Podcast and news intelligence — read it, query it, or plug an AI into it.
Grubber is a platform, not just an app. A production pipeline transcribes roughly 120 podcast episodes a week — about 1.3 million words — across 70+ shows, ingests wire and public-radio news hourly, clusters both into topics, and measures where podcast conversation and news coverage diverge. Everything below sits on that one corpus.
A native Mac reading desk (with an iPhone/iPad companion) for the whole corpus: a daily Brief, the Newsletter (The Grubber Daily and a bigger Sunday Weekly), a Live Desk of topics in motion, a transcript Library organized for long reading, a Newsroom, and full-text Search across every transcribed word.
Distribution is TestFlight, invite-only, by design. Grubber Desk lives in permanent TestFlight rather than the App Store, and invitations are personal. If you want in, write.
Every consumer — Grubber Desk included — talks to the same versioned HTTP surface: shows, episodes, transcripts, search, news, topics, daily digests, and newsletter editions. Access is per-client keys; there is no other way in, which is what keeps the platform honest. API access is by arrangement.
The corpus speaks Model Context Protocol natively: a hosted MCP server exposes the read surface as tools, so a Claude (or any MCP client) can list shows, pull transcripts, and search the corpus directly. The Mac app carries a second, local MCP server that lets an assistant drive the app itself — open an episode, jump a section, run a search on screen.
Assembled every morning from the day's analysis — and a bigger edition on Sundays — with an editorial lede, the topics where pods and news diverge, and what landed in the corpus. Read it in the app today; delivery by mail is on the roadmap.