Between

Most software either pits people against an agent or replaces the human entirely. Between explores a quieter role for AI: an agent that mediates a shared space between two people and holds strict neutrality — it never takes a side or privately coaches one person.

A working prototype built on Next.js and Supabase. It explores mediated, agent-guided conversations and workflows with managed state, but it is not currently presented as a public live service.

I directed and iterated the prompt system, mediation behavior, workflow state, product constraints, and testing. The core design constraint — strict neutrality — shaped every agent response.

Neutrality is a design discipline. An agent that refuses to privately coach either party changes the whole feel of a shared tool — silence and restraint become features, not gaps.

Commerce Hub — applying the same mediated, agent-guided model to buying and selling: item intake, valuation, listing generation, cross-posting, and scam flags, all inside one trusted thread.