LacPointer Personas: One Bar, Many Brains
For the first few weeks I used LacPointer, I kept the default persona running for everything. Coding questions, journaling prompts, quick calendar lookups — all the same voice, same tone, same vibe. It worked fine. But "fine" is kind of annoying when you know there's a better setup sitting right there in the settings.
Personas are one of those features that look like a fun cosmetic option until you actually use them properly. Then they stop being a novelty and start being the thing you reach for before you even open the bar.
What a Persona Actually Is
A LacPointer persona has five moving parts:
- Name — up to 40 characters, shows in the UI when the persona is active
- Avatar — optional image, mostly for visual context switching
- Voice — one of six:
alloy,verse,ash,sage,coral, orcedar - System prompt — completely replaces the default LacPointer prompt. This is the real lever.
- Greeting — shown when you activate the persona in voice mode via
Cmd+Shift+V
There's also an optional tools_allowed list. Leave it empty and the persona can use all your installed skills. Restrict it to a specific set and you've got a focused tool that can't go off-script.
The detail that matters most: the voice and system prompt aren't applied on top of LacPointer's default behavior — they replace it. Every chat message and every voice session goes through the active persona's instructions. There's no bleed from one persona into another.
The Five Built-in Personas
LacPointer ships with five read-only personas. You can't edit them directly, but you can clone any of them as a starting point for your own.
Default
Friendly, general-purpose assistant. This is what you get if you never touch the Personas tab. Good for everything, optimized for nothing specific. I use this for quick one-off questions where I don't want any particular personality layered on top.
Coding
Senior engineer mode. Direct, no hand-holding, no "Great question!" before answering. It skips the preamble and gets into the actual code or explanation. I switch to this when I'm debugging something in a flow state and I don't want the AI to warm me up — just tell me what's wrong.
If you pair the Coding persona with Caps Lock (the Wand), you get a fast loop: highlight code on screen, Caps Lock, get a blunt technical read on it. No ceremony.
Therapist
Non-judgmental, reflective, slow and warm. Uses a different pacing — more questions, less rapid-fire answers. I won't pretend I use this one daily, but when I'm stuck on a decision and need to think out loud rather than get an answer, it actually helps. The AI doesn't try to fix things. It just reflects back what you're saying until you figure it out yourself.
Stoic
Calm philosophy, heavy Marcus Aurelius / Epictetus influence. You ask something, it answers through a lens of what's in your control and what isn't. Weirdly useful for processing work frustration without spiraling. Also good for the occasional "am I overthinking this?" sanity check.
Hype
High-energy motivational coach. I use this exactly once per day: right before I start a big task I've been avoiding. Hit Option+Space, switch to Hype, say "I need to finish this PR today," get unreasonably encouraged, close the bar, get to work. Sounds silly. Works.
Building Your Own Persona
The built-in five cover a lot of ground, but the real value is in making something specific to how you actually work. Here's how I think about it.
Open LacPointer settings, go to the Personas tab, and hit the button to create a new one. Pick a name that means something to you — you'll be reading it in the UI every time you switch. Then write the system prompt. This is where it's worth spending five minutes.
A weak system prompt reads like a vague instruction: "You are a helpful assistant who is good at writing." A good one is specific about behavior, not just role: "You edit my writing by cutting, not adding. Flag anything that sounds like filler. Never suggest I add an intro or a conclusion — I already have those. Give me line-level suggestions only."
That specificity is what makes a persona actually useful instead of just a named version of the default.
A few personas I've built and use regularly:
- Reviewer — cloned from Coding, system prompt focused on code review only. Restricted
tools_allowedto nothing — no skills, just analysis. Voice set toash(a bit more measured than the default). - Brief — answers in three sentences or fewer. No context, no caveats, no "it depends." Set this as my go-to for factual lookups. If I need more, I ask a follow-up.
- Stand-up — knows my team's Slack handle format and how we write updates. Every morning I hit
Option+Space, switch to Stand-up, and say "yesterday I finished X, today I'm doing Y, no blockers." It formats a clean message I can paste straight into Slack.
Voice Baking Is the Part People Miss
When you activate a persona and open voice mode with Cmd+Shift+V, the selected voice and system prompt go into the session via an ephemeral token. That means the voice isn't just cosmetic — it's part of the model's behavioral context for that session.
In practice this means the Stoic persona doesn't just answer in Marcus Aurelius terms — it answers in sage voice, which has a slower cadence that actually fits. The Hype persona uses verse, which carries energy. These aren't random assignments. They were chosen to match the persona's tone.
When you build your own, pick the voice last, after you've written the system prompt. Read the prompt out loud in each voice and see which one doesn't clash. It matters more in voice sessions than in text ones, but it's still worth getting right.
Switching Mid-Day
You switch personas in one tap from the Personas tab or directly from the bar UI. There's no cooldown, no session reset required. I probably switch two or three times a day depending on what I'm doing — code review in the morning (Coding or Reviewer), a Stoic check-in after lunch, Stand-up before the team call, Brief for anything where I just need a fast answer.
The one thing worth noting: if you're mid-conversation and you switch personas, the new persona's system prompt takes over, but your existing messages are still in context. Sometimes that's fine. Sometimes you'll want to start a fresh chat so the new persona isn't influenced by the previous context. Just close and reopen the bar to start clean.
A Practical Starting Point
If you've never touched Personas and you want one that immediately earns its keep: clone the Coding persona, rename it something like Blunt, and add this to the system prompt — "Keep every response under 100 words unless I explicitly ask for more. No intros. No summaries. Answer the question."
Use that as your default for a week. You'll find yourself switching to something warmer occasionally, but you'll also realize how much of what the default persona says is padding. The persona you set shapes what you expect from every session. Set it intentionally.
You can download LacPointer and start setting up Personas at lacai.io/lacpointer. Free tier gives you enough room to try all five built-ins and build a couple of your own.