Describe.
The person authors a memory. A place, a season, a person, a room, a weather shift, a trace of light. The starting point is not a product preference. It is what the person chooses to say.
System / In four moves
How memory becomes signal.
Scentient is a loop, not a model. A person authors a memory with consent. The system treats it less as text to summarize than as a latent atmosphere to represent. That representation can be rendered through scent as a physical proof surface. The person accepts, edits, or rejects what comes back. The graph is the asset.
The four stages
Describe. Translate. Refine. Realize.
Translate.
The system reads for latent atmosphere: warmth, brightness, openness, texture, intimacy, aliveness — along with scene cues such as weather, light, place, and season. The output is a sensory-affective signal, not a claim about how someone feels. The signal is the unit of compute.
Refine.
The person edits what feels closer. Rainier. Dimmer. More open. Less sweet. Less nostalgic. Each edit moves the signal and re-ranks what could render closer to true.
Realize.
The signal becomes something testable: a scent, atmosphere, or partner-rendered experience. Scent is the first physical proof surface because it sits close to memory, place, and affect. The artifact matters because it creates a response the graph can learn from.
The personal loop
The signal closes on itself.
Five passes through the loop — each one a consented comparison between an authored memory, a generated representation, and what felt closer to true.
- 01Choose the memory.
Start with what matters. A place, a person, a season, an atmosphere, a scene. The system does not begin with a product. It begins with authored memory.
- 02Shape the atmosphere.
Adjust light, texture, weather, proximity, intensity. Each adjustment clarifies the signal without claiming to read emotion.
- 03Translate the signal.
The system maps the memory into sensory direction. A candidate render appears with a confidence signal, not a final answer.
- 04Refine what feels true.
Keep what resonates. Change what misses. Rejection is not failure. It is signal.
- 05The graph gets clearer.
Future translations start closer to the person's own language, memory patterns, and prior refinements. The graph becomes more calibrated without pretending to know the person.
Immersive Rooms / Early capture surface
Begin with a room. Build toward a world model.
The first interface is not a claim to reconstruct a person's inner life. It is an immersive panorama room: a simple surface where a person can shape the place around a memory through light, texture, weather, proximity, and atmosphere. Those edits become structured signal for Scentient's affective memory graph.
Starting surface
Panoramas make memory spatial. They let a person point, compare, and refine: this corner, this light, this closeness, this weather, this room. Each adjustment can make the authored memory more legible to the system.
Longer arc
A true world model cannot stay visual. Human memory is not only image and language. Olfaction has unusually direct ties to brain systems involved in memory and affect, which is why scent can make a place or moment feel immediately present.
Scentient role
Scent becomes a grounding channel. Scentient uses that relationship carefully: not to read emotion, but to help translate authored memories and environments into sensory direction.
The Sensory Interpreter
Atmosphere becomes sensory direction.
The influence is world-model research: represent the structure that matters, not the surface description alone. Scentient translates environmental cues — temperature, light, space, texture, proximity — into sensory-affective dimensions a render system can use. These dimensions make a remembered scene computationally addressable. They do not claim to know how someone feels.
InputTemperature7.2 / 10
InputLight5.8 / 10
InputTexture6.5 / 10
OutputWarmth7.1 / 10
OutputIntimacy8.4 / 10
OutputAliveness6.3 / 10
The Vocabulary / The platform, in plain prose
Eight primitives. No code.
The substrate, written as a public glossary — enough to understand the system, not enough to disclose the machinery.
01
The Profile.
A person-owned memory context. It gathers consented signals from authored memories, edits, renders, and responses so future translations can begin closer to the person's own language.
02
The Signal Vector.
The system's compact representation of a memory's sensory-affective shape. It turns atmosphere into something that can be compared, refined, and rendered.
03
The Atmosphere.
The remembered world around the memory: weather, light, place, season, texture, proximity. Atmosphere is how Scentient keeps the memory from becoming plain text.
04
The Render.
The outward expression of the signal. Today, scent is the first physical render surface. Over time, the same logic can support ambient, spatial, or adjacent sensory outputs.
05
The Feedback.
The person's response after the render lives in the world. What felt closer. What missed. What changed later. Feedback is how the system learns without pretending to read emotion.
06
The Consent Envelope.
The permission layer around the memory and its derived signal. Scope, portability, deletion. Sovereign affective memory in plain terms.
07
The Loop.
Authored memory to signal to render to response to refinement. The loop is what turns a one-time artifact into compounding learning.
08
The Graph.
The asset. A growing map of consented memory-response patterns that makes future translations more precise. At maturity, it can support a personal model for sensory-affective translation. The graph is the moat.