New + recent threads
A Messages-like list. Shortcuts stay compact. No composer until New.
Ask Claire is a conversation list, not a manifesto. New starts a thread. Answers arrive as objects you can tap—sources, people, open loops—and the same Claire sits quietly inside a chat until you ask her to help you reply.
A Messages-like list. Shortcuts stay compact. No composer until New.
Inviting someone into the research room—avatars, chips, scoped sources.
Answers are objects: sources, follow-ups, never a sourceless paragraph.
Structured cards you can track. Observation separated from inference.
One ribbon. Inline promises. No reply strip until you ask.
The in-chat sheet is Ask Claire with Maya already focused.
The stale-chat moment: who is waiting, then one suggested next line.
Shorter, warmer, clearer—on the text you already typed.
No close button on a tab. No input until you open New or a thread. A lime New control, a vertical thread list, and compact shortcuts—Catch me up, Find open loops, Check the tone.
Across your connected conversations
Deck by 10 · 3 sources
Friday still open · Telegram
2 promises · 1 unanswered
September 18 · Maya + studio
The tab is a conversation list. New, recent threads, and three compact shortcuts. The input appears only after you open New or a thread.
Use @ to focus on a person or chat
Claire never sends messages from here. She only reads connected conversations and cites what she used.
Typing @ opens people and chats—not a hidden filter. Selecting one becomes a focus chip on the next screen.
User turns stay simple. Claire turns are cards: a kicker, a scoped header, sources that open the original message, and follow-ups that start the next useful move.
Claire never sends from here
Ask about a plan, something she said, or the tone of this thread. Sources will stay inside Maya unless you @ someone else.
Focus is visible before you send. Relationship memory frames the person, then stays out of the way.
Looking at Maya
Maya asked to move the public launch to September 18. The studio thread agreed the extra week was the point—not a delay for its own sake.
Default answer language: kicker, short finding, tappable sources, then the next useful move.
Promises, questions, plans
You said this yesterday · WhatsApp
Asked Tuesday · Telegram
Maya proposed · not confirmed
Open loops are cards, not a paragraph. Track lands in Promises. Open jumps to the source chat.
Looking at Maya
Replies are short and warm. Most turns are one or two sentences.
Maya seems to want confirmation more than a long plan right now.
“Yes—we’re set. I’ll send the deck by 10, and we can take comments first.”
Tone is never a personality diagnosis. Observation, inference, then a matching next line.
Looking at Maya and Noah
Tell Maya that Noah is free Friday, and offer Thursday afternoon as a backup.
Two people in focus become a comparison, not a blended summary. The conflict is the product.
Inside a conversation Claire has three intensities: a quiet ribbon, an on-demand sheet that is the same Ask Claire thread pre-scoped to this person, and send-time help on the draft you already wrote.
WhatsApp · active now
Product review tomorrow · deck still open
Ambient Claire only: one ribbon and an inline promise. No auto-generated reply strip.
Same thread as the Ask Claire tab, already scoped. Claire never sends from here.
“Yes—we’re all set for tomorrow. I’ll send the updated deck by 10, and we can use the first few minutes to cover your comments.”
Drafts generate when the sheet opens—not on every inbound message. Shortcuts stay compact.
Studio team · last 4 days
You haven’t answered. She needs a yes or a new time.
Maya said the deck is enough. No reply needed.
“Friday still works on my side—happy to take Priya’s notes first.”
The stale-chat moment: who is waiting, what is noise, then one line that closes the loop.
In Telegram on Tuesday he said Friday at 10 works if Maya can shift. He has not repeated that in this thread.
The unique Claire move: stay in Maya, pull Noah, and mark the foreign source as another conversation.
WhatsApp · rewriting
Highest new in-chat ROI: rewrite the text already in the composer. Tiny context, happens at send time.
WhatsApp · before send
This thread has been short and warm. Your draft is more formal than usual.
Regret prevention, not a lecture. Observation first, one tap to match the thread, send still belongs to you.
Research and drafts stay reversible. Use reply drops text into the composer. The person still hits send. Auto-reply that sends is not a launch AI surface.
Every answer names the messages it used. Sources open the chat with a highlight. Foreign sources say they came from another conversation.
User-tapped actions only. Suggested replies generate when the sheet opens or when someone asks for help—not on every inbound message.
Focusing a person should feel like inviting them into the room: avatars, lime chips, and an answer header that says who Claire is looking at.
Tone checks separate what the messages show from what Claire is guessing. No personality diagnosis. A matching next line is the payoff.
In chat, keep the Quick Context ribbon and inline promises. Do not also launch a smart-card tray. Presence is quiet until asked.