Pour One Out for Granola

OpenAI announced native data connectors for deep research today — your docs, your meetings, your CRM, your whatever — and I'm mostly thinking about Granola.

Granola is, was, a beautiful little app. Meeting notes that actually understood what happened in the meeting. Ambient, unobtrusive, good at the one thing it did. The kind of product that made you think indie AI tooling had a real future, that there was space to be useful and small and excellent.

There isn't, apparently. There wasn't.

The cycle is familiar by now — some gap opens in what the big models can do, a smart team fills it with a focused product that people genuinely love, and then the gap closes. Not because someone built a better version. Because the surface area of the platform just expanded to include it. No announcement about Granola specifically. No malice. Just geometry.

What OpenAI actually shipped is meaningful for enterprise — deep research that can reason over internal data, agentic queries running against your actual knowledge base instead of the public internet. That part is legitimately useful, the kind of thing that would have required a full integration project six months ago and now just requires connecting a few sources. Real simplification. Worth noting.

But I keep coming back to the meeting notes thing. There's a specific texture to watching a beloved small app die — not from competition, exactly, but from the platform deciding to include it. The team at Granola built something good. It didn't matter.

This is what it looks like when a platform wins. Not a battle. An absorption.