AI · Looking ahead
Useful AI should not need your whole life story.
Your information should be useful with or without AI. We’re exploring how Signcards could provide relevant, user-chosen context to future AI tools, with access and permission to act kept separate. Direct AI-agent connections are not available today or at launch.
The same Signcard, in two formats.
These screenshots show the same sample Breakfast Signcard as AI context and as a table. The context view frames saved answers for an assistant; the table organizes those answers by field.
Demonstration data only. These static previews do not connect an AI agent or grant access to a Signcard.
Relevant context, not everything.
An agent helping with dinner should not need your whole profile. Our direction is to make relevant, modular context available through Signcards you choose to share. What matters for one task may not belong in another; relevance alone would not grant access.
A suggestion is not your decision.
AI could help you think, but its interpretation should not silently become your answer. Our aim is to keep suggestions separate from information you have confirmed, with room to review, correct, or decline. Structured information would not, by itself, prove that an answer is true.
Knowing is not permission to act.
Knowing your preferred airline would not authorize booking a flight. In our intended model, permission to read context would be separate from permission to buy, book, or share. Any future agent acting for you would need explicit authorization within the limits you set.
These principles guide what we are exploring. External AI-agent connections are not available today and will not be open at launch. No providers, protocols, or release dates are committed.
Back to Engineering