New in CardSnap: Meeting Memory, 28 languages and smarter search
What shipped recently: meeting notes that turn into commitments with real dates, answers that tell you how sure they are, an interface in 28 languages, and instant search that understands plain questions.
Meeting Memory
Add a memory to any contact after a meeting, in whatever words come out. No form, no fields, no template. Type or speak it.
CardSnap turns that into three things: a short executive summary, a list of who was in the room, and the commitments made, each with an actual due date rather than a vague soon.
Nothing is saved until you have looked at it. The extraction is shown first so you can correct a date or drop a line, then it goes on the record in one tap.
Later you can ask what you promised somebody and get the real answer: the date, the meeting, the exact promise, and whether it is still open.
Ask AI now tells you how sure it is
Every answer carries a confidence label. A confident answer says so. A partial one says so too, which matters more, because a vague reply that reads confidently is worse than no reply.
When it is not sure, you get a Try again, search wider button. That drops the narrow filter and re-reads a wider slice of your records, which usually turns a maybe into a specific answer.
Answers about the product, not just your contacts
Ask AI also answers how-to questions about CardSnap itself, such as how to save cards to Google Contacts or how automatic saving works, with the screen named and the steps numbered, quickest route first.
28 languages
The interface is available in 28 Asian and international languages, including Malay, Thai, Vietnamese, Tagalog, Bahasa Indonesia, Chinese, Japanese, Korean, Hindi and Tamil.
Voice input goes further and handles how people actually talk, including Manglish and Singlish, so you do not have to flatten your speech into formal English to be understood.
Instant search
Search now understands plain requests such as high priority contacts in logistics, or people I met in Singapore. Most of those are answered instantly from your own records without spending an AI action.
It also stopped treating everyday words as country codes, so a phrase like my high priority contacts searches what you meant rather than filtering to one country.
Quieter running costs
Behind the scenes, images are downscaled before reading, a second read is skipped when the first one is already clear, and each task is routed to the cheapest model that can do it properly. You see the same results, and the service stays affordable as it grows.
