Final Project / Files
Project Files
Everything you need to rebuild NeuroAR lives on this page: the CAD, the boards, the firmware, the software, and the decals, all cleaned up and ready to download.
If you just want everything at once, this is the button:
Everything is released under the Fab license, (c) Yusuf Kusibati, June 2026. The license text travels inside every bundle, and the code files carry it as a header.
CAD and Mechanical
The glasses housing: the frame, temples, snap-fit covers, hinge, display module, prism holder, and wire concealers, designed in Fusion 360 around my own lenses.
To print it: Matte Black PLA, 15% infill with 2 wall loops for the body, and remember the cover wants to be thin with low infill so the XIAO's WiFi can breathe (the reasoning is in dev log §11).
The combiner is a 45x20x2 mm piece of mirror acrylic, laser cut, and it slots into the prism holder at the fixed angle.
Electronics
The main board is the ADS1292 biopotential board that lives in the right temple, designed in KiCad and milled on single-sided FR4. The 32-TQFP package needs 0.2 mm traces, so check your mill can do that before committing a chip.
The Spiral 0 EEG board is the earlier test board with the XIAO ESP32-S3 Sense, the SAMD11C14A sampler, OLED, button, and electrode header. The bundle has the KiCad files, the gerbers and milling g-code, all three firmware projects, and the CMSIS-DAP uf2 for flashing the SAMD11 over SWD.
Firmware and Software
The EEG SNN + dashboard package is the heart of Spiral 2. It contains the MicroPython firmware for the XIAO (ADS1292 driver, feature extraction, LIF-SNN runtime, Display Studio renderer), the Python training tools, the browser dashboard with its Node server, and the MicroPython firmware binary for the board.
To run it: copy the files in the firmware folder to the board with mpremote, then npm start inside the dashboard folder and open localhost:5176. The README inside walks through data collection, training, and deployment step by step.
It ships a neutral heuristic starter model, not my trained weights, and no recordings. An EEG model is person- and device-specific (it depends on the individual, the electrode placement, and the exact ADS1292 board), so a model trained on me would not transfer to you. Record your own labelled data with the dashboard and train your own, the pipeline is built for exactly that.
The Poke MCP bridge is the Spiral 3 package. It contains the XIAO MicroPython Poke UI, the Cloudflare Worker/Durable Object relay, the local FastMCP USB bridge, the deploy scripts, and package notes with the token/API-key setup. The public package uses CHANGE_ME_DEVICE_TOKEN placeholders instead of my live token.
The agentic glasses interface is the Week 15 dashboard generation: USB/WiFi configuration, camera, microphone, OLED, teleprompter, and the first EEG packet stream.
The early embedded experiments that picked the architecture are also here: the Muse-to-stepper blink test, the Muse EEG on the XIAO with BLE and OLED, and the OV2640 camera streaming test.
Decals and Identity
The NeuroAR logo and the sticker set for the glasses and the powerbank, cut on the vinyl cutter. The cut-ready zip has the layered files sized for the Anker 5000 mAh powerbank.
BOM and License
Rebuilding NeuroAR, In Order
- Print the housing from the STEP file and laser cut the mirror-acrylic combiner.
- Mill the ADS1292 board from the KiCad files and solder it (patience with the 0.2 mm traces, I ruined my first chip rushing this).
- Order the parts from the BOM and assemble: the board, OLED, lens, prism, and combiner all snap into the housing, no screws.
- Flash MicroPython onto the XIAO with the binary in the SNN package, then copy the firmware files over.
- Run the dashboard, check signal quality, record your own labeled data, and train your own model. The starter weights are deliberately a heuristic, your model has to come from your own head :D
If something doesn't make sense, the development log has the full story of every part, including the failures.
NeuroAR is released under the Fab license: (c) Yusuf Kusibati, June 2026.