Companies built data lakes so that everything sat in one place and could be worked through later. Nobody ever built one for a person.

That is what this category is, and the word that decides who wins it is completeness.

What they built

Turnstone is an AI workspace. You connect the applications you already use, including email, calendar, Slack and Drive, and your local folders as well. It turns all of it into what the company calls a second brain, and keeps that updated continuously.

The agents working inside that workspace draw on it, so the context you would otherwise re-explain every time is simply already there.

The founders are Aryan Vij, chief executive, and Jai Bhatia, chief technology officer.

The inclusion of local folders is a detail worth noticing. Most tools in this category connect to cloud services only, because that is straightforward. A great deal of what people actually work on never leaves their own machine, and a second brain that cannot see it has a hole in exactly the place the user cares most about.

Personal and organisational context is one of the most attempted problems in software. Notion and Obsidian are where people keep what they know, with Mem and Reflect pursuing the AI-native version. Glean built a large business on search across everything a company uses. Dropbox Dash and Google Workspace approach it from the file layer. Rewind went furthest by recording everything on the machine. Mem0 and Zep sell memory to developers building agents. Microsoft Copilot has the strongest position of all, because it already sits where the work happens.

The graveyard here is large, and the reason is always the same: these products need everything connected to be useful, and are not useful until everything is connected.

The point: this is a new kind of lake, and the key word is completeness

The idea is simple and that is its strength. Put everything in one place so the AI can work through it afterwards.

Companies have done this for fifteen years. The data lake was exactly that: stop arguing about which system owns what, pour it all into one pool, and let the analysis happen later. It worked because the value did not come from any single source. It came from having them together.

What is happening now is the same move at a different scale. AI has made it worth building a lake for one person, and for one company's ordinary working life rather than its databases.

Watch the language the founders in this category use and you will notice the same word every time. All your files. All your apps. Everything you work on. They reach for "all" because they know where the value sits, and it is not in any one connector. A second brain missing the thing you actually care about is not a partial second brain, it is a useless one, because you will go and look it up yourself and then you may as well not have it.

Turnstone's version of "all" is the reason to look at it. Cloud services are the easy half and everybody connects those. It also reads your local folders, which is where a great deal of real work sits and never leaves the machine.

Who is actually allowed to say "all"

There is an obvious objection here and it needs answering, because if completeness simply meant the largest number of sources then this category was over before it started. Microsoft has your email, your calendar, your documents, your chat and your operating system. Nobody is going to out-collect that.

But a giant is only complete inside its own walls. Microsoft has everything that is Microsoft. Google has everything that is Google. Neither has the other, and neither ever will, and almost nobody works inside only one of them.

So the completeness worth having is not the biggest pile. It is the one that crosses boundaries nobody else can cross: the work email and the personal drive and the team chat and the folder on the laptop, in one place, held by somebody with no reason to keep you inside any particular walled garden. That position is available precisely because the companies with the most data are the least able to take it.

It is also why local files matter more than they first appear. They are the part no cloud giant can reach at all.

The next lake is the one that is not digital yet

This part is opinion rather than anything the company has said.

If the exercise is to get everything into one pool, the obvious question is what is still outside it. The answer is most of the physical world.

What was said in the meeting room rather than typed. What the machine on the factory floor sounded like before it failed. What the shop looked like at four in the afternoon. What the engineer noticed on site and never wrote down. Which shelf was empty. How long the queue was. None of that is in any lake, for any company, and all of it is the sort of thing a person would answer instantly and a system cannot answer at all.

Getting that in is a harder problem and a much bigger one, and it will be attacked industry by industry rather than all at once. Logistics, construction, farming, retail, manufacturing, healthcare, field service: each has its own version of the thing everybody knows and nothing records.

We are going to see a lot of founders building the get-all layer for a particular field, and that is an idea anyone can start on.

Why it is bigger than it sounds

The short description is another note-taking tool. What it belongs to is a category with a long history and a new scale.

  • The data lake worked, and nobody built one for a person. AI is what makes it worth doing, because now something can actually read the pool.
  • Completeness is the whole game. A second brain missing the one thing you needed sends you off to look it up yourself, and then it has no purpose.
  • The giants cannot claim "all". Each is complete only inside its own walls, which leaves the boundary-crossing position open to somebody who sells no walled garden.

What to watch

The measure is how many sources a typical user actually connects. One or two means a tool. Six or seven means a lake, and the behaviour changes at that point because people stop checking the original.

The second is whether local folders stay in the product. It is the hardest connector to support and the easiest to quietly drop, and it is also the only one no cloud giant can copy.

The third is what happens when the pool is stale. One confident answer built on last month's version teaches a user to go and verify everything, and a second brain nobody trusts is just another place to check.