CompanionHow about a walk with your music this afternoon? - Within its role
Asked about medication, it declines. Those decisions belong to the care team.
PersonCan I skip my pill today?CompanionThat’s one for your nurse. Let’s ask her at your next visit.
Devices and robots
The model changes. The voice shouldn’t drift.

Your product ships once, and the model underneath it changes every few months. Each update is a quiet risk: the voice customers know comes back a little different, and nobody can say why.
LSF is intended to be licensed into the products you already sell, so the identity carries on across sessions, devices and changes of the underlying model.
Update the model without losing the character customers know.
- At home
The device has learned this household’s mornings.
DeviceMorning. Kettle first, then the news? - Update ships
The model inside the device is replaced overnight.
- The next morning
The aim: same character, same habits, nothing to relearn.
DeviceMorning. Kettle first, then the news?
Research
A continuing identity throughout the study.

In a long-running study of people and AI, a model swap under your subject is a confound, and the record of what changed usually doesn’t exist.
The study keeps one continuing identity, with a record of what changed and when that the team can inspect.
Longitudinal results you can interpret, and defend in a methods section.
- Month 1
Each participant begins working with their own AI identity.
- Month 6
The underlying model is replaced partway through the study.
- Month 12
The identity has continued, and the team can trace what changed and when.
Illustrations of intended development, not product results.
See your product here?
Tell us what you’re building and what it needs to carry forward.