THALis: early architecture alternatives

EEG / reservoir / hand / maybe EMG + feedback... lots of possible pieces. keep the options here without making all of them prerequisites for the first version

brainstorm structure: divide the prosthesis into mechanics, control, sensing and proprioceptive/feedback components. The September–October document considers EEG, myoelectric input, sensor fusion and low-energy/adaptive learning rather than a single settled implementation.

initial sketch: EEG electrodes → high-dimensional reservoir weighted by tractographic information → an SNN-style processing stage → joint output with myoelectric integration. It also mentions activity modes, a limited hand/forearm scope, Raspberry Pi/printed hardware, a visualization GUI and online learning. These are proposed components; the later CSP/LDA implementation does not establish that this entire early sketch was built.

EEGsynth processing modules

questions already present in the brainstorm: continuous versus discrete movement representation; signal-to-noise limitations; calibration time; online calibration; sensor fusion; feedback. These are more useful as a design-decision log than as a retrospective claim that the initial concept was the final architecture.

reading pointers: open-source hand assembly guide, control-algorithm reading, feedback reading.

review question: which proposed components were later retained, replaced or deferred, and what evidence drove each choice? Keep the appended previous-year MesoNet plan separate from the THALis brainstorm; the same document contains both.