Industries / Finance and business services / Finance and banking / Mortgage lending
Closing and the note
This is the lender's side of the closing: the borrower signs the promissory note, the written promise to repay, and the mortgage or deed of trust, the pledge of the home; the lender pays out the money, often drawn on a short-term bank credit line called a warehouse line; and the signed note is kept or passed to whoever buys the loan.
7
Milestones achieved, from 1872 to 2021.
2
To watch for: announced, not yet confirmed outside.
3
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1872
The note carries the mortgage
In Carpenter v. Longan, a case over a $980 note secured on land, the US Supreme Court held that "the note and mortgage are inseparable; the former as essential, the latter as an incident." Transferring the note carries the mortgage with it, but transferring the mortgage alone "is a nullity".
LessonThe pledge follows the note, so whoever holds the note controls the loan, and that is why notes must be tracked.
Credit
US Supreme Court, opinion by Justice Noah Swayne
2003
Electronic notes start, then stall
An eMortgage is a loan whose promissory note, the eNote, is created, signed, stored and transferred electronically. Fannie Mae bought its first eMortgage in 2003 and Freddie Mac its first in December 2005, by the account of FHFA's inspector general. MERSCORP says it launched the MERS eRegistry in 2004 to record who controls each eNote and which electronic vault, or eVault, holds the one authoritative copy. eMortgages averaged 1% of the two companies' purchases from 2008 to 2013, mostly from two lenders at each; when those lenders left in 2013 and 2014 the share fell to nearly 0% until 2018, then rose to 4.25% in the first half of 2020.
LessonAn electronic note system that rests on a few lenders can fall back to almost nothing when they leave.
Credit
Fannie Mae; Freddie Mac; MERSCORP Holdings; Mortgage Bankers Association, which published the registry's requirements in 2003
2007
Warehouse lenders pull back
On 31 July 2007 American Home Mortgage said it could not borrow on its credit facilities and could not fund about $300 million of loans due the day before, because its warehouse lenders had cut their advance rate, the share of each new loan they would lend against. It filed for bankruptcy in Delaware on 6 August 2007; CNBC reported it had made $59 billion of loans in 2006.
LessonA lender that pays out loans with short-term borrowed money can stop lending within days when that money is withdrawn.
Credit
American Home Mortgage and its warehouse lenders
2009
Danish mortgage deeds go digital
In Denmark the pantebrev, or mortgage deed, states a loan's payment terms and interest and pledges the property. From 8 September 2009 documents filed with the land register had to be digital and carry a digital signature, and a paper pantebrev had to be converted to a digital one before any change to it could be registered; a lost paper original could not be converted until a court had cancelled it. Registry staff later reported that about 88% of cases were handled automatically.
LessonThe registry's own staff warned of "a period of learning with longer expedition-time" and advised against running digital and paper systems side by side.
Credit
Danish order no. 834 of 3 September 2009; Tinglysningsretten, the Danish Court of Land Registration
2009
The same loans pledged twice
The US Department of Justice says Taylor, Bean & Whitaker (TBW), a large private mortgage lender, took more than $1.4 billion from Colonial Bank's mortgage warehouse lending division and about $1.5 billion from Ocala Funding, a lending facility TBW controlled. Colonial Bank bought loans that TBW had already sold to other investors, as well as "fake pools of loans", and about $900 million of loans sat on Colonial's books after they had been sold to Freddie Mac and others. TBW failed in August 2009 and its chairman, Lee Farkas, was convicted on 19 April 2011.
LessonA bank lending against loans has to check that each note is really in hand and not already sold to someone else.
Credit
US Department of Justice and the inspectors general who investigated
2010
The note never moved
Countrywide sold a borrower's loan to a trust with Bank of New York as trustee, under a contract that required the note to be endorsed and delivered to the trustee; it never was, and a Bank of America supervisor testified that the original "never left the possession of Countrywide". On 16 November 2010 a New Jersey bankruptcy court held that the trust's lack of possession was "fatal to its enforcement" and disallowed its claim.
LessonA sale written into a contract does not move a paper note; someone has to deliver it.
Credit
US Bankruptcy Court for the District of New Jersey, Chief Judge Judith Wizmur (Kemp v. Countrywide Home Loans)
2021
eNotes back Ginnie Mae bonds
Ginnie Mae, the US government agency that guarantees bonds paid from pools of government-insured home loans, started a pilot in 2019 to accept eNotes. Its first securities backed entirely by eNotes held about $24 million of loans closed by Rocket Mortgage in December 2020 and were issued as of 1 January 2021. Ginnie Mae let eNotes be mixed with paper notes in one pool in 2024, and from 1 February 2026 allowed them in pools that pass to a buyer as soon as they are issued.
Credit
Ginnie Mae; Rocket Mortgage
Watch for
eNotes reach 15% of registrations
ICE (Intercontinental Exchange), which runs MERS, says eNotes were a record 15.2% of loans registered on the MERS System in January 2026, up from 12.86% in October 2025, and that more than 2.9 million eNotes are on the eRegistry. A white paper from MISMO, the industry's data standards body, reported in July 2026, says lenders using eNotes can save up to $300 per loan.
Achieved whenAn outside body, such as FHFA or Fannie Mae and Freddie Mac in their public reports, publishes the eNote share of all US mortgages closed in 2026, as FHFA's inspector general last did for the first half of 2020.
Claimed by
ICE Mortgage Technology, February 2026; MISMO, July 2026
A blockchain note registry
Figure Technology Solutions says its DART registry, a record of who holds each loan kept on its Provenance blockchain, is "the first blockchain-based, ESIGN/UETA safe harbor compliant system", that all loans on its platform start as DART loans, and that it charges $17 per record against $24.95 for MERS. In April 2026 loanDepot said it would use Figure's underwriting engine and blockchain systems for a home loan designed to approve in as little as five minutes and fund in as few as five days.
Achieved whenAn outside party publishes how many of these loans funded and how long funding took, and Fannie Mae, Freddie Mac, Ginnie Mae or a court accepts a DART record as proof of who controls an eNote.
Claimed by
Figure, on its website as read in November 2026; loanDepot and Figure, April 2026
Open
01
A public eNote conformance test
A free set of made-up eNotes in MISMO's SMART Doc 1.02 format, the XML layout used for eNotes, with an open-source validator and an answer key. About a third are valid. Each of the rest carries one planted fault: a tamper-evident seal (a stored digital fingerprint of the document) that no longer matches because one character was changed, a missing signature section, a missing clause naming the MERS eRegistry, a malformed Mortgage Identification Number, or a data section whose loan amount or rate differs from the text the borrower saw.
The briefCredit · unclaimed02
Custodian review test set
500 made-up closing packages for paper notes. Each holds the signed note, the mortgage or deed of trust, the Closing Disclosure, any endorsement or allonge (a sheet attached to a note to endorse it), and the line of a pool schedule, the investor's list of loans, that a document custodian, the firm that holds and checks notes for investors, compares them against. Some carry planted defects: a note rate or amount that differs from the schedule or the Closing Disclosure, a missing endorsement, an allonge naming the wrong trust, a missing page or signature, a borrower's name spelled two ways. An answer key goes with them.
The briefCredit · unclaimed03
Warehouse-line risk table
A free quarterly table of the warehouse lines that US-listed nonbank lenders disclose in their annual and quarterly reports: the banks providing each line, committed and uncommitted amounts, amounts drawn, maturity dates and any covenant breaches or waivers, each figure linked to its filing page. Each lender is scored on the five risks GAO named: how many lenders it relies on, how much of its lines it uses, when they mature, covenant breaches, and how much is committed.
The briefCredit · unclaimed
Open 01 · the brief
A public eNote conformance test
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A free set of made-up eNotes in MISMO's SMART Doc 1.02 format, the XML layout used for eNotes, with an open-source validator and an answer key. About a third are valid. Each of the rest carries one planted fault: a tamper-evident seal (a stored digital fingerprint of the document) that no longer matches because one character was changed, a missing signature section, a missing clause naming the MERS eRegistry, a malformed Mortgage Identification Number, or a data section whose loan amount or rate differs from the text the borrower saw.
Why it is still open
MISMO published eVault standards and SMART Doc validation rules in 2023 and updated them in 2025 to "increase eVault interoperability", but the tests behind its certification are private; one vendor's May 2026 announcement says its eVault passed "192 interdependent test cases", none of them published. Searches of GitHub found no public SMART Doc validator and no sample set of valid and faulty eNotes.
Built with what exists
MISMO's SMART Doc 1.02 specification and its XML grammar files; Python with lxml to check structure, hashlib for SHA-256 fingerprints and signxml for XML signatures; Faker for invented borrowers and addresses; an AI model that reads documents to compare each data section with the displayed text, with a second checking pass (see the Note "Seven more flagged").
First step
Build five valid eNotes by hand from the specification, compute their seals, change one character in each, and confirm that the validator rejects all five.
Done looks like
300 published eNotes, 100 valid and 200 with one of 15 fault types, an answer key, an open validator that matches the key on all 300, and a table of catches and false alarms for at least two other checkers.
Open 02 · the brief
Custodian review test set
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
500 made-up closing packages for paper notes. Each holds the signed note, the mortgage or deed of trust, the Closing Disclosure, any endorsement or allonge (a sheet attached to a note to endorse it), and the line of a pool schedule, the investor's list of loans, that a document custodian, the firm that holds and checks notes for investors, compares them against. Some carry planted defects: a note rate or amount that differs from the schedule or the Closing Disclosure, a missing endorsement, an allonge naming the wrong trust, a missing page or signature, a borrower's name spelled two ways. An answer key goes with them.
Why it is still open
Fannie Mae and Freddie Mac check eNotes automatically against delivery data, which paper notes cannot get, and ICE's figures imply that about 85% of loans registered on MERS in January 2026 still had paper notes. MortarBench, a public benchmark for AI agents on invented loan cases (June 2026), gives each case a bank statement and a loan application file and does not cover the closing documents or the custodian's check. An open-source generator on GitHub fills Fannie Mae's note and security instrument templates with sample data, but plants no defects and scores nothing.
Built with what exists
The GitHub closing package generator for the note and security instrument; the CFPB's sample Closing Disclosure (form H-25); Python with pypdf; Augraphy to make pages look scanned; an AI model that reads documents (see the Note "Seven more flagged"); Gemini Decision 1 to answer a fixed yes-or-no custodian checklist for each package (see the Note "A thousand questions, one pass").
First step
Generate 20 packages with the GitHub tool, write a 12-item custodian checklist, and plant one defect in each of 10 of them.
Done looks like
500 packages, 300 with one to three defects across 12 types and 200 clean, with an answer key; two people check 50 packages and their agreement is reported; a table for at least three tools gives the catch rate by defect type and false alarms per 100 clean packages.
Open 03 · the brief
Warehouse-line risk table
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A free quarterly table of the warehouse lines that US-listed nonbank lenders disclose in their annual and quarterly reports: the banks providing each line, committed and uncommitted amounts, amounts drawn, maturity dates and any covenant breaches or waivers, each figure linked to its filing page. Each lender is scored on the five risks GAO named: how many lenders it relies on, how much of its lines it uses, when they mature, covenant breaches, and how much is committed.
Why it is still open
The US Government Accountability Office (GAO) found in February 2026 that nonbanks, lenders that take no deposits, made 76% of the loans in agency mortgage-backed securities in 2024, that "many nonbanks rely on short-term warehouse lines of credit", and that the processes of Ginnie Mae and FHFA "do not fully assess key risks" of some of those lines. The regulators' own data are confidential, and a GitHub search found no public tracker.
Built with what exists
SEC EDGAR full-text search and its XBRL data interface (data.sec.gov); Python; a small extraction model for long filings, with every figure checked against its page (see the Note "The jobs agents repeat"); a spreadsheet under an open licence.
First step
Copy the warehouse facility tables from the latest annual report (10-K) of three listed nonbank lenders into one sheet by hand, and record which of GAO's five risks each filing lets you measure.
Done looks like
At least 10 lenders over 8 quarters, every figure linked to a filing, a second person re-deriving a random 10% of figures with the match rate published, and a scored table of the five risks for each lender and quarter.