Answer the diligence question once, not four times

Two funds ask about model bias in the same fortnight, in different words, and neither of them knows the other asked. A third wants the cohort economics you built in March for somebody who then passed. The grant body wants a quarterly report on a work package whose evidence is the same document a fund is waiting on. Somewhere in a thread from May there is a good answer to all of it, written at eleven at night, and finding it takes longer than writing it again.

So you write it again. Slightly differently, because you are a different person in August than you were in May, and now there are two versions of your position on the same question, both sent to investors who talk to each other. None of this is disorganisation. It is what happens when the only index is you, and you are also the person doing the raise.

Every fund, every grant programme, every milestone, every question and every document is a node. The relationship between two of them is written on the line that joins them, so Nordlys does not merely sit near Model explainability, it requests it. When the second fund asks the same question in different words, you join it to the answer that already exists. There is now one place that question is answered, and both funds are looking at it.

The reasoning lives on the node itself, in plain markdown, and it is the only copy. Nothing is duplicated into a summary that then drifts. When the loss curve assumption changes, it changes in one place and it has changed everywhere it was ever relied on.

Then the part a folder cannot do. Your term sheet is conditional on a licence. The bank will not sign until that licence lands. A grant you have not won yet requires you to hold it. Three facts, learned three months apart from three people, and in a structure they are one visible chain rather than three things you have to be holding in your head at the same moment to notice.

An Oslo company doing AI credit decisioning for small business working capital, raising EUR 7.5M off the back of a convertible bridge taken to reach a regulatory decision rather than raise into one. Six funds are live. One has issued a term sheet, one has passed, one is a strategic asking for terms the lead will not accept. Three non-dilutive programmes run alongside, and the money they bring changes what the round has to cover.

The round is not gated on investor appetite. It is gated on a credit institution licence with an indicative decision date, a partner bank that will not sign until the licence lands, and an independent model validation that three separate audiences are waiting to read. All three are visible as one chain, and so is the awkward part: a term sheet that expires and a decision date that is indicative have four weeks between them in the good case and no relationship at all in the bad one.

The two most useful nodes in the space are the ones a tracker would never hold. One is the fund that passed, kept because a pass is information and the reason they gave is worth having when the next round starts. The other is a founder vesting agreement minuted in 2023 and never executed, which nobody remembered and which surfaced out of a routine cap table reconciliation. Finding that yourself costs a conversation. Having a fund find it costs the conversation and the credibility.

At the end of a partner meeting somebody says send us the detail. What you have is a board they cannot open, a folder they will not read and a deck that was never meant to carry it. So you spend two days building a third artefact out of the first two, and from the moment you send it, it is out of date.

A space publishes as a link or as a single file that opens in an ordinary browser, attachments included. The recipient installs nothing, signs up for nothing and pays nothing. It is the same structure, laid out as a document, with the working notes left out and the reasoning kept in.

It is not a second copy. Nothing is written twice, so the version an investor read last week and the version your team is working from cannot drift apart. Update the answer and republish, and the person who is going to ask you about it in three weeks has been reading the current one.

The seed and Series A deck is a fixed sequence that partners expect in a known order, and the reason it is a rigid form is that it is read by people who look at forty of them a month. It should stay in the tool that makes slides. We are not going to tell you otherwise.

What this replaces is everything behind it: the boards, the trackers and the folder of half-current documents that the deck is assembled from, that nobody outside your company can follow, and that you rebuild by hand every time somebody asks a question you have already answered. The deck is fifteen slides. The raise is nine months.

Nobody wants a single structure holding the entire company. The raise is one space. The product decisions are another, the regulatory work a third, the hiring plan a fourth. Spaces are unlimited, the bridges between them are free, and each holds twenty nodes, which is roughly where a space stops being one job anyway.

That separation is what makes the leave-behind possible. You send the investor the raise, not the company: the pipeline, the evidence and the open risks, without the product argument you had in June or the compensation band you are still deciding. What you publish is a whole space, so the boundary you draw when you make the space is the boundary you can safely send.

Everything about holding the structure is on the free plan. No account, no card, no trial running out, no feature quietly withheld until you upgrade. Attachments are not limited and never will be.

The paid tier is $120 a year and buys exactly one thing: sending a space to somebody who does not have Filamental. On most of this site that is a footnote, because most people never need it. For a founder in a live process it is the entire point, and it costs less than an hour of the lawyer who is going to read what you send.