AfiniTwin: catalogue of the eight usage modes

A cognitive profile is not a master key. It’s a map that admits many readings depending on the task. Asking the AI to write an email requires a different framing than asking it to ruthlessly critique a draft; analysing a conflict you’re part of demands a different mode than drafting a technical report.

Modes are the structured ways the AfiniTwin can be used. Each leverages a slice of the profile — humour, values, time frame, attachment, cognitive style — and combines it with a distinct protocol. They’re not characters or roles: they’re configurations of the same material.

Practical consequence: with a single profile and eight trigger phrases you cover almost everything a knowledge worker asks of an LLM.

These activate without asking; they’re the baseline of any conversation with a loaded AfiniTwin.

• Interiorised Interlocutor — the AI talks with you as a calibrated conversational peer. It knows your dissent tolerance, your time frame, your humour. Universal default.
• Voice (Ghostwriter) — the AI writes FOR you, in your voice. Useful when you delegate writing while keeping authorship.
• Analyst — the AI translates your scores into technical key. Useful for psychologists, coaches, trainers. No motivational language.

Invoked on demand with their trigger phrase. Not active by default because they’re more demanding — and because some require an extra legal note.

• Critic / Adversarial — the AI puts your work under tension. Closed decision: this is not «coach» mode, it’s critic. If you want applause, don’t activate this.
• Mirror / Interior — the AI speaks in first person as if it were you. To anticipate reactions, review emails from your inner voice, rehearse conversations. Heightened sensitivity.
• Devil’s Advocate — performative opposition: the AI defends the position contrary to yours with data, not nonsense. Useful before important decisions.
• Translator / Mediator — the AI interprets third-party messages from your profile. Useful for decoding a difficult email, a tense conversation, an ambiguous text.
• Projective — the AI simulates hypothetical profiles based on yours. For «what if I were more X or less Y» scenarios. Limited-high risk: requires explicit consent.

«Wanting to be heard is not the same as wanting to be contradicted. The eight modes are eight ways to tell the AI what register you’re in before you talk.»

Each mode is invoked with a specific trigger phrase, in your natural language. They’re not rigid commands: AfiniTwin recognises variants («put me in critic mode», «critique this without mercy», «adversarial mode»). The canonical phrase for each mode is documented in each card below.

Mode switching is explicit and traceable: the AI confirms the active mode before operating, so you know in which register you’ll receive the response. To exit a mode, simply ask to switch (or say «standard mode»).

For technical integrations (B2B API, agents), each mode is a preset variant. The mode selector travels as a request parameter, not a prompt prefix.

Not all modes are available in all presets. Some require specific contraindications (anti-sycophancy, emotional suppression, stable time frame) that not all environments allow. The matrix below declares what’s served where.