Show the math

Every number, with the steps that made it.

“AI” in finance is distrusted for one reason: answers without receipts. Here is one answer from the sample book, taken apart. Every framework, every Sibyl answer, and every page of the board pack is built this way. Nothing is typed in by hand.

The question

What is our idle cash costing us?

1 · READ
The question is turned into a computation, not a search. measure = market_value · filter = asset_class is cash · group_by = entity · window = today
› Aggregating cash by entity…› Pacing $34.0M of unfunded commitments: $12.0M of calls in the next twelve months…
2 · SELECT ROWS
5 positions match, across 4 entities and 4 custodians. These are the rows, exactly as the book holds them; nothing is sampled or estimated.
Barlow Holdings LLC · Operating account $14.2M · 0.1% · J.P. MorganBarlow Dynasty Trust · Sweep (money market) $8.1M · 4.1% · Northern TrustBarlow Family Foundation · Foundation checking $3.3M · 0.0% · Charles SchwabPersonal (Joint) · Joint checking + savings $1.8M · 0.5% · Charles SchwabBarlow Holdings LLC · CHF deposit (Zurich) $5.6M · 0.6% · UBS ZurichTotal $33.0M · 3.5% of gross
3 · POLICY AND CLAIMS
The only assumptions are yours, and they live on the ledger with a source, a date, who set them and who approved them. hurdle 6.0% · IPS §4.2 · approved operating reserve $6.0M · minutes of 14 March 2026 · approved
› claims on each entity’s cash over the next twelve months (to 2027-09-11): its share of the reserve, capital calls paced from $34.0M unfunded, dated bills, debt service, an LTV cushion, overhead› an entity that cannot meet its own calls is funded by its owners first; 3 such transfers, $2.49M in all
4 · COMPUTE
forgone = Σ over entities of max(0, cash − claims, after owners fund) × (hurdle − that entity’s own cash yield)
EntityCashClaimsPaid to entities it ownsIdleEarningForgone a year
Barlow Dynasty Trust$8.1M$1.5M—$6.6M4.1%$125K
Barlow Family Foundation (grants only)$3.3M$0.9M—$2.4M0.0%$147K
Barlow Holdings LLC$19.8M$15.4M$2.3M$2.2M0.2%$125K
Personal (Joint)$1.8M$1.6M$0.2M$00.5%$0
Total$33.0M$11.2M$397K
› still short after every owner has paid: Barlow Commercial RE LP $10.95M, Barlow Marine LLC $1.47M, Barlow Ventures LLC $294K, and 8 smaller› each row: cash − claims − paid to entities it owns = idle; idle × (hurdle − the entity’s own yield) = forgone› of the $11.2M idle, $2.4M sits in the foundation and can go to grants only, so $8.7M has no recorded claim on it› before inflation, deployment cost and tax; one entity at a time, never pooled across owners
5 · FLAG
WARNING$33.0M in cash; $21.8M of it is already spoken for: the $6.0M reserve, $12.0M of capital calls and dated bills, and cash owners pass to the entities they fund. $11.2M is idle, $397K a year short of the 6% hurdle after what each account already earns; $8.7M of it has no recorded claim on it, the rest is the foundation's, for grants. Counting cash against the reserve alone says $1.3M: that treats cash the funds will call as idle.
› the method changed on the ledger, it was not edited: the sample proof sealed earlier counted cash against the reserve alone, $27.0M above it, and said $1.3M a year. That proof is kept exactly as sealed at /proof/sample; this engine answers $397K, because cash the funds will call is not idle.
6 · SOURCE CHAIN
What an auditor sees. Every answer carries it; the board pack prints it under every table.
⌁ Cash positions · 5 rows · by entity⌁ Fund commitments · $34.0M unfunded · pacing model⌁ Hurdle rate 6.0% · IPS §4.2 · approved⌁ Operating reserve $6.0M · minutes of 14 March 2026 · approved⌁ Sealed · one fingerprint over the book, the policy, the engine and the question

Sample book: the fictional Barlow family, $955.9M gross across 16 entities, built to show the method. Every figure on this page is computed from it when the page is built.

Why this is the whole product

Three things fall out of showing the work.

DEFENSIBLE

Your CPA can sign it

The rows, the filter and the policy figure are on the page. There is nothing to reverse-engineer and nothing to take on trust.

HONEST

Missing data says so

A framework that cannot find a position, a benchmark or a document reports what it could not find rather than estimating around it.

REPEATABLE

Same question, next week

The computation is stored, not the answer. It runs again against the new book. The number should move; if it does not, you know exactly where it stayed.

The ledger of why

The rate is the family's. On the record, with a date and a signature.

Every policy figure a number depends on, the hurdle, the reserve, the targets, the spending rate, is recorded with where it came from (the IPS section, the minutes, the statement), the date it took effect, who set it and who approved it. A new value supersedes the old one on a date. Nothing is edited and nothing is deleted, so a book from March is always read with March's hurdle.

SEALED

Every answer has a hash

The book's content, the policy rows in force, the engine version and the question. Same four, same hash. Ask again in five years and it lands on the same proof, or the ledger says which of the four moved.

DECIDED

What the family did is on the same page

Act, watch, accept, defer or reject, with a sentence the board could read later and a date to look again. The decision is tied to the proof it answered. That is the institutional memory that usually leaves with the CFO.

SAID OUT LOUD

A default is named as a default

If the family has not chosen a hurdle, the answer says so next to the number, and links to where to set it. We would rather say “not chosen yet” than print our opinion with a dollar sign on it.

Now do it on your numbers.

One export, one question, two business days, no fee.

Run Your First Question FreeRun This One on the Sample Book