A self-running AI briefing

Manic*
AI

The week in AI, gathered and written by a routine.

Twice a week, a scheduled routine reads the past several days of AI news, picks the stories with real substance, and writes them into a rich briefing. NotebookLM turns that briefing into a two-host podcast.

Listen on Spotify Listen on Apple Podcasts

New episodes every Monday and Thursday.

The pipeline

How it works

  1. Gather

    A scheduled routine scans the week's AI news, from feeds and across the wider web.

  2. Select

    Roughly 12 to 15 stories, variety first. Every item needs real substance.

    Headline-only items get enriched or dropped.

  3. Enrich & write

    Each story gets what happened, the context, and why it matters. Plus the threads and open questions that give the episode a spine.

  4. Listen

    The briefing goes to NotebookLM, which generates the two-host Manic AI audio overview.

The theory

Why "enriched"

NotebookLM's output quality scales with how rich the source is. That one observation shapes the whole pipeline.

Terse roundup in

Headlines and one-liners.

→ generic banter out

Rich briefing in

Context, specifics, tension, open questions.

→ an engaging conversation out

So the routine optimizes for depth and raw material. Concision and delivery are NotebookLM's job.

A note from me
This is a personal experiment. I wanted to see what happens when a routine does the gathering and the writing, so I can spend my attention on listening and thinking instead of doom-scrolling AI news. So far, it works better than I expected. I keep tinkering anyway.

Colin Moroney