Doesn’t the fact that something admits its faults make you trust it more?
Last week, I tried out Granola during a client call and it tagged one of the speakers incorrectly for the first ten minutes. Frustrating, sure, but what stayed with me was the little note that the transcription might require some fixes, and the two clicks required to do so. I trusted the tool more after that, not less.
It got me thinking about how few products ever own up to being uncertain. Most give you a confident answer and hope you don’t bother looking further. The products I’m willing to pay for, such as Raycast, PostHog, and Supabase, always explain themselves: an execution plan, a raw event, a log line.
But I cannot know if this is really a trust trigger or something I see just because I am a developer. A friend of mine who is not a tech-savvy person said that she would prefer an application working well rather than being told that it could be wrong.
Where are you on this one? Do you like products to doubt themselves or not to? Has any product won your trust through a mistake?
Replies
i think your friend has a fair point too. Most people want things to work without having to question every result, so the challenge is finding the right balance.
Point to the uncertain speaker label and put replay and correction beside it
That gives someone a way to check the result without reading logs. A vague warning on the whole transcript leaves them guessing what to trust
Think it's less dev vs non-dev and more about stakes. For low stakes I want magic, for client calls or logs I want truth plus uncertainty.