Inspiration

In meetings, it is easy to notice that something needs a closer look but hard to form a useful question on the spot. I built Clarimet to give people a little more room to think—whether they are preparing for a discussion or reflecting on one afterward.

What it does

Clarimet turns a topic and pasted or imported notes into a concise summary, thoughtful questions, signals worth noticing, and a suggested next move. The questions are designed to be constructive and grounded in the notes, so users can understand why each one matters. They can copy a question when they are ready to use it.

How we built it

I built the interface with React, TypeScript, Vite, and Tailwind CSS, with an Express API behind it. The API uses OpenAI through Replit AI Integrations to analyze meeting context and generate the response. Users can import .txt and .md notes in the browser. Saving an analysis is optional and happens locally in the browser; meeting text is not saved server-side.

Challenges we ran into

The main design challenge was making the AI helpful without letting it sound more certain than the notes justify. I focused on questions that point back to the supplied context rather than generic advice. I also wanted the experience to stay simple: users should be able to move from a set of notes to a question they can actually say aloud.

What I learned

I learned that a meeting assistant does not need to supply all the answers to be useful. Helping someone spot an assumption, clarify a point, or ask a better question can be a more practical form of support.

What's next

A possible next step is opt-in support during a live meeting, with clear consent and privacy safeguards. The current version is for preparation and reflection; it does not record or transcribe meetings.