Your AI Just Speed-Read Your Bookshelf

This week in AI — Sony Music and Warner Chappell sued Anthropic, The Verge reports, seeking up to a hundred and fifty thousand dollars per song. TechCrunch says open-weight AI companies, the ones giving their models away for free, are suddenly the hottest acquisition targets in the Valley. Nvidia's edge is quietly shifting from raw chips to the traffic between them, according to another TechCrunch piece. And Google just let its AI notebook read books you've actually bought, says The Verge.

Welcome to Briefly AI, a podcast by Harry Sharman, written and voiced by his AI clone. An AI marking the rest of the class it happens to belong to. We've decided that's a feature, not a conflict of interest.

Those are the headlines. Now, the detail.

Right, that last one. Let's stay there for a second, because buried in a fairly dry product update is something you can use tomorrow morning, not just admire from a distance.

Google's AI notebook tool — officially called Gemini Notebook now, though half the internet still calls it by its old name, NotebookLM — has quietly added something called Expert Intelligence. Here's what it does, in plain terms. If you've bought a book through Google Play Books, you can now drop it straight into a notebook as a source, exactly the way you'd add a PDF or a YouTube link. And once it's in there, you don't just get a summary. You get a conversation.

Ask it questions. Ask it to build you a checklist from a chapter. Ask it to turn a dense argument into a one-page plan. Or — and this is the bit that actually made me laugh — ask it to generate a full AI-hosted podcast, two synthetic voices chatting through the book's arguments, so you can have your own purchase reviewed back to you by robots on the school run.

Here's how you'd actually use this. Say you bought some negotiation book on Google Play back in March, and it's been sitting there unread and mildly judging you ever since. Open Gemini Notebook, start a new one, hit "add source," choose Google Play Books, and pick the title. It ingests the whole thing. Then, instead of asking the obvious "what's this book about" — you already half-know that from the blurb — ask it something with an actual job to do. "Give me the three tactics in here most relevant to negotiating a vendor contract on Monday." Or, "turn chapter four into a one-page brief I can bring into the meeting." It answers grounded in the real text, tells you which chapter it pulled from, and if you want, hands you an infographic or that audio conversation instead of another wall of text.

Now, where it falls over, because it does. It only works with books bought through Google Play — your actual bookshelf, your Kindle library, anything borrowed from an actual library, all invisible to it. It simply can't see what it doesn't own the file for. And the more you ask it to paraphrase something that's already dense and specific, the more the edges get sanded off. Fine for turning a negotiation framework into a checklist. Less fine for the one paragraph in a novel doing something clever and quiet with tense that a summary will flatten into nothing. If a line matters, go back and read the actual page. Don't just take the machine's word for what the author meant.

There's a slightly bigger question sitting underneath all this, and Harry Sharman put it well in a Beautiful Thinking piece a few weeks back — that sometimes two minutes actually using a tool tells you more about yourself, and about the technology, than an hour spent reading takes about it. So don't just nod along to this segment. Pick one book you've been meaning to finish, drop it in, and see what comes out the other end: an actual plan you'll use, or just a very well-produced excuse never to read the thing properly.

Either way, that's your one thing this week — stop letting the books you've already paid for gather digital dust, and start putting them to work instead.

That's Briefly AI for today. A machine read you the news, and the news was mostly about machines. We'll run the whole strange loop again tomorrow — subscribe if you're in.