Brain-State Circuit Replay / Resonance

Photos record what we saw. Brain-State Circuit Replay asks whether we can reconstruct what we were.

Public boundary: the repo presents this as Brain-State Circuit Resonance because the scientifically careful target is bounded, partial state resonance — not exact playback, literal time travel, or current human brain-state reconstruction.

Open the live prototype: https://rola2005-klc.github.io/brain-state-circuit-replay/

This repository is building the conceptual, scientific, and engineering foundation for a speculative neuroscience / BCI thesis. The public frame is resonance, not playback: if photos, videos, music, and smell can trigger memory, could future closed-loop interfaces help a person partially re-approach the integrated self-state of a past moment while keeping uncertainty and safety visible?

The emotional problem is The Unreturnable Present Paradox / 当下不可归悖论: when a person is very happy or inside an important moment, the thought "I can't go back anymore" can create anticipatory grief that blocks the person from fully living that moment.

The interactive simulation is secondary. It supports the thesis by making assumptions visible: target states, technical replay modes, similarity, uncertainty, reconstruction error, and safety risk. The main artifact is the foundation: what the concept means, what science supports, what is plausible extrapolation, and what remains speculative.

This project does not claim literal time travel, medical treatment, exact human memory replay, exact visual playback, or EEG reading synaptic traces.

Start here: one-minute path

One-sentence thesis: Photos record what we saw. Brain-State Circuit Replay asks whether we can reconstruct what we were — with the current public frame restricted to partial, uncertain resonance, not exact replay.

Emotional problem: The Unreturnable Present Paradox / 当下不可归悖论 — the fear that a meaningful present cannot be returned to can create anticipatory grief inside the moment itself.

Core object: not scene/video/image replay, and not a single emotion label. The target concept is a past-associated integrated self-state: body state × attention × emotion topology × context × self-model × meaning × configured ignorance / the ability to feel the world that way.

Open these three artifacts first:

1. Live prototype / 3D mindmap: https://rola2005-klc.github.io/brain-state-circuit-replay/
2. Process journey: https://rola2005-klc.github.io/brain-state-circuit-replay/journey.html
3. Full thesis doc: docs/unreturnable-present-paradox.md

Audience-specific one-minute path:

• Researchers: thesis doc → 3D mindmap science clusters → limitations/future research.
• Internship reviewers: project brief → process journey → live prototype → tests/synthetic-data caveats.
• General viewers: project brief → process journey → 3D mindmap.

For a polished non-technical overview, use the project brief: https://rola2005-klc.github.io/brain-state-circuit-replay/project-brief.html

Boundary in one line: weak “time travel” means bounded state reinstatement; medium means subjective similarity; strong/literal time travel is only a metaphor and is not claimed here.

Presentation Path

Use project-brief.html as the clean entry point. It gives the hook, problem, thesis, demo boundary, scientific anchors, system concept, ethics, current artifacts, and next steps without opening raw Markdown in the browser.

Recommended flow:

1. Start on project-brief.html and read the hook plus boundary.
2. Open journey.html and walk through the six-step process.
3. Open index.html only after the boundary is clear: it is a concept map and simulation scaffold, not brain-derived output.
4. Open future-research.html as the closing horizon: what current science cannot do yet and what a responsible research path could ask.
5. Keep the source Markdown docs available for deeper questions.

Audience paths:

• Researchers: start with the full thesis, then read the native memory trace explainer, research map, future research directions, and data model.
• Internship reviewers: start with the project brief, skim the process journey, inspect the 3D mindmap, then close with the future research page, foundation, presentation outline, tests, and synthetic data scripts.
• General viewers: start with the project brief, follow the process journey, then use the 3D mindmap before the future research horizon.

Project thesis

We may not be able to go back in time, but we may learn how to revisit parts of brain-body-self states.

The central theory is that a meaningful past moment is not only a stored image. It is a distributed integrated self-state involving interoceptive prior/body state, salience weighting/attention, affective manifold/emotion topology, situational embedding/context, self-model, meaning, and configured ignorance: the specific boundary of what the past self did not yet know, expect, or have to account for, which made the moment feel open.

The project treats return as an affective trace encounter rather than archive restoration. Retrieval can update memory through reconsolidation, so the goal is not to recover an unchanged old self; it is to let the present self safely re-meet a past-associated self-capacity.

The refined idea is that no previous external brain-state recording is strictly required for the concept because plasticity and LTP can be treated as a native biological record. That record suggests trace existence, not selective access. Experience changes synapses and circuit dynamics, but the hard problem remains retroactive addressability and controlled reactivation: how to access the right trace, reactivate it safely, and measure what happened without pretending it is exact.

A future closed-loop system might approximate that state by:

1. using photos, audio, place, smell, sleep cues, or context as external memory cues;
2. treating plasticity and engram-like traces as the biological record;
3. representing a target moment as a distributed brain-body state signature, with integrated self-state treated as a conceptual synthesis rather than an established neuroscience metric;
4. applying cueing, neurofeedback, or carefully bounded stimulation as controlled nudges;
5. monitoring similarity, reconstruction error, and safety risk;
6. showing why exact replay is not the target, while partial state resonance may be scientifically imaginable as a bounded approximation.

Why this is interesting

Humans already experience accidental state resonance:

• a song brings back a whole period of life;
• a smell triggers childhood memory;
• trauma cues recreate body states;
• athletes re-enter flow states.

This project asks whether future neurotechnology could make partial state resonance intentional, measurable, and safe while treating emotion as structured information rather than noise. The design target is Resonance over Recreation: any past-associated state intervention should return the person to the present, not trap them in the past.

Future versions should express resonance as convergence toward a declared reference attractor relative to a null baseline, and avoid pseudo-precise confidence when calibration is unknown.

Scientific backing

This is a speculative project, but it is now grounded in adjacent real research areas:

• LTP and plasticity: experience can leave durable changes in synapses and circuit dynamics.
• Memory engrams: animal studies show that activating tagged neural ensembles can evoke memory-like behavior.
• Targeted memory reactivation: sounds/odors paired with learning can bias later memory reactivation, especially during sleep.
• Closed-loop neurostimulation: human studies increasingly decode cognitive/emotional states and adapt stimulation in response.
• Hippocampal memory prostheses: MIMO-style models have been explored for supporting memory encoding/recall.
• Neurofeedback: real-time fMRI/BCI feedback can train partial self-regulation of brain activity.

Important boundary: EEG is a non-invasive state correlate. It may help monitor broad state dynamics, but it is not a circuit/engram readout and does not read synaptic memory traces.

Current repo role

The repo currently contains three layers:

1. Presentation foundation: thesis, pitch outline, scientific boundaries, and roadmap.
2. Technical concept: native memory traces, technical replay modes, EEG limits, state-estimation limits, and synthetic data caveats.
3. Supporting demo: a browser simulation that makes the concept tangible without claiming biological validation.

The realistic floor is cue-triggered recall and Targeted Memory Reactivation. The frontier is direct engram reactivation and guided endogenous replay as a technical mode, not exact playback. The simulation exists to keep those categories explicit. Current claim maturity is E1/E2: literature-grounded concept plus toy scaffold, not a validated brain-derived system.

Run locally

python3 -m http.server 8765

Open:

http://127.0.0.1:8765

Run tests

node test.js
python3 -m unittest discover -s tests

The expected test outputs include All simulation tests passed. from the Node toy simulation tests and standard unittest output from the Python synthetic-data tests.

Pre-publish safety check

Before any public push or nightly publish, run:

python3 scripts/safety_check.py

This local gate checks that agent-inbox/ comment artifacts are ignored/untracked, scans repo text-like files for credential patterns, verifies local HTML references and anchors, and runs the JavaScript and Python toy-model tests. It exists because the public GitHub Pages site is concept/thesis + thought map + simulation scaffold only; private local comment inbox data must stay local.

Data-backed toy model

The repo includes a generated synthetic dataset and decoder baseline:

python3 scripts/generate_synthetic_data.py --samples-per-state 120 --seed 42 --out data/synthetic_state_samples.csv
python3 scripts/evaluate_decoder.py --samples-per-state 120 --seed 42

Current baseline: nearest-centroid decoder reaches about 90.83% accuracy on 120 held-out synthetic test rows. Synthetic classifier results illustrate pipeline behavior only and are not empirical validation. This does not prove real brain-state decoding; it proves the project has an explicit state representation, protocol assumptions, and a reproducible evaluation path.

Synthetic data layer: docs/data-model.md

Known limitations

After advisor review, two structural limitations are flagged and explored in the repo:

1. Decoding power. Telling apart a target "happy memory" state from background noise and other emotional states. A noise-robustness sweep (scripts/evaluate_decoder_noise.py) shows accuracy decays from ~91% on clean synthetic data to near chance well before realistic noise levels.
2. Stimulation coverage and precision. Where memory lives and whether a finite electrode array can evoke it without recruiting unrelated neurons. A toy 2-D simulation (scripts/simulate_stimulation_coverage.py) visualizes the precision/coverage tradeoff; a literature map (docs/stimulation-coverage-precision.md) surveys engram localization and modality tradeoffs.

Full writeup and future plan: docs/limitations.md. The key stance is that current technology does not yet allow real memory decoding or safe, precise engram stimulation; near-term work should test coarse state convergence, cue-driven reactivation, neurofeedback, uncertainty displays, and safety gates rather than claiming replay.

Repository structure

index.html              # static website entry point
project-brief.html      # polished project brief
journey.html            # one-minute process journey page
future-research.html       # creative but bounded future research directions
style.css               # responsive visual design
src/journey.js          # keyboard/click process journey interaction
src/simulation.js       # pure simulation logic
src/app.js              # UI rendering and interaction
test.js                 # Node tests for simulation behavior
tests/test_synthetic_data.py # Python tests for synthetic data generation/decoding
scripts/generate_synthetic_data.py # synthetic state dataset generator
scripts/evaluate_decoder.py # nearest-centroid decoding report
scripts/evaluate_decoder_noise.py # decoder noise-robustness sweep (limitation 1)
scripts/simulate_stimulation_coverage.py # stimulation precision/coverage toy sim (limitation 2)
data/synthetic_state_samples.csv # generated toy dataset
data/decoder_noise_sweep.csv # noise-sweep results
data/stimulation_coverage_sweep.csv # coverage-sweep results
docs/limitations.md     # advisor-flagged limitations and current responses
docs/stimulation-coverage-precision.md # engram localization & stim modality literature map
docs/theory.md         # user's core photo-to-brain-state-replay theory
docs/concept.md         # presentable concept note
docs/presentable-foundation.md # thesis, pitch, framework, and boundaries
docs/presentation-outline.md # 8-slide presentation structure
docs/unreturnable-present-paradox.md # refined thesis and ethical boundary
docs/native-memory-trace-reactivation.md # EEG/LTP/engram/TMR explainer
docs/research.md        # literature-backed research map
docs/future-research-directions.md # future capabilities current science cannot yet provide
docs/data-model.md      # synthetic data schema and baseline
docs/development-plan.md # long-term implementation plan
docs/roadmap.md         # product/research roadmap

Important disclaimer

This project is an educational and speculative interface prototype. It is not medical advice, not a diagnostic tool, and not a validated neuroscience model.

Future direction

The long-term version is not just a nicer interface. The serious frontier is a set of research problems current science cannot solve yet:

• encoding-imminence fingerprinting: detecting when a lived moment is becoming deeply encoded without turning life into surveillance;
• configured ignorance modeling: representing what the past self did not yet know without paternalistic information control;
• affective topology mapping: modeling the shape and attractor structure of feeling rather than reducing it to valence/arousal;
• closed-loop multimodal cue orchestration: coordinating smell, sound, light, haptics, posture, and narrative context while maintaining a present-anchor;
• sleep-integrated resonance assistance: collaborating with native sleep replay without disrupting sleep or consent boundaries;
• personal corpus state archaeology: using writing/art/context as reflective scaffolds for past states, not as an oracle;
• reconstructive loosening protocols: asking how unwanted state-attractors could be loosened voluntarily, reversibly, and safely.

Full page: future-research.html and source note: docs/future-research-directions.md.