We Built the Memory Work Has Always Needed

For years we have built Groupthink on one bet: the conversation is the system of record. Work happens in talk. The context that runs a company lives there, in the decision someone made out loud, the thing said in passing that turned out to be the whole story. And until now, it evaporated the moment the meeting ended.

This month two investors wrote the same thing down, a week apart. David Haber at a16z put it plainly: everything is recorded now, and the next system of record is being built out of the living context inside how people talk. Sarah Guo named the part that matters most the untrainable, the value models can’t reach because it only shows up after surviving the real world, the moat capital can’t collapse. We think they are right.

We would add one thing. The untrainable is not a fixed reserve the models are catching up to. It refills. When people talk they invent, they say things that did not exist a minute before and that neither of them could have written down alone. That never stops. So a record built on real conversation keeps getting richer for as long as people keep having them, and the thing it is made of is the one thing a model can never generate on its own.

We learned the rest the hard way. Being right early is the same as being wrong. Before we narrowed Groupthink to a notetaker, we prototyped most of this: an AI that drafted a company’s handbook from its own documents, a demo that pulled a team’s OKRs straight out of their all-hands instead of making anyone type them, AI coworkers that sat in meetings. It was too early. So we retreated to a “good enough” notetaker and spent a year in the commodity fight everyone warned us about.

What we got wrong was never the technology. It was timing, and timing is a question about user behavior. None of this works until people record their conversations by default, and back then they didn’t. Then the same wave that commoditized us fixed it: every notetaker that shipped trained a few million more people to record everything they say at work. Frustrating to live through, and it is the ground the whole vision stands on. The notetakers poured the foundation and stopped on top of it. We can’t stop there. Recording is the on-ramp. The memory is the point.

So, the Groupthink we are building. Most of it works today.

Your record is personal and it compounds. Bring your notes from whatever you already use, forward a recap, paste a transcript, type one line like you’d text a friend, and the people in them get richer in your record without you filing anything. The morning before a meeting, the context you need comes to you.

Then you talk to it. Ask your own record what you decided with someone back in March, or when a relationship started to cool. Your tools ask too, through an open connection any assistant you already use can read, so your context follows you instead of living in one more app.

Then it federates. The record you own composes, with your consent, across a team and across the tools each person already lives in. A company’s living memory, grown from the bottom up out of the records the people made and own, instead of installed from the top down over their heads.

All of it stays private by default, and it stays yours. It answers to the person before the company, and we wrote that into the data model, not a setting we can flip later. In one version of this future, your employer records everything you say and owns the file. In ours, you do. We are building that one.

And it keeps climbing. When nothing a team learns hits the floor and the untrainable keeps refilling, the limit stops being how much people can hold in their heads. Companies get to do more with smaller teams carrying more ambition and range than used to be possible, because less of the day goes to remembering and re-deriving and more of it goes to the work only people can do. The ceiling rises, and then it rises again.

We made this for you. Come see what it can do.