Vibe Coding a Live Collaborative Assistant for Team Meetings

By Will Melton · January 17, 2026 · 1:10:42 read

Category: Business & Economy

Show: Live Vibe Coding With Will Melton

In this live vibe coding session on Twitch (twitch.tv/willlmelton), I’m building an AI-powered meeting assistant from the ground up—live, unscripted, and in real time. Chapters: The core idea is simple to explain and hard to execute well: an AI participant for meetings that listens quietly, transc

In this live vibe coding session on Twitch (twitch.tv/willlmelton), I’m building an AI-powered meeting assistant from the ground up—live, unscripted, and in real time.

Chapters:

The core idea is simple to explain and hard to execute well: an AI participant for meetings that listens quietly, transcribes everything as it happens, and only responds when explicitly invoked by name or by a button press. No interruptions. No constant prompts. Just a system that respects how people actually collaborate while capturing the information that usually disappears the moment a meeting ends.

Most teams face the same issue: meetings generate decisions, tasks, questions, and commitments, but those outputs rarely make it cleanly into the systems where work continues. Notes are incomplete, context is lost, and follow-ups fall through. This build focuses on reducing that gap between conversation and execution.

In this session, I start with product framing and then move directly into implementation using AI-assisted development tools. The emphasis is on building real foundations, not a throwaway demo. That means data models, authentication, permissions, consent handling, and extensibility are all part of the work you see happening live.

What’s being built and shaped during this stream:

• A meeting-first system designed for in-person and browser-based use

• Real-time audio capture and transcription

• Silent-by-default AI behavior with explicit invocation

• Structured outputs like tasks, decisions, summaries, and key points

• Organization-based accounts with role-aware access

• Admin tooling for users, plans, and settings

• A global navigation system connecting product and admin views

One of the most important pieces built early is a product roadmap that lives inside the product itself. This roadmap is not a simple checklist. It tracks priority, status, readiness, dependencies, assignments, comments, and attachments. Items move through stages like information needed, in planning, planned, in progress, in review, and released. The roadmap becomes the system used to manage the build as it happens.

By the end of the session, the roadmap is populated with nearly twenty prioritized items representing the remaining work. Foundational items like organization multi-tenancy and the meetings data model are treated as blockers because they unlock everything else. Features that depend on those foundations are staged for upcoming sessions, while items requiring business decisions are clearly marked so they don’t get buried.

The architecture assumes responsible handling of recording consent. Location-aware prompts and visible recording indicators are designed in from the start to account for different state consent requirements. Control is a major theme throughout the build: users choose when the AI speaks, what it remembers, and how much it’s allowed to do.

What comes next after this stream:

• Finalizing organization-level data isolation

• Completing the meetings data infrastructure

• Wiring up real-time transcription end to end

• Adding artifact generation with editing and approval

• Integrating Stripe for trials and paid plans

• Beginning confirmed execution workflows

• Preparing the foundation for future integrations

You can follow the build and related projects at:

https://studio.willmelton.com

Here you can find the Roadmap Replication Guide, a copy and paste prompt for Lovable to create your in-app product roadmap: https://github.com/WillMelton/lovable-code-guide/blob/main/ROADMAP_REPLICATION_GUIDE.MD Be prepared to optimize the roadmap to your preferences.

If you’re in Richmond and want to connect with the local AI community, check out:

https://AIReadyRVA.com

Follow @AIReadyRVA

This session is built in collaboration with ChatGPT and Lovable, using AI both as a planning partner and a development environment.

If you’re interested in building real products in public—especially where AI, workflow design, and software architecture intersect—this stream is for you.

Live on Twitch: https://twitch.tv/willlmelton

More builds and breakdowns coming here on YouTube.