Thanks for the feedback. The objective is to spread kindness, regardless of the tool. Also, the final AI-generated message is provided to the user to read and then send, so "dictated but not read" is not applicable.
I think writing an inspirational message, and knowing a stranger will read your exact words, is just a fulfilling as being on the receiving end. For that reason I recommend you lose the AI component, except perhaps as a translation mechanism.
This was originally that way, but when I ran some tests, the main reason of not sending it continuously was the typing part. So, I tested this method and people became more active.
I’d rather have pre-defined templates than some AI magic button. That way you can set a mood and intent for messages without relying on an llm ( which can open up messages to be rather unkind).
Great feedback, will consider it for the next update. BTW, if you haven't noticed, you can write a message in any language yourself and send to AI to translate it optionally if it is not in English then send it or send it directly. This is the most authentic way, the AI-crafted message is for the users who prefer simpler way to spread kindness.
Doesn't a message written by an AI carry significantly less weight, though? I'd much rather read someone written by an actual human (this applies to most things, really), and would hold it in much higher regard.
thanks for the suggestion, I did:
"You’re right that the means can change the message. People often value perceived effort and authenticity, so provenance matters. The evidence is mixed: AI drafts can be rated more empathetic than typical human responses, but when recipients discover a message was machine‑authored, the benefit can vanish unless it’s disclosed and human‑edited.
So the app is designed for that reality: (1) hand‑written for people who prefer them; (2) AI‑assisted drafts that must be reviewed by the sender; and (3) clear, simple disclosure"