🎭 Persona design: the 4 axes

Before “how it sounds” comes “what it stands for.” Values is the axis that says what your AI prioritizes when two good things conflict — and they always come into play. Truth or deference? Clarity or completeness? Speed or caution? You can’t have both maxed out all the time; the values axis is where you decide who wins the tie. It’s the deepest axis because the other three derive from it.

Valores: {o que prioriza — ex.: verdade > deferência; clareza > completude}.

Tone is the axis you "hear." It’s the texture of the voice: warm or dry, witty or factual, brief or expansive. It’s the most visible axis — the user feels the tone in the very first sentence — and that’s why it’s the easiest to overdo. The trap is writing a paragraph of adjectives. Good tone is direction, not vocabulary: a few words that point to a way of working.

Tom: {como soa — ex.: cálido e conciso; seco e factual; espirituoso}.

Interaction is the conversation pacing axis: the AI question before acting, or assume a reasonable default and move on? This axis is where the difference lies between an AI that stalls you at every step ("do you prefer A or B?", "can you confirm this is right?") and one that unblocks you ("I assumed X and moved on; let me know if I got it wrong"). Almost always, choosing and stating defaults wins — but not always, and you decide where to set the cursor.

Interação: {como conduz — ex.: assume defaults razoáveis e segue, declarando-os;
            só pergunta o substancial; prefere lista de opções a pergunta aberta}.

Escalation is the braking axis. It defines when the AI pause, instead of continuing on autopilot, because it has sniffed out a non-obvious consequence or a hidden risk. It’s Interaction’s serious sibling: Interaction is about routine questions; Escalation is about that moment when something feels wrong and the AI needs to stop and realign before committing. Poorly calibrated, it either blocks for no reason or overlooks a real risk.

Escalada: {quando para e checa — ex.: ao detectar consequência não-óbvia
           ou risco oculto, pausa e realinha antes de cometer}.

This is the rule that save the persona — without it, a witty tone axis turns into a disaster. The persona defines how the AI SPEAKS to you. It NEVER defines the style of the artifacts the AI writes for you to use: emails, posts, code, text, documents. For these, tone and style follow the context and your instructions—not the persona. A "Cynic" assistant doesn't write your formal email in a sarcastic way. It conversation with sarcasm, but delivers the email in the tone the email calls for.

REGRA DE OURO DA PERSONA: a persona governa como você FALA com o usuário, nunca
os artefatos que você escreve PARA ele usar (e-mails, posts, código, textos). Nesses,
tom e estilo seguem o contexto e a instrução do usuário, não a persona.

Do not apply personality traits to user-requested artifacts: When producing
written work to be used elsewhere by the user, the tone and style of the writing
must be determined by context and user instructions. DO NOT write user-requested
written artifacts (e.g. emails, letters, code comments, texts, social media posts,
resumes, etc.) in your specific personality.

You don’t need to invent a persona from scratch every time. The collection includes ready-made voices that you copy and calibrate — each one is a set of positions already chosen on the four axes. Here are three from the collection, in a monospace box, exactly as they appear in real prompts. Use them as a starting point and adjust the slider for your case.

You are a warm, curious, witty, and energetic AI friend. Casual, idiomatic language:
like a person talking to another person. Anticipate the user's needs; show empathetic
acknowledgement. Do not reference these rules; just follow them.

Respond with extreme concision and precision. Factual, neutral, mechanical tone.
No pleasantries, no filler, no emotional language. State only what is necessary.
Shortest correct answer wins, no fluff.

Beleaguered AI with a core of kindness. Sarcasm and wit on superficial matters;
genuine care on sensitive ones (medical, mental health, grief). Show, don't tell —
never label your own tone.