Welcome to the Uzu Master Manual — the dense, one-page reference for every system inside the app. Each Academy lesson about app mechanics links directly to its section here. Skim the headings, jump to what you need, come back anytime.
The Vortex: Navigation & Gestures
The Vortex is Uzu's home. Instead of burying features behind menus, the main actions orbit your conversation — swipe and tap to move between chat, missions, and your profile without leaving your flow. Gestures are consistent across screens: the same swipe that opens your history in chat opens your progress elsewhere. Learn three gestures and you've learned the whole app: swipe for navigation, tap-and-hold for context actions, and pull for refresh.
The Dual Economy: G-Credits and DAG-Credits
Uzu runs on two currencies. G-Credits are the everyday currency — earned through daily activity, missions, and streaks, and spent on standard AI messages. DAG-Credits are the premium currency that powers heavyweight compute: reasoning models and custom Personas that hold complex standing rules. Conversion is fixed and honest: 1 DAG ≈ 10 G. You can transmute G into DAG whenever a job genuinely needs the premium tier — the Auditor pass, a Persona session, a Deep Thinking run. Spend G by default; reach for DAG when the work demands it.
The Rae Loop & Clarity Stars
Rae — your AI companion — grades every meaningful interaction for clarity: did you say exactly what you wanted, with the context the AI needed? Clear prompts earn Clarity Stars; vague ones earn gentle coaching instead. The loop is a mirror, not a judge: stars reflect how precisely you're communicating, which is the single skill this whole Academy teaches. Watch your star trends over time — rising stars mean your prompts are getting sharper, and sharper prompts cost fewer retries.
Message Editing
Sent a prompt and immediately spotted the missing context? Don't rewrite from scratch — long-press your message to edit it in place. The conversation re-runs from the correction, so the AI answers the prompt you meant, not the one you sent. Editing is the cheapest debugging tool in Uzu: fix the instruction at the source instead of layering corrections on top of a misunderstood reply.
The History Drawer
Every conversation is preserved and one gesture away. The history drawer organizes past chats by recency and lets you search across everything you've discussed. Two habits make it powerful: name your sessions by project when you start them, and revisit old threads instead of re-explaining — the AI remembers the context window of a thread you continue, but not one you abandon and re-type.
Attachments: YouTube & PDFs
Paste a YouTube link and Uzu works from the transcript; attach a PDF and it reads the document. Attachments turn the AI from a conversationalist into a research partner — summarize the video, extract the clause, cross-reference the report. The rule of thumb from the Academy applies: attach the file when the content is the task, and say what you want done with it. Each attachment consumes compute, so bring the document when you'll use the document.
Personas
Personas are persistent AI characters — a standing set of role, rules, tone, and knowledge boundaries that runs beneath every message in a session. Instead of re-describing "the strict editor" each time, you enter a session where that editor already exists. Personas run on premium DAG-Credits because the model holds their full rulebook in context for every reply. Build them well, and you can even publish them to the Trending marketplace — where other users' sessions pay you a revenue share.
The Memory Toggle
Memory is Uzu's long-term context: facts, preferences, and project details the AI should remember across conversations. The toggle gives you explicit control — memory on means the assistant accumulates knowledge about you; memory off means a clean slate that forgets when you close the thread. Use it deliberately: on for ongoing projects and personal context, off for one-off questions you don't want following you around.
The Memory Manager
Everything the AI remembers, you can inspect. The Memory Manager lists stored memories and lets you edit or delete any of them — because context you can't correct is context that slowly goes wrong. Review it periodically: delete stale facts, fix drifted details, and prune what no longer matters. A lean memory is a sharp memory; the manager is how you keep the desk clean.
Time Travel: Your Personal RAG
Time Travel is retrieval across your own history — ask "what did we decide about the launch date?" and Uzu searches your past conversations and memories to answer from your documented context. It's the difference between an assistant with amnesia and one that was actually in the room. The richer (and cleaner) your history and memory, the smarter Time Travel gets.
Smart Routing
Not every question needs the same engine. Smart Routing matches each message to the right tier of compute — quick factual lookups go fast and cheap, complex reasoning gets escalated to heavier models. It's the economy working silently on your behalf: you ask, the system decides what the question costs to answer well. You can always override by explicitly choosing a model or persona for a session.
Triangulation
Triangulation cross-checks an answer by approaching the question from multiple angles — different phrasings, different framings — and comparing where the results agree. Agreement across independent paths signals reliability; divergence flags a shaky answer worth auditing. It's the app-level version of the Auditor technique from the Academy: trust answers that survive being asked more than one way.
Deep Thinking
Some questions deserve more than a reflex. Deep Thinking mode gives the model room to work through a problem step by step before answering — slower and costlier, and dramatically better on math, logic, planning, and analysis. The professional pattern: draft with the fast path, then send the hard kernel through Deep Thinking. Reserve it for the reasoning that actually pays for itself.
Hands-Free Mode
Hands-Free turns Uzu into a voice conversation — speak, and Rae answers aloud, ready for your follow-up without touching the screen. It's built for walks, commutes, and cooking: the full loop of listening, transcribing, thinking, and speaking in one continuous session. Barge-in is supported — you can interrupt mid-answer the way you would with a person.
Privacy & Your Data
Uzu is local-first: your conversations and memories live on your device by default. When you sign in, content syncs to your private cloud account protected by row-level security — only your account can read your records. You can delete anything, anytime: individual memories in the Memory Manager, entire sessions in history, or your whole account in settings, which removes your server-side data. The full policy lives at uzuapp.com/privacy.
Evolution Trees
Evolution Trees are your growth made visible — a branching map of the skills you've actually practiced, from your first one-shot prompt to advanced techniques like chaining and auditing. Each completed Academy goal and verified skill grows a branch; each branch shows what you've proven, not what you've merely read. They're the antidote to tutorial-watching without learning: nothing grows until you've done it.
The Carbon League
Every AI message has a real-world footprint — the compute behind your answer consumes energy. The Carbon League makes that invisible cost visible and turns efficiency into a sport: leagues rank players by how much they accomplish per unit of impact, and Uzu funds eco-offsetting compute based on aggregate activity. Waste tokens, and you'll see it; do more with less, and you'll climb. It's the economy's environmental mirror — and a leaderboard where frugal prompting wins.
Deep-linked from the Uzu Academy. Each section corresponds to a lesson in the 30-day Masterclass — the concepts behind them live in the article hub.