A mixed reality device that predicts how another person's brain is processing a conversation and makes that invisible signal visible.
Every tool we have for understanding other people - therapy, psychology, medicine - requires their cooperation and self-awareness. But humans are poor reporters of their own inner states. We mask, misread, and misremember what we feel.
Existing emotion-recognition systems try to solve this by reading faces. They map expressions to labels: "sad," "angry," "neutral." But facial expressions are performances, not ground truth. They're culturally variable, consciously controlled, and weakly correlated with actual neural state.
Nooscope doesn't read the face. It reads the situation the way the brain does.
On March 26, 2026, Meta FAIR released TRIBE v2 - a foundational model trained on 1,115 hours of fMRI data from 700+ participants. TRIBE v2 predicts, with zero-shot generalization, how a human brain responds to any video, audio, or language stimulus.
This unlocks something that was impossible the week before: predicting neural engagement from the outside, non-invasively, in real time - without an fMRI machine, without EEG electrodes, without touching the person at all.
Nooscope is the first device built on this capability.
Nooscope is a lightweight MR glasses system that:
- Captures the audio and visual context of a live human interaction via a forward-facing camera and microphone
- Predicts the neural response of the person you're talking to using TRIBE v2 - which brain regions are most active, what kind of processing is happening
- Overlays a subtle, non-intrusive signal in the wearer's field of view - not a label, not a diagnosis, but a spatial map of neural engagement updating in real time
The overlay is intentionally ambiguous. Nooscope doesn't tell you what someone is feeling. It shows you how intensely their brain is processing the moment - and lets you respond accordingly.
[ Forward camera + mic ]
│
▼
[ Scene capture - 3s rolling window ]
│
▼
[ TRIBE v2 inference ]
- Video encoder: V-JEPA2
- Audio encoder: Wav2Vec-BERT
- Language encoder: LLaMA 3.2
- Output: predicted fMRI activation map (whole-brain, ~70,000 voxels)
│
▼
[ Neural engagement scoring ]
- ROI extraction: visual cortex, auditory cortex, language regions, limbic system
- Engagement index: weighted activation across regions
│
▼
[ MR overlay ]
- Color-coded ring around person: warm = high activation, cool = low
- Updates every 2–3 seconds
- Rendered via monocular AR display (Brilliant Labs Frame / custom ESP32 module)
| Existing emotion AI | Nooscope | |
|---|---|---|
| Signal source | Facial expressions | Sensory context → neural prediction |
| Output | Emotion label ("sad", "angry") | Neural engagement map |
| Ground truth | Behavioral proxy | Predicted cortical activation |
| Cultural variability | High - expressions differ across cultures | Low - neural response to stimuli is universal |
| Requires cooperation | Yes - person must express visibly | No - prediction is situational, not personal |
| Model | Computer vision classifier | TRIBE v2 foundational brain model |
Autism therapy - People with autism struggle to read social cues intuitively. Nooscope gives them a neural signal without requiring them to interpret facial expressions. The first assistive device built on a brain model rather than a behavior model.
Medical training - Doctors and nurses learning to read patient distress in high-stakes situations. Simulated patient interactions with real neural feedback, without real patients.
High-stakes communication - Negotiation training, leadership development, conflict resolution. Anyone whose work depends on reading the room - and who currently has no tool to practice it.
Long-term: consumer MR layer - As smart glasses become mainstream (Meta Ray-Ban, successors), Nooscope becomes a software layer - an app on the platform, not just a device.
| Component | Purpose |
|---|---|
| ESP32-CAM or Raspberry Pi Camera | Forward-facing scene capture |
| INMP441 MEMS microphone | Audio capture |
| Raspberry Pi 5 | Edge inference - TRIBE v2 pipeline |
| Brilliant Labs Frame (or custom) | Monocular AR overlay display |
| 3D-printed glasses frame | Wearable integration |
| Python + PyTorch | Inference pipeline |
| Node-RED or custom dashboard | Real-time monitoring (dev mode) |
This hardware instinct comes directly from a prior project: a wearable health and safety monitoring system for underground miners, integrating MAX30102 (HR/SpO2), MQ-series gas sensors, MPU-6050 accelerometer, and SOS trigger into a two-unit wearable (bracelet + smart jacket) communicating via MQTT - patent pending.
- TRIBE v2 local inference - pipeline setup
- Scene capture module - rolling 3s video/audio window
- Neural engagement scoring - ROI extraction from predicted activation maps
- Real-time overlay - color-coded engagement ring
- Wearable integration - ESP32-CAM + AR display mount
- V0 demo - live interaction with overlay updating in real time
Active development. Started April 2026, one week after TRIBE v2 was released.
The noosphere (from Greek nous, mind) is Pierre Teilhard de Chardin's concept of the sphere of human thought - the layer of consciousness that envelops the Earth the way the atmosphere does. A scope that sees into it.
Nooscope is built on the belief that the most powerful technology lives at the intersection of deep human understanding and rigorous engineering.
CS student. Hardware builder. Psychology obsessive.