What Current Science Cannot Do Yet

Why this page exists

The destination is imaginable before it is reachable

The questions Resonance is built around — what it would mean to re-approach a past integrated self-state, whether the quality of a moment can be partially reconstructed, and how configured ignorance could be preserved or honored in a future system — are not questions current science can answer. They are questions current science can approach, and approaching them honestly requires naming the distance between here and there.

This page does not describe a product roadmap. It describes open research directions that, if they became tractable, would give the Resonance framework somewhere to go.

Cautious source-backed watchlist

Current BCI / Neuro Signals to Watch

These are near-term radar items for framing questions, not evidence that Resonance can decode memories, identify engrams, or infer private experience. They are useful only if kept inside the project boundary.

• Neuralink surgical robot / implantation workflow reports: treat as secondary and low-confidence until an official source is added to this page. The interesting signal is infrastructure: reports describe efforts to make electrode-thread implantation more repeatable and scalable. That matters for future BCI access, but implantation repeatability is not decoding, and it says nothing by itself about memory reconstruction. Secondary report
• arXiv 2605.04680, "Multi-Level Bidirectional Biomimetic Learning for EEG-Based Visual Decoding" (2026-05-06): watch as an EEG visual-decoding representation paper, not as exact memory reading. The relevant boundary for Resonance is that reconstructing or aligning visual representations from EEG remains a lossy model output tied to task data, not access to a private remembered scene. arXiv
• arXiv 2605.03874, "Spatiotemporal Convolutions on EEG signal -- A Representation Learning Perspective on Efficient and Explainable EEG Classification with Convolutional Neural Nets" (2026-05-05): watch for efficient EEG state-classification methods and explainability claims. The boundary is classification: separating task or state labels from EEG features is not the same as reading a lived self-state or recovering a circuit trace. arXiv
• arXiv 2605.03169, "NeuralSet: A High-Performing Python Package for Neuro-AI" (2026-05-04): watch for possible tooling and scaffold relevance. A PyTorch-ready neuro-AI data interface could inform future simulation, provenance, or multimodal dataset organization, but it would not validate the current toy model or supply biological evidence by itself. arXiv

Presentation boundary: cite these as signals to monitor, not claims to import. The current project remains a concept and synthetic simulation scaffold.

Future direction 1

Encoding-Imminence Fingerprinting

Current gap: current EEG, fMRI, wearables, and affective-computing systems cannot detect when a person’s brain-body system is beginning to treat the present as deeply worth encoding.

Why it matters: Resonance needs a target. Without knowing what made a moment feel unreturnable while it was happening, a system would aim at generic positive affect instead of a specific past-associated self-state.

Research path: longitudinal naturalistic studies could combine high-density EEG, pupil data, cardiac phase, movement, environment, and later self-report to search for multi-signal clusters that precede strong autobiographical encoding.

Risk: ranking “moments worth encoding” can become surveillance over life. Non-detection and user refusal must be valid outcomes.

Future direction 2

Configured Ignorance Modeling

Current gap: archives record what someone wrote, saw, did, or knew. They do not record the shape of what a past self did not yet know: unresolved futures, uncertainty boundaries, and open assumptions.

Why it matters: importing present-day certainty into a past-associated state would destroy part of what made that state itself. Configured ignorance is not a missing variable; it is part of the phenomenology.

Research path: AI interpretability, formal epistemology, and personal timeline modeling could attempt to estimate a past belief boundary from writing, calendar context, messages, and metadata while preserving uncertainty.

Risk: any system that suppresses or masks present knowledge risks paternalism and manipulation. Consent must be voluntary, granular, and revocable.

Future direction 3

Affective Topology Mapping

Current gap: affective computing often reduces state to valence, arousal, dominance, or a few labels. It cannot map the shape of an emotional state: coupling, instability, attractors, and transitions.

Why it matters: integrated self-state is a shape in feeling-space, not a mood score. Resonance would need to approach the right basin, not just a state with similar valence.

Research path: topological data analysis and dynamical-systems modeling could be applied to continuous physiological, behavioral, and self-report data to identify basins and uncertainty boundaries.

Risk: similar-looking state geometries may feel radically different from inside. Any topology must carry uncertainty and allow rejection by the user.

Future direction 4

Closed-Loop Multimodal Cue Orchestration

Current gap: current Targeted Memory Reactivation uses limited pre-selected cues and does not dynamically orchestrate smell, spatial audio, light, temperature, haptics, posture, and narrative context from a real-time resonance signal.

Why it matters: the conservative floor of Resonance is cue-guided recall. A future version would need richer cue constellations that adapt gently rather than optimizing intensity.

Research path: smart environments, olfactory display, spatial audio, wearable haptics, and physiological feedback could become a closed-loop cue sequencer.

Risk: immersive cue environments can blur reconstruction and reality. Every session needs an unmistakable present-anchor and exit condition.

Future direction 5

Sleep-Integrated Resonance Assistance

Current gap: sleep replay and TMR research can bias consolidation in limited ways, but cannot target a specific integrated self-state, monitor subjective consolidation quality, or adapt across nights toward a past-associated state signature.

Why it matters: sleep is already the brain’s native memory-processing environment. The least invasive future path may be collaboration with that process.

Research path: closed-loop sleep-stage detection, acoustic/odor cueing, and multi-night feedback could test whether gentle cues bias memory processing without disrupting sleep architecture.

Risk: sleep interventions act while the person cannot consent in the moment. Pre-consent, withdrawal rules, and sleep-health protection are core requirements.

Future direction 6

Personal Corpus State Archaeology

Current gap: models can imitate style, summarize text, or infer sentiment. They cannot reliably reconstruct the integrated self-state that produced a diary entry, song, photo, or project.

Why it matters: this is the most near-term bridge before advanced BCI: writing, art, music, and captured context can become a resonance scaffold without claiming direct brain access.

Research path: personal archives could be mapped over time using language models, metadata, optional physiological annotations, and user feedback loops. The goal is reflective scaffolding, not oracle-like inner-state claims.

Risk: models will confabulate plausible state descriptions. The person’s own judgment must remain the authority.

Future direction 7

Reconstructive Loosening Protocols

Current gap: reconsolidation research can sometimes attenuate emotional force, but cannot precisely and reversibly loosen the gravitational pull of an unwanted state-association without risking distortion or overreach.

Why it matters: if Resonance studies intentional movement toward a past state, it must also ask about the inverse: how someone might voluntarily loosen a state they no longer want to be pulled back into.

Research path: future reconsolidation science, cueing, and state-space modeling might study attractor strength: how strongly a state pulls the person back, and whether that pull can be reduced safely.

Risk: identity modification, coercion, and irreversible change remain hard red lines.