Does it help if I only use ChatGPT and never plan to switch?

It helps. You skip re-explaining at the start of every chat: paste your PCP once as custom instructions or use it in custom GPTs. If you ever switch or try Claude, you already have it.

Is there a risk the model treats me "by profile" and gets it wrong?

It is the most serious risk, which is why the PCP ships with an interpretation protocol: the profile is not identity, observed behaviour rules, traits calibrate tone but never prescribe. The proxy reinforces that rule in every call.

What if I am very different at work and in private life?

Your PCP captures stable traits; you add context when you ask. "Do it in work mode" or "in personal mode" is enough for the model to adjust. On the Professional plan we also add contextual extraction: the system learns which domains you use and weights them.

Are my data used to train models?

No. Your profile is injected at runtime as a system prompt and discarded after each response. No fine-tuning on your answers. You can export, delete or cancel the subscription whenever you want.

← Back to use cases

Professionals with a full head

For people who use AI several hours a day and need it to understand how they work — not to parrot what they already know.

You are a consultant, an engineer, a lawyer, a journalist, a designer, a product manager, a doctor. You have internalised that general-purpose AI saves you time. You have also internalised that every conversation starts the same way: explaining yourself. Your sector, your seniority, your risk tolerance, the way you make decisions. You say it once, it forgets. You say it again, it still forgets. Afini does that for you, once, with data.

The pains you bring

Re-explaining yourself every chat

Five minutes at the start of every thread: "I am senior, skip the obvious bullet lists, write in tone X, assume knowledge Y". Multiply by threads per day, by days per year. That is your life.

Replies that flatter instead of thinking

You ask for feedback on a tricky email. You are told it is brilliant. You know it is not. You ask for criticism of a strategic call. You get validation. You know that is wrong too. Sycophancy in general-purpose models is a problem OpenAI and Anthropic have publicly acknowledged. If your profile says "high openness, low agreeableness, comfortable with disagreement", the model knows and pushes back.

Switching models and starting from scratch

ChatGPT memory does not export to Claude. Claude memory does not reach Gemini. Every time you try a new model or switch for price, capability or availability, you start at kilometre zero again. Your portable JSON profile does not.

Internet personality tests are useless

MBTI, enneagram, colour codes. You know them. You also know serious psychology dismisses them. You want data. Big Five with norms from over a million respondents, attachment with a validated scale, values with our own instrument. Not horoscopes.

What Afini solves

One single session, 60–300 items

You start with the Big Five (IPIP-NEO, public domain, Johnson norms from 145,388 + 619,150 respondents). You pick the rigour: 5€ for 60 items or 20€ for the full 300. In 10 minutes you have percentiles. The other layers (attachment, humour, values, time, cognitive style) get added later from the subscription, whenever you want.

Profile auto-injected into every conversation

Your Essential plan (€14.99/mo) includes an authenticated proxy to Claude. The proxy reads your compiled profile and injects it as a system prompt before every call. You see "My AI"; the model sees a full PCP of 800–1,400 tokens. No training on your data: it is only injected at runtime.

Portable JSON: switch to any AI

For 249€ once, you take the PCP home as structured JSON with an interpretation protocol. Paste it as a system prompt in any playground (Claude, GPT, Gemini, Mistral, Llama via Ollama). Your profile stops being hostage to one vendor.

Explicit anti-sycophancy

The proxy injects a directive block reminding the model: the profile is not identity, observed behaviour beats traits, no validation for the sake of it, push back when the evidence calls for it. Your calibration formula uses A (agreeableness) × N (neuroticism) to dial in how much pushback fits — without overdoing it.

What you are probably asking

Does it help if I only use ChatGPT and never plan to switch?: It helps. You skip re-explaining at the start of every chat: paste your PCP once as custom instructions or use it in custom GPTs. If you ever switch or try Claude, you already have it.
Is there a risk the model treats me "by profile" and gets it wrong?: It is the most serious risk, which is why the PCP ships with an interpretation protocol: the profile is not identity, observed behaviour rules, traits calibrate tone but never prescribe. The proxy reinforces that rule in every call.
What if I am very different at work and in private life?: Your PCP captures stable traits; you add context when you ask. "Do it in work mode" or "in personal mode" is enough for the model to adjust. On the Professional plan we also add contextual extraction: the system learns which domains you use and weights them.
Are my data used to train models?: No. Your profile is injected at runtime as a system prompt and discarded after each response. No fine-tuning on your answers. You can export, delete or cancel the subscription whenever you want.

Your case, with your own data

Start my profile

Cancelable, exportable, no training on your data.

Use cases

Couples who want to talk better

Two profiles. One conversation protocol. No sermons about what love is.

Parents who prefer understanding to lecturing

Your kid is not like you. Knowing this does not turn parenting into a manual, but it takes the drama out of half the fights.

Therapists who want an instrument, not a replacement

Big Five with norms from a million respondents, ECR-R, HSQ, ZTPI. Comparable data your patient can bring you done. No AI telling you how to work.

Coaches who prefer data to self-help

For executive, life, sports or team coaches. A serious instrument for serious clients. No DISC, no colour wheels, no promises of radical transformation in six sessions.