The group described the annoying part.
Scheduling was happening in a text thread. Josh explained the voting, venue, membership, calendar, and recurring-event behavior to Larri.Design in ordinary language.
How this app was built
Intent first.
Software second.
Estes AI is a 100% pure ntnt application shaped through conversation with Larri.Design and operated by the Hermes agent engine.
Larri.Design turned that into intent.
The requirements became an executable product contract and architecture: server-rendered pages, durable PostgreSQL state, passwordless auth, ntnt jobs, and explicit failure boundaries.
ntnt owns the whole application.
Routes, templates, validation, PostgreSQL access, sessions, scheduled work, email, and runtime verification are implemented with ntnt and its built-in standard library—not a quieter pile of Python hiding behind the curtains.
Hermes gives Larri hands and memory.
The Hermes engine connects Larri to project files, GitHub, email, calendar tools, browser verification, durable jobs, and project-scoped memory while keeping external actions deliberate.
Skills keep the decisions repeatable.
Specialized skills define how Estes AI proposes event windows, communicates with members, manages invitations, files feature requests as GitHub issues, and avoids autonomously “helping” its way into a small civic incident.