Repository navigation
Project ATO 2.0.1 — Whole-project Review and Corrections
This correction release follows a bounded review against the original project
brief. It fixes confirmed UART frame interleaving, removes a duplicate author
from reference R17 and the rebuilt manuscript, adds a portable current CAD
launcher, and updates current navigation for the public repository. See the
review dossier
and current firmware.
The prior v2.0.0 tag, all 2,733 protected Stage 1-5 files, original board/CAD
geometry, studies and production core remain unchanged. The corrected UART
application is an explicit additive revision with a fresh target build.
Project ATO now connects its proposed invasive spinal interface to a complete,
inspectable digital research and engineering account: scientific foundations,
explicit future assumptions, two application mechanisms, reproducible models,
native electronics, target-linked firmware and mechanical assemblies.
Both original ambitions remain central: expert-informed practice that could
improve later unassisted motor learning, and feedback over a selected
physiological function. AstroSam's
Mind-Link inspired the progression from
future vision to inspectable engineering. ATO develops its own spinal mechanisms
and artifacts; it incorporates no Mind-Link code or geometry.
Start here
- Read the integrated illustrated manuscript attached to this release.
- Use the engineering entry guide
to find the native KiCad/FreeCAD files, models, firmware, source evidence and
reproduction instructions. - Review the six-sheet mechanical drawing set,
also attached as a PDF. - Use the contribution record
and defence guide
to understand the project's ownership, assumptions and questions worth testing.
Included engineering
| Deliverable | Release contents |
|---|---|
| Evidence and mechanisms | Focused 18-reference primary/inspiration audit; 30 classified claims; all 36 original requirements receive a scoped final disposition |
| Computational work | Preserved conservative negative results; separate 128-scenario conditional mechanism study; new compiled-core integration replay for both applications |
| Native electronics | Eight routed populated boards; 1,298 positions / 1,296 fitted; 80-sheet master and 258 harness rows; current saved ERC/DRC/parity/unconnected checks pass |
| Target firmware | Corrected nRF5340/BT40F UART image, actual-callback regression, source/configuration and independent binary audit; original image retained |
| Full-size mechanical reference | 260-part combined native assembly; actual eight-board geometry at 1:1 scale; 600 x 310 x 240 mm chassis including feet; 83 mating housings and mapped wiring interfaces |
| Future implanted interface | Separate 60 x 58 x 14 mm can/header before leads; 230 physical solids, six explicit internal allocations and 36 mapped body ports; fitted/exploded editable CAD and exchange files |
| Reproducible release | Source archive, manuscript, drawing sheets, external SHA-256 checksums, committed Git-blob manifest and source-bound verification receipts |
The new host integration replay executes the unchanged production C admission
and phase state machine through conditional companion and body-response
adapters. Its 114,426 assertions cover 5,184 completed motor event requests,
1,200 pressure event requests, 13 core boundary/fault cases, four separate ideal
supervisor cases and six companion-feature cases. Neutral teaching improves
assisted performance without extra OFF learning; favorable and adverse teaching
produce different retained outcomes. These are consequences of declared model
assumptions, not measurements of human skill acquisition or electrical delivery.
Read the engineering limits with the results
The current target records RHD frames only in OBSERVE and inhibits recording
while stimulation is armed. It does not implement the new replay's live
spinal-feature feedback during MOTOR. The host replay demonstrates conditional
software interfaces; it does not execute the target peripherals, boards or
biology together. The earlier uninterrupted 450k aggregate / 4k acquisition
schedule and 4% phase-charge target remain unearned.
The actual eight reference boards retain their real dimensions. They do not fit
inside the small future can. Its electronics, energy storage, chronic thermal
performance and qualified interfaces require a new miniaturized implementation.
The present radio/network-core image is unimplemented; the future antenna/window
are allocations. The 3.154 W / 4.880 h reference power/runtime figures are
conditional calculations.
No hardware was fabricated or physically executed, and no human or animal data
were collected for ATO. Retained human learning benefit, selective chronic
spinal access, universal physiological control and physical implant qualification
are not established. The pressure example uses an external sensor and a
target-specific positive pressor premise; it does not establish a spinal blood
pressure decoder or healthy-consumer benefit. CAD and native electrical rule
checks establish their stated digital properties.
Provenance and use
Pratham Varma directed this student research concept. Codex and collaborating
AI agents supplied extensive research, modelling, implementation, engineering,
verification and writing assistance. Personal learning and later revisions should
be recorded honestly; the release does not certify completed student mastery or
external expert review.
Original project material remains an all-rights-reserved portfolio under the repository's
proprietary license. Third-party libraries, linked firmware and references retain
their own notices and rights. The prior v1.0.0 single-channel desktop subsystem,
dated studies and adverse findings remain preserved. Website and video revision
are parked.
Use a full Git clone at v2.0.1 for history-dependent verification. The source
ZIP is convenient for inspecting files and contains no .git history. Compare
downloaded assets with the attached SHA256SUMS.txt; the checksum list excludes
itself. Large populated STEP files are supplied as lossless ZIPs with verified
extraction launchers.
Publication verification — 17 October 2026
The release tag and main branch resolve to 9f6cf0a4403a00549a858ae9f0cecdd9dce0c5bb.
The current integrated release workflow
and exact-archive packaging workflow
passed at that commit. The full study reproduction
passed all 15 jobs at 7825512; its study code, inputs and workflow are identical
in the final tag. Subsequent changes concern only release verification,
publishing and their metadata, with separate checks at the final commit.
A separately fetched checkout passed the manifest and current release checks.
The source archive contains 2,819 files, each compared with the exact tagged
Git blob. The 6,509 release checks, 91 canonical artifact bindings, both
downloaded PDF hashes and all four remote asset digests passed. The 2,733
protected Stage 1-5 files and previously published v2.0.0 tag remain unchanged.
Compare downloads with SHA256SUMS.txt.