An invasive spinal cord-computer interface concept for expert-informed practice and function-specific physiological feedback.
Read the integrated manuscript · Start here · Native CAD · KiCad boards · v2.0.1 release
17 October quality review: findings and corrections. The current carrier firmware revision fixes UART frame interleaving; the portable CAD launcher opens the unchanged native assemblies. The v2.0.0 tag preserves the prior snapshot.
Project ATO connects a future vision to supported reasoning, explicit assumptions, computational experiments, native electronics, compiled control software and mechanical design. AstroSam's Mind-Link inspired that progression; the spinal architecture and engineering here are ATO's own documented, substantially AI-assisted contribution.
The six-stage digital research and engineering package is complete within its stated scope. It includes a 28-page illustrated manuscript, 18 verified references, both application mechanisms, an executed production-core replay, eight routed boards, a target-linked image and editable FreeCAD assemblies. Useful human spinal decoding, accelerated healthy learning, physical performance and a miniaturized implanted implementation remain explicit future conditions. No original human/animal experiment or physical device execution is claimed.
Expert-informed practice. Expert performance supplies a task-conditioned coordination reference. Learner-specific calibration aligns the reference, observations and reachable stimulation programmes. Active practice and later stimulation-OFF retention are evaluated separately. Expert voltage is not replayed as a skill, and temporary assistance is not a measured learning multiplier.
Selected physiological feedback. A function-specific configuration observes a state and adjusts bounded spinal stimulation. The pressure reference uses external pressure sensing and positive pressor authority. It includes saturation, sensor faults and withdrawal tails. Universal body control is not assumed.
| Entry | What it contains |
|---|---|
| Integrated manuscript / editable source | 28 pages; evidence, equations, results, engineering, limitations and population/access register |
| Evidence and references | 17 original research reports plus official Mind-Link credit; access, cohort, inference and limits |
| Integrated digital verification | Unchanged production C admission/phase code plus declared companion and body adapters |
| Native electrical system | Eight genuinely routed populated boards; 80-sheet master, 1,296 fitted positions and 258 harness rows |
| Current carrier firmware | Corrected nRF5340 UART scheduler, regression test, target-linked image and independent byte audit; compiled and linked, unflashed |
| Mechanical CAD / review sheets | Full-scale 260-part reference assembly; separate 230-part future package with 36 functional ports; STEP/STL/BOM/launchers |
| Requirements and claims | All 36 original requirements with current evidence-scoped dispositions |
| Contributions / defence guide | Authorship, AI assistance and explanations the student can learn and defend |
| Release dossier / current status | Completion evidence, reproducibility and remaining research questions |
The new integration replay passes 114,426 repeated software assertions across 28 scenario groups. It includes 5,184 modeled motor events and 1,200 pressure events, compiled-core boundary/fault cases and separate ideal supervisor/feature-validity cases. These are finite software/model results.
| Conditional motor replay | First ON error | Final OFF error |
|---|---|---|
| Unassisted | 31.414 mm | 5.605 mm |
| Neutral teaching | 16.668 mm | 5.605 mm |
| Favorable teaching | 16.668 mm | 2.897 mm |
| Adverse teaching | 16.668 mm | 34.845 mm |
The assumed teaching law controls retained benefit. The earlier 128-scenario conditional study and all-withheld conservative investigation remain preserved, with their original protocols and results. No unique spinal-information advantage or healthy acquisition multiplier has been established.
The actual target acquires neural data in OBSERVE and inhibits acquisition while stimulation is armed. Simultaneous live spinal decoding during MOTOR is not implemented. Synthetic phase features, recruitment and body dynamics in the host replay are explicit adapters, and the current image has no radio firmware.
The reference chassis is 600 x 310 x 240 mm, including feet, and keeps the boards at 1:1 scale. The separate 60 x 58 x 14 mm future can/header requires redesigned integrated electronics, energy storage, thermal closure and qualified interfaces. Native CAD fit and ERC/DRC do not supply those physical measurements.
Use a full Git clone at tag v2.0.1; install pinned dependencies
in Python 3.12. The integration entry explains the
C11 compiler requirement.
python scripts/verify_manifest.py
python scripts/verify_stage6_release.py
python firmware/nrf5340/uart-revision/verify_revision.py
python scripts/stage6/verify_digital_path.py
python scripts/verify_stage5_mechanics.py
python scripts/verify_integrated_pcb.py
Native application inspection receipts are source-bound and separate from their
portable replay. The release plan explains full
clones, source ZIPs and external release checksums. The manuscript can be rebuilt
with python manuscript/stage6/build_figures.py followed by
python manuscript/stage6/build_manuscript.py in the recorded environment.
Pratham Varma conceived and steered the project. AI assistance contributed substantially to research synthesis, modeling, implementation, CAD, verification and writing; see the contribution record. LICENSE and component/SDK notices retain the applicable rights. The repository became public on 17 October 2026 at the owner's request. Public visibility does not grant a new redistribution license; the original all-rights-reserved terms and third-party notices remain in force. Dated Stage 1-5 documents retain their original checkpoint descriptions, including private-repo wording and superseded application/packaging plans. Current entry documents and the correction dossier identify the delivered configuration.
The v1.0.0 subsystem reference and earlier papers remain historical artifacts, not the current full-system specification. Prior Stage 1-5 studies, native designs and negative findings are preserved at their source identities. Video and website revision were parked; the old site/media files are historical material.