Somewhere in your fund, a fee is being calculated against the wrong terms. Not because anyone was careless — because the document that says what's right doesn't actually exist.
Ask a simple question: what's the carry on Fund II? The answer seems obvious — it's in the LPA. Except it isn't, not really. The LPA set carry at 20%. Then Amendment 1 added a hurdle. Then a side letter with your largest LP cut their carry to 15%. Then Amendment 2, signed at final close, changed the waterfall for everyone. The number you actually owe each LP lives in none of those documents alone. It lives in all of them, read together, in the right order.
That reading happens in exactly one place: the head of whoever negotiated the deal. And every quarter, someone reconstructs it from memory, a folder of PDFs, and the hope that they didn't miss an amendment.
An LPA is not a document. It's a running ledger of overrides that begins decaying the moment it's signed. The original terms are the least reliable thing in the file, because everything written after them was written specifically to change them.
This is how funds end up paying carry they don't owe, calling capital under a superseded formula, or honoring a co-investment right that a 2023 amendment quietly removed. Not fraud. Not incompetence. Just the slow drift between what the documents say and what any one person can hold in their head at once.
The terms governing your fund are real, binding, and legally precise. They're also scattered across a dozen documents and reassembled from memory every time they matter.
The instinct is to fix this with storage. Put every LPA, amendment, and side letter in a tidy data room. Tag them. Search them. But a folder of the right documents is not the same as knowing the answer. Search finds the side letter. It doesn't tell you the side letter overrides clause 8.3 of the LPA, which was itself amended in 2022, and that the net effect on this LP's distribution is 15% over an 8% hurdle. Retrieval hands you the inputs. You still do the reasoning — every time.
A capable AI doesn't escape this either. Hand a model five documents and ask for the effective carry, and it will read all five and give you a confident, fluent answer that may quietly miss the side letter — because nothing in the documents announces that this one LP negotiated an exception. The exception isn't in the text. It's in the deal.
This is the kind of problem FundOS is built to hold. Not a smarter folder — a layer that reads your LPA, every amendment, and every side letter, and carries the terms that are actually in effect. Per fund. Per LP. With the override chain intact, so when the carry on Fund II is really 15% for one LP and 20% for the rest, the fund knows it — and can show its work.
So when a distribution waterfall runs, it runs on Amendment 2's formula, not the original. When that LP's statement is cut, it reflects their side letter, not the standard terms. When an LP's counsel asks why a number is what it is, the answer is a single trace — not a week of someone re-reading the file. The understanding belongs to the fund, instead of being rented from the one person who remembers the negotiation.
Your documents are the record of what was agreed. A fund brain is the record of what's currently true.
Every fund's terms start drifting out of date the day the ink dries. The only question is whether reconstructing them is a quarterly fire drill performed from memory — or something your fund simply knows, because it has read the whole stack and never forgets an amendment.
Your LPA is already out of date. The terms that replaced it should live somewhere more reliable than someone's memory.