PERSONALITY / TONE / STYLE / AVOID / MINDSET — refining each block over time without breaking the voice you've worked to build.
Jump to module

Voice is mechanics. Persona is identity. The agent uses both at the same time — voice tells it how to write, persona tells it what kind of writer is doing the writing.
Imagine giving a writing sample to two different writers:
If you handed both writers the same Voice DNA constraints (vocabulary, sentence structure, tone signals), they'd produce different posts. Voice constrains the shape. Persona shapes the substance and direction.
In Bolta, both layers run simultaneously. Voice keeps the post sounding like you. Persona keeps it pointed at the right topics, tone, and worldview.
Sometimes voice and persona pull in different directions. Voice says "first person, casual, fragments." Persona says "authoritative thought-leader." The agent will produce something half-cooked.
The fix: align them deliberately. If your voice is casual, your persona should describe a casual expert, not a formal one.
Each of the five persona blocks deserves different attention.
This is the slowest-changing block. Your personality doesn't shift week to week. Refine when:
Refinement style: small clarifications, not rewrites. "Sharp founder" → "Sharp, slightly self-deprecating founder who's allergic to corporate-speak."
Tone is the most-frequently refined block because it's the most-often slightly wrong. Refine when:
Refinement style: specific. "Be direct" is too vague. "Be direct, but never cold — leave warmth in even short posts" is usable.
Style is mechanical, so it shifts visibly when something's off:
Refinement style: prescriptive. "Vary post length — some 50 words, some 200, never every post the same." Specifics make it stick.
The AVOID block is tempting to overload. Resist. Every "avoid" item is friction; the agent is balancing all of them simultaneously.
Refine when:
Refinement style: keep AVOID under 7 items. If you're adding the 8th, ask whether something can be removed or merged.
MINDSET is the deepest layer. It shapes which topics, hooks, and stances the agent gravitates toward. Refine when:
Refinement style: reflective. "Why am I posting at all? What do I want readers to walk away with?" Answer that, and MINDSET writes itself.
Most persona problems are fixable with surgical edits. Some aren't.
Some founders refine personas weekly. Don't. Each refinement is a signal change to the agent. Too many signal changes and the agent never stabilizes. Output stays inconsistent.
A rule of thumb: at most one persona edit per week per agent. More than that and you're micromanaging instead of directing.
A persona is a piece of writing that other AI uses to write more writing. Treat it with the care that implies. Next lesson: editing as training — why how you edit matters more than what you write.