jmooves fixes the chat answers that could end mid-sentence
A small proposed safeguard would have ensured finished AI replies show their full text, not a stalled partial version.
jmooves identified an unnerving edge case in Mike's chat experience: while an assistant reply was being revealed smoothly on screen, the stream could finish before the display caught up. The result was an answer that looked complete to the system but visibly stopped partway through for the reader.
The proposed change would have made completed and replayed responses immediately display in full, while preserving the gradual reveal during live generation. It was closed without being merged, so the upstream Mike product does not currently receive this protection.
Spotted something wrong? Or know the PR text has fresher detail than the writeup above?