The Bar Feels Different Depending on Who's Answering
I used to treat LacPointer as a single thing — open it with Option+Space, type a question, get an answer. That's fine. But it missed something obvious: the way I want AI to talk to me at 9am when I'm debugging a race condition is completely different from how I want it to talk to me at 4pm when I'm burned out and staring at a design decision I've been putting off for a week.
That's the actual problem Personas solve. Not "here's a cute branding feature." It's about the AI meeting you where you are instead of always coming at you from the same angle.
The Five That Ship by Default
LacPointer comes with five built-in personas, all of which you can use immediately and clone if you want to tweak them.
- Default — the friendly general assistant. Good for everything, opinionated about nothing. This is what you get if you never touch the Personas panel.
- Coding — senior engineer mode. Direct, no hand-holding, no filler. You ask it why your reducer isn't updating state and it tells you exactly what's wrong, not a paragraph of "great question, let's explore this together."
- Therapist — non-judgmental, reflective, unhurried. I use this when I'm trying to think through a decision out loud. It doesn't rush to an answer; it asks the right follow-up questions.
- Stoic — calm philosophy in the Marcus Aurelius / Epictetus vein. Surprisingly useful when you're spiraling about a deadline or a bad code review. It reframes things without being dismissive.
- Hype — high-energy, motivational. I won't pretend I use this one every day, but right before shipping something I've been working on for weeks it's exactly right.
The built-in ones are read-only but you can clone any of them as a starting point for your own.
What Actually Changes When You Switch Personas
This is worth spelling out because it's not just a different tone of voice. When you activate a persona, its system prompt replaces the default LacPointer prompt entirely. That means the entire framing of how the AI interprets your messages shifts. It's not a soft overlay.
On top of that, the persona's voice carries into every realtime voice session. LacPointer supports six voices: alloy, verse, ash, sage, coral, and cedar. The Coding persona uses a different voice than the Therapist persona by design — and when you hit Cmd+Shift+V to go hands-free, you hear that distinction immediately. It's a small thing but it matters when you're using voice regularly.
You can also set a greeting per persona — the message that plays when you activate it in voice mode. I have mine set to something short and specific so I know exactly which context I'm in without looking at the screen.
Building Your Own
The real value kicks in when you build a persona for something specific that the five defaults don't cover. Here's the full set of fields when you create one:
- Name — up to 40 characters. This shows in the switcher.
- Avatar — optional image. Useful if you have several custom personas and want to tell them apart at a glance.
- Voice — pick from alloy, verse, ash, sage, coral, cedar. Try them all before committing; they're more distinct than the names suggest.
- System prompt — this is the whole thing. Write it like you're briefing a new hire who will only ever read this one document. Be specific about tone, format, what to skip, what to prioritize.
- Greeting — what the persona says when activated in voice mode. Keep it short.
- Tools allowed — optional. Leave it empty and the persona can use everything: Notion, Calendar, GitHub, Slack, any installed Skills. Or restrict it to a specific subset if you want a focused, distraction-free mode.
I have one custom persona I call "Reviewer." Its system prompt tells it to act like a skeptical but fair senior engineer reviewing my work — find what's wrong first, then acknowledge what's solid. I switch to it when I want to pressure-test a plan before I commit to it. It's blunt in a way the Default persona isn't, but not destructive like asking for criticism in a vacuum tends to be.
A Few System Prompt Tips From Actually Using This
The system prompt field is freeform and that's both the power and the trap. A few things I've learned:
Tell it what format to use. "Respond in bullet points when listing steps, prose otherwise" is a specific instruction that actually changes output. Vague prompts like "be helpful and concise" don't do much because the base model already tries to be that.
Tell it what to skip. I added "never open your response with a restatement of my question" to my Coding persona and the difference was immediate. These small exclusions matter more than positive instructions sometimes.
Restrict tools when focus is the point. If you're building a writing persona or a journaling persona, you probably don't want it reaching for the GitHub skill mid-conversation. Use the tools_allowed list to scope it down.
Clone before editing. The five built-in personas are read-only but cloneable. If you want a Coding persona that's 90% the same but slightly adjusted, clone it and edit rather than starting from scratch.
Switching in Practice
The switcher is one tap inside LacPointer. I keep my most-used personas at the top. The flow I actually run most days: open with Option+Space, switch to Coding first thing, work through the morning, flip to Stoic or Therapist if I hit a decision that needs thinking through rather than executing. It takes about two seconds to switch and the context shift is real enough that it's worth doing.
If you haven't opened the Personas panel yet, it's worth spending fifteen minutes there. Build one persona that's specifically tuned to your most common use case — even a small, specific system prompt is better than the generic default for repeated tasks. That's the version of LacPointer that actually fits how you work.
You can grab LacPointer at lacai.io/lacpointer if you're not already using it.