Observe
Process a time-ordered stream and maintain a compact internal state.
Lab · proof-stage research
Eidos Brain explores how a streaming system can preserve meaningful change without pretending to be autonomous authority.
The research combines prediction error, surprise handling, anomaly receipts, and human review. It remains a proof-stage architecture experiment, separate from the primary Eidos Works service offer.
Current maturity
Research with receipts. Not a finished commercial platform.
- Input
- Live or replayed streams represented as bounded signals.
- Processing
- Prediction error and surprise scoring identify changes worth preserving.
- Output
- Metrics, incident receipts, plots, and proof artifacts for human review.
- Authority
- Human-reviewed; no claim of autonomous operational control.
Research method
Learn the expected stream, measure the residual, preserve the part that changes meaning.
Predict
Estimate the next signal and measure the residual between prediction and observation.
Review surprise
Compare error with recent error behavior so unusual change can be inspected.
Emit receipts
Keep raw and calibrated metrics visible beside human-readable evidence.
Limits and next experiment
The work does not prove general intelligence, consciousness, universal prediction, or autonomous decision-making.
Current research value comes from reproducible domain proofs, false-positive discipline, compression and anomaly-preservation measurements, and incident explanations. The next experiment should be chosen from a documented proof gap, not a broader claim.
Read the public case note