A persona is a saved set of instructions that changes how the AI answers — its role, its structure, its voice. The expert prompting is baked in, so a plain question gets an expert answer. Personas never change where facts come from: the scope picker still decides that, citations still work, and the grounding rules always win.
Click the ✦ Persona chip beside the scope picker in the AI drawer and pick from the built-ins — thirteen of them, each with its own starter suggestions:
A persona lasts one conversation: + New returns to the plain assistant, so yesterday's Devil's Advocate never colors today's answers. Each answer is tagged with the persona that wrote it, and ⇄ Second opinion under any answer re-asks the same question as a different persona — the fastest way to feel what they do.
Each built-in carries scope-tuned instructions: on My wiki the Devil's Advocate quotes the places your own notes undercut the idea, while on Anything it argues from base rates and counterexamples at full strength. You never see this switching — you just pick the persona.
Manage personas… (at the bottom of the picker) creates, edits, and deletes your own. The editor gives you a scaffold — role, goal, rules, tone — plus optional per-scope variants, your own starter suggestions, and a model pairing so a persona can always answer from a specific installed model. Editing a built-in opens an editable copy, so you can see exactly how they're written. Best of all: ✦ Draft the instructions with AI — describe the persona in a sentence and the local model writes the instructions, the wiki-grounded variant, and six starter suggestions for you.
Edit with AI and Synthesize notes each have a compact ✦ Voice menu — the same personas, applied to rewriting and synthesis. A ✍️ Writing Partner synthesis reads very differently from a 🔬 Research Analyst one; the sources and citations stay identical.
When you save an answer as a page, it notes which persona wrote it — so future-you knows the 😈 Devil's Advocate was doing the talking.