Adopts Scott Makeig's editorial review of the JOSS paper.

Adopted

• Drop the unsupported EMG claim; reframe to EEG/MEG. No published work validates AMICA as an EMG source-separation method; the AMICA-plus-EMG literature uses AMICA to remove EMG as an artifact from EEG. First sentence now reads "electroencephalographic and magnetoencephalographic (EEG/MEG)" with foundational citations (Makeig et al. 1995; Vigario et al. 1997; Iversen & Makeig 2019). Statement-of-need and audience lines follow suit.
• "standard method" -> "widely applied method"; "activity" -> "activities".
• AMICA "produces the most dipolar" -> "produces both the least dependent and the most dipolar" decompositions (delorme2012 measured both dependence and dipolarity).
• Name the reference implementation's author ("written by Jason Palmer"); "driven from" -> "callable from" MATLAB/EEGLAB.
• "AMICA's usual 2000 iterations" -> "the AMICA heuristic default of 2000 iterations", plus a note that source separation keeps improving slowly at higher iteration counts (Frank et al. 2023, BIBE: MIR increases to at least 4000 iters, PMI to at least 3000).
• Newton note clarified: the Newton update speeds convergence; the divergence is framed as initialization-basin sensitivity, not a Newton defect.
• Cite Frank et al. 2025 (NER) at the definition of k = frames/channels^2 (their kappa) and Frank et al. 2023 (BIBE) for the MIR/PMI benchmarking metrics.
• State of the field: complements the reference Fortran AMICA used with EEGLAB (not EEGLAB itself); "shares EEGLAB's output format" -> "uses the same output format as that Fortran version".
• Acknowledgements: "his advisor Ken Kreutz-Delgado".
• Five new references added to paper.bib (Makeig 1995, Vigario 1997, Iversen & Makeig 2019, Frank 2023, Frank 2025).

Verification

• 1557 words (body), 0 em-dashes, no orphan citations, all new bib keys cited.