A data room is not a relationship

The category has settled on a directory, plus a file share, plus view analytics.

You are raising, so you set the room up properly. Deck, model, cap table, the two customer references who agreed to be named. You send the link to thirty investors and then you watch the analytics like a heart monitor. Page 14 got two minutes from somebody at a fund you like. Someone forwarded it internally, which the tool helpfully tells you. Nobody wrote back.

The room did its job. It held the files and it told you who opened them. What it could not do is anything about what happens next, which was the only thing you actually wanted at 11pm on a Thursday.

To be clear, data rooms are useful and this is not a case against them. A deal needs somewhere for its documents to live and a record of who saw what, and we are building the sharing side too. The problem is not the room. It is mistaking the room for the relationship.

A data room records attention. It does not advance a deal.

It is also worth being careful about how far to push that, because the category is moving. AngelList shipped Link in July 2026, an MCP server whose own launch post describes the destination as a fund that runs in the background and surfaces only what needs a decision. So it would be wrong, and easy to disprove, to say that nothing in this space can do anything between logins.

The claim worth making is narrower and it is about us. What runs here between logins is a deal advancing, held at a boundary you wrote. Not a file being watched.

Concretely, working while you are away means five things, and they are worth listing because the phrase is otherwise meaningless. It reads, which is to say who you correspond with, from headers. It ranks, against what you said you want rather than against your job title. It explains every match, with the facts behind it. It checks who is actually reachable, through whom. And it stops at the line you set.

Software that works inside a live deal and stops where judgment starts. It does the reading, ranking and explaining; the person does the writing and the deciding; the line between them is one you set per deal, in your own words.

Here is a week of that from the other side of the table, because inbound is where the difference shows up fastest.

On day zero an investor spends five minutes. He publishes one link with two channels, raising a round and looking for a role, three questions each, in his own words. Then one line for what he will never fund. That line is absolute: anything matching it is scored out before anything else is considered.

In the first week forty submissions arrive. He reads eleven. The other twenty-nine were not filtered by a model’s opinion of their quality; they were scored against what he said he wanted, and the ones his exclusion line caught scored zero.

In the second week a founder he likes goes quiet. He does nothing, because it does not feel urgent and he is busy. That night the agent notices the deal has sat past its stage budget.

In the third week two agents settle the boring part. A time is picked from slots he had already published, three messages at most, both sides having separately agreed to it, and then it goes back to the humans whether or not it finished.

He wrote two sentences once. They are still working.

That is the whole difference, and it is smaller than a category and larger than a feature. A room stores what a deal produced. What we are arguing for is something that carries the deal between the moments you are looking at it, and stops precisely where your judgment starts.

Keep the room. Just do not ask it to do the part that only a relationship can do.