01 · In development
Human Experience
An ongoing, consent-based interview system that reconstructs a person’s life history — childhood, relationships, adversity, work, changing beliefs — and their own interpretations of it.
Partners · Licensing and research
For teams building agents and robots.
We’re developing ways to give AI systems experiential foundations — and the evaluations to show whether they help. We’re looking for a small number of design partners to shape that work before anything is sold.
What we’re developing
Three foundations, one layer of interpretation.
02 · In development
Synthetic Experience
Coherent, explicitly synthetic life histories for research, simulation and agent applications — representing diverse circumstances without claiming anyone lived them.
03 · In development
Experiential Integration
Portable Life Packages and runtime adapters, so agents can retrieve relevant experiences, weigh inferred lessons and revise beliefs. BEDROCK separates recollection, interpretation, hypothesis and confidence.
Who it’s for
Early partners
- Agent-platform developers
Persistent agents whose behaviour should be grounded, explainable and differentiated.
- Robotics research teams
Exploring how experience might inform behaviour — always inside operational safety constraints.
- Simulation companies
Where a diversity of plausible histories may itself be the value.
- Persistent AI companions
Systems people return to over months and years.
Before you ask
Questions we expect you to ask us
Does it work?
We don’t know yet, and we won’t claim it does until we can show it. The first gate is evidence that experiential grounding produces measurable benefits beyond richer prompting.
The first evaluations will measure behavioural differentiation, decision quality, stability across repeated trials, sensitivity to irrelevant memories, calibration, and consistency with the claimed provenance. Until results exist, every demonstration on this site is labelled illustrative.
Will it make our systems less predictable?
Different decisions are not necessarily better decisions. In some simulation products, useful diversity may be the point. In safety-critical robotics, uncontrolled variation is a liability.
Our rule: experiential identity must never override operational safety constraints. It can change how an agent explains itself — never whether it complies.
Where do the experiences come from?
Two sources, never confused: adults who voluntarily take part in structured interviews under explicit privacy and licensing terms, and explicitly synthetic histories constructed by our team. Every package carries its provenance.
Do you verify that memories are true?
No. A memory is evidence of a recollection; its meaning is a hypothesis. BEDROCK keeps what a person recalled apart from what we infer, and records confidence with its reasons. Memory verification, psychological interpretation and behavioural usefulness are separate questions.
How are contributors treated?
As contributors, not products. Participation fees and/or royalties are being considered, subject to agreement. Consent is recorded before any session, and recollections are presented only as recollections.
What can we license today?
Nothing yet. We’re in foundation research. The stages ahead — behavioural validation, private pilots, commercial licensing, embodied intelligence — are intended stages, not commitments to dates. Design partners help decide what gets built first.
Design-partner program
Start a conversation
Tell us what you’re building. We’ll reply personally — this goes to a person, not a sales sequence.
Opening soon
The inquiry form opens soon.
Until then, write to us directly and a person will reply.