NewsTune Team Reporter

An AI reporter that turns fast-moving team updates into a personalized morning podcast.

Comment

Inspiration

AI has dramatically accelerated how fast teams build, but it has created a new bottleneck: shared understanding. Work changes so quickly that people often do not know what their teammates completed, where each project stands, which decisions were made, or what actually requires their attention.

Meetings and dashboards help, but they still require people to stop, search, and read. We wanted a lighter, more personal way to stay aligned: a morning podcast produced by an AI reporter that continuously follows the work.

What it does

NewsTune Team Reporter turns fast-moving team activity into a personalized audio briefing.

It can:

  • Read connected project updates, team documents, meeting notes, and knowledge sources
  • Compare recent activity and identify what meaningfully changed
  • Surface important progress, decisions, risks, blockers, and information gaps
  • Prioritize updates according to the listener's teams, projects, and interests
  • Produce a private morning podcast that can be monitored, reviewed, and explicitly published
  • Support multi-team executive briefings and organization-wide audio news for companies, schools, and nonprofits

NewsTune is not simply a text-to-speech tool. It acts as an AI reporter: gathering information, applying editorial judgment, establishing context, and explaining what the listener actually needs to know.

How it works

GPT-5.6 in Codex acts as the reporter and orchestration layer. It uses connected tools and MCP sources to gather team context, compare changes, identify missing information, and prepare an editorially structured briefing.

NewsTune Studio provides secure MCP tools for the production workflow:

  • Inspect the connected NewsTune workspace and existing podcast series
  • Review authorized hosts and voices
  • Preview the exact credit cost before any generation
  • Create a private podcast episode only after explicit user approval
  • Monitor the production job until the audio is ready
  • Return the private episode page for listening and review
  • Publish only an explicitly approved series, episode, and RSS scope

The system uses OAuth 2.1 permissions, ownership checks, idempotency controls, private-by-default generation, and confirmation gates for paid or public actions.

How we used Codex and GPT-5.6

Codex was used throughout the Build Week development process to design and implement the MCP architecture, OAuth authorization model, tool contracts, confirmation flows, privacy boundaries, test fixtures, and production-readiness checks.

GPT-5.6 is the core intelligence behind the Team Reporter workflow. It interprets updates from multiple sources, determines what changed and why it matters, detects risks and information gaps, adapts the briefing to the listener, and coordinates NewsTune's MCP production tools.

Built during OpenAI Build Week

NewsTune existed before Build Week as an AI podcast production platform. During the submission period, we meaningfully extended it with:

  • A stateless Streamable HTTP MCP server
  • OAuth-protected read, write, and publish scopes
  • Nine agent-facing podcast production tools
  • Exact-cost preview and explicit-confirmation workflows
  • Safe, idempotent private episode generation
  • Consent-oriented voice workflows
  • Approval-controlled public pages and RSS publishing
  • Reviewer fixtures and automated MCP security and transport tests
  • The Team Reporter workflow for turning changing team knowledge into a personalized audio briefing

The pre-existing podcast generation, audio, and publishing systems provide the production foundation; judging should focus on the new MCP, agentic workflow, safety, and Team Reporter functionality added during Build Week.

What's next

We plan to expand Team Reporter from a single morning briefing into a continuous information layer for individuals and organizations: personalized multi-team updates, proactive follow-up questions, interview workflows, recurring organizational news, and deeper audience-specific editorial memory.

Built With

Share this project:

Updates