MNTR

Mentorship does not scale, and every attempt to scale it dilutes the thing that made it work. A good mentor's value is not information — it is the questions they ask and the judgement behind them.

MNTR is an attempt to encode that: AI chat, playbooks, structured courses and a mission system, all shaped around a particular expert's way of thinking rather than around generic advice.

The onboarding does not ask for a job title or a company size. It asks, by name: what are you actually building toward — not the safe answer, the real one, the one that keeps you up at night. And it offers a graceful way out: "Not yet — I'll come back to this."

That is the whole product's thesis compressed into one screen. A mentorship tool that opens with a form is going to get form answers, and form answers are exactly what makes generic advice generic. The escape hatch matters as much as the question — pressuring someone into a half-honest answer would poison everything downstream of it.

The palette is a single signal green on near-black, with a warm off-white for light surfaces. The brand is parameterised as one variable — --brand-hue: 145 — and the rest of the system derives from it.

Restricting the palette to one accent means green always means the same thing: this is the action. On a dark ground with a hundred screens to keep coherent, that consistency is worth more than a wider palette would be.

A hundred screens sit on an Express and PostgreSQL back end — three hundred and thirty-five server files and a hundred and forty-two migrations, with Drizzle holding the schema and a hundred and ninety-three tests holding the behaviour.

The AI is not one model. Gemini and Perplexity do different jobs — reasoning and retrieval — while Stripe meters the credits that are the product's unit of value rather than the seats, and Twilio carries the parts of the mentorship that happen off the platform.