Bidirectional peripheral neural interface — product research program (email edition)

Preserved source identity

Source PDF: output/pdf/novel-biotech-bpni-product-research-program-2026-08-12-email.pdfPrivate repository source

SHA-256: 7913b4f0a49b128e11f47619929ff480f7f55b593d2b5ea3dd08b5f4dc26382e

Companion used for text-layer verification: output/pdf/novel-biotech-bpni-product-research-program-2026-08-12.pdf · 36c8faa666ceb7d39d9f61ed49f9e4099d7d5e7deab1f23a4c5bd5e0b3046a43

A hash establishes byte identity, not truth or present authority. All ten pages were rendered and inspected. This email PDF has no extractable text layer and is visibly lower-resolution; its pages match the full-resolution product-program PDF page-for-page, so the transcription was checked against that companion source. On page 8, the first letter of “Two” is visibly clipped at the left edge in both PDFs; the complete heading is preserved here from the companion PDF text layer.

Source page 1 of 10 · executive summary

Accepted control package; no product selected

The source said the program had an accepted evidence baseline, layered functional model, neutral architecture frame, 28-question backlog, and gated roadmap through mid-June 2027.

Direction preserved

Direct peripheral-nerve forward path plus afferent return path; not reduced to prosthesis motion control.

Evidence bounded

Short human components exist; chronic exact-route direct ENG plus stimulation remains open.

Options stay neutral

Electronic, biohybrid, and combined families remain research branches with zero selected.

Next gate is not a device

Science must define a dimensionless simulator contract, then Backend must review it.

Source page 2 of 10 · system model

The layered idea becomes a chain of distinct claims

Forward path: central nervous system → peripheral nerve → recording interface → computer

1. Sense: candidate activity.
2. Confirm: existence / validity within a bound.
3. Reject artifact: keep unresolved state.
4. Condition: preserve provenance.
5. Classify: uncertainty retained.
6. Compute: propose; never authorize.

Return path: computer → stimulation interface → afferent peripheral nerve → central nervous system

1. Encode: bounded return content.
2. Command: proposal only.
3. Authorize: allow / inhibit / review.
4. Condition: no waveform or values.
5. Stimulate: delivery state.
6. Acknowledge / effect: report; do not infer benefit.

Source page 3 of 10 · evidence control

The accepted Science packet contains 34 bounded claims

QA PASS ≠ claim PASS.

Core boundary

B00 supplied N=2 component evidence over 4–5 weeks. Only S3 completed the limited closed-loop task: mixed neural + EMG median-USEA decode, separate ulnar-USEA stimulation, 41/47 trials, possible proprioception/efference-copy contribution, and exact timing unknown.

OPEN — chronic exact-route human system

No accepted record established chronic fully implanted direct natural ENG plus peripheral-nerve stimulation in one human platform.

UNKNOWN — same-interface duplex timing

Separate, alternating, same-session, time-interleaved, and simultaneous duplex were distinct claims.

FAIL — broad platform claims

Universality, broad novelty, feasibility, safety, efficacy, readiness, freedom to operate, and availability were not established or were prohibited at this breadth.

Source page 4 of 10 · problem landscape

Seven problem classes were narrowed — none became a product indication

The source warned that its bar lengths were visual routing emphasis only — not evidence scores, effect sizes, or rankings.

• 5 deferred: retained only as bounded questions or comparators; evidence prerequisites remained open.
• 1 research branch kept: motor loss survived for comparator narrowing only; not indication, architecture, or product selection.
• 1 rejected for now: diagnostics was a reversible prerequisite-gap disposition and did not fire KILL-04.

Invasive comparator rule: if a removable, temporary, usual-care, or lower-risk external system provides essentially the same bounded benefit, the invasive branch must stop or be reframed.

Source page 5 of 10 · architecture research

Three families remain neutral, unselected, and evidence-asymmetric

Only presently defensible common “platform” layer: schemas, provenance, uncertainty, calibration lifecycle, monitoring, and imported fail-closed semantics.

No common biological map, return code, anatomy, endpoint, or universal calibration followed.

Source page 6 of 10 · research roadmap

Eight dependency-ordered phases lead to a research closeout — not a launch

Can proceed now

Maintain the six research-control artifacts, questions, traceability, and exact Science imports.

Waiting

Branch-specific Science, dimensionless simulator contract, Backend review, and governance imports.

Blocked

Architecture or indication selection, simulator implementation, device visual, testing, fabrication, implantation, development.

Source page 7 of 10 · question backlog

Twenty-eight questions convert ambition into falsifiable decision points

Priority composition: P0 / evidence and problem gates — 10; P1 / discriminators and lifecycle — 10; P2 / program and artifact readiness — 8.

Kill / pivot discipline

• Missing evidence remains OPEN or UNKNOWN.
• Only Science defines or fires scientific thresholds.
• Killing one bounded branch does not kill all branches.
• Equivalent lower-risk benefit rejects the invasive branch.
• No family advances on aesthetic appeal or “living” status.
• KILL-04 remains prospective and unfired.

Question → evidence: named owner, dependency, required source record, route, population, duration, endpoint, and limitations. Evidence → decision: one bounded downstream decision, with uncertainty preserved and no claim upgrade. Decision → stop / pivot: a conditional routing consequence tied to accepted criteria — never elapsed time.

Source page 8 of 10 · gate control

Two gates are closed. Nine still prevent selection or visualization.

Science

Claims, comparators, routes, failure modes, kill criteria, simulator assumptions.

Backend

Runtime, technical security, offline/non-actuating behavior, monitoring, fail-closed execution.

Legal / Ethics / Regulatory

Privacy, consent, ethics, access/equity, coercive use, reversibility, regulatory and freedom-to-operate judgments.

Integrator + user

Later bounded research routing and exact visual authorization — never scientific truth by preference.

Visualization remains blocked. Report diagrams explain accepted records. They are not a device concept, architecture depiction, implant, blueprint, or G10 bypass.

Source page 9 of 10 · dormant blueprint program

Twelve explanatory candidates exist only as a gate plan

Every future exact artifact needs: prerequisite-owner acceptance → Independent QA reproduction → Integrator + user authorization → bounded nonfunctional production → final QA.

Product cannot self-authorize.

Source page 10 of 10 · custody & next action

The accepted program is reproducible, bounded, and ready for the next research gate

Independent QA: P0 / P1 / P2 / P3 = 0 / 0 / 0 / 0. Product QA turn 019ff63b-18d2-73a2-8009-5a80ca8dd8e1 reproduced all six hashes and made no writes.

Overall: REVISION. Release, product development, visualization, testing, and external action remained unauthorized.

1. Science simulator contract. Versioned, dimensionless assumptions, parameters, sources, and thresholds. Separate G0/G1 closure required.
2. Backend review. Only after Science acceptance: offline, non-actuating, runtime security, monitoring, and fail-closed requirements.
3. Research branch decision. Integrator + user decide one bounded problem or multiple branches. This remains research routing, not product selection.

Caveat preserved from the source: this report summarized accepted repository records. It was not clinical, regulatory, legal, safety, efficacy, novelty, patent, freedom-to-operate, feasibility, or development advice. No human testing, biological work, implant development, or availability was represented.