Robert is sharp. His calendar, health portal, notes, and smart home are not talking to each other. He connects the dots manually. Every day.
Naavi connects them for him.
Robert is not struggling with technology. He is struggling with the gap between his tools.
Six tools. Six separate apps. Zero awareness of each other.
Robert holds all of this in his head and connects the dots manually.
Naavi does not replace Robert's tools. It sits above them — reading everything, remembering everything, and surfacing only what matters.
Robert speaks naturally. Naavi understands English and French, handles mid-sentence switching, and responds in whichever language Robert used.
Over time, Naavi builds a Cognitive Profile — Robert's daily rhythms, relationships, health context, and preferences. It learns without being taught.
Every morning, before Robert reaches for his phone, Naavi has already reviewed his day and assembled a brief. The answer is ready before the question.
Every morning, Naavi opens with an answer — not a question. It has already reviewed Robert's day so he does not have to.
Tuesday, March 17 · 08:15
What Naavi is doing
"Morning. Four things on my radar — Dr. Patel at 2, a medicat…"
Naavi has already reviewed the calendar, health portal, and pending threads — before Robert says a word.
"Dr. Patel today at 2pm, Centretown Medical on MacLaren. It's…"
Connects today's appointment to the last visit. Surfaces a useful preparation note.
"Your Metformin runs out around April 10. If you want to avoi…"
Gives the deadline in days, not just a date. Offers the next step as a question — never an instruction.
"Enjoy the walk. Good luck with Dr. Patel."
Knows he walks in the morning. Closes without a summary — Robert was there.
Design principle
Naavi opens every morning check-in with the count and a question — never a greeting, never a health quiz. Robert is a peer. He came for the answer.
Naavi is structured in layers — each with a clear responsibility, each invisible to Robert.
Robert's calendar, health portal, smart home, notes, and voice memos — the tools he already uses every day.
Robert's iPhone. All data syncs locally first, so Naavi works even without internet. His data stays on his device by default.
Naavi reads from every tool — silently, in the background, every 15 minutes. It normalises everything into one shared picture.
Claude — a large language model — reads Robert's full context and decides what matters, what to say, and when to say it.
Naavi's persistent memory. It grows over time — learning Robert's rhythms, relationships, preferences, and patterns.
The Cognitive Profile grows over time — within 90 days, it knows Robert well enough to anticipate rather than just respond.
The Cognitive Profile is not a database of facts. It is a living map that grows with every conversation.
Robert is 68. His health data, his routines, his relationships — these are not data points. They are his life. Naavi treats them that way.
Robert's health context, routines, and relationship data never leave Canadian servers. Regulated under PIPEDA and Ontario's PHIPA.
Naavi listens, transcribes, then discards the audio immediately. Only the text of what Robert said is retained — and only the text he'd expect.
Robert chose not to connect his banking. Naavi will never ask again. Boundaries are not guidelines — they are hard rules built into the system.
Robert can export or delete his entire Cognitive Profile at any time. Purged within 24 hours. No retention, no backup exceptions.
Robert sets his boundaries once. They are enforced in code — not in guidelines, not in privacy policies. The banking integration he declined will never appear again.
When Robert says "parle-moi en français", Naavi switches immediately — no announcement, no delay. It uses Canadian French vocabulary and stays in whichever language Robert used last.
English
"Dr. Patel today at 2pm, Centretown Medical on MacLaren."
Français
"Le Dr Patel est à 14h aujourd'hui, à la clinique Centretown sur la rue MacLaren."
Same information. Robert's language.