Luckily, the feedback on the first release of Chess for Glass has overall been positive. Criticism I have seen, and also agree with, focus on (1) mistakes made by voice recognition with similarly sounding files (since b, c, d, e, and g all sound similar) and (2) placing the voice menu over the chess board.
The next release addresses the first issue with the ability to enable a phonetic alphabet for voice input. In this mode, voice input uses the phonetic alphabet for files (alpha, bravo, charlie, etc) without further from or to designation. Example: "ok glass, pawn, echo two, echo four" plays the pawn from e2 to e4. Although uncommon in chess notation, this form of voice input is less prone to sound ambiguities.
As for the second issue, I am afraid that showing the voice menus over the board will stay for a while though. One of the reasons I wrote Chess for Glass was to demonstrate some features of the GDK (the Glass SDK) since I work in that team in Google[x]. Because the GDK voice menus follow rather strict UX requirements, I am currently simply using those rather than using a custom-made version that would put the text more to the right. This is likely to improve over time though, so stay tuned.
Finally, the upcoming update has all sorts of minor fixes and improvements "under the hood", mainly focused on third party engine support. For example, engine analysis is now outputted in a much better format, as shown below.