From a 10-K to a fund

Spec · ship on the filing

How the strip pays

The product is mezzanine on a human P&L, not a loan with a coupon and not equity. The filing already prices it. The LP still cannot see the path. This is that path, using Maya Chen (The Operator) as the worked example. Build it into the 10-K so the investor return is as visible as the verdict stamp.

• We buy a capped share of underwriting FCF, not a percent of wages and not a claim on assets.
• Check is the largest amount that still (almost) clears the gross hurdle at a share ≤ 30%. Better books get a bigger check, not a higher IRR. Expected IRR on an auto-sized strip will sit on the hurdle. That is by construction, not a coincidence.
• Cap is 2.0× the check. Once that is paid, the strip dies. Term is eight paying years if the cap never hits.
• Gross hurdle 19.5% = 10% LP-net + 3% fees + 3% expected loss + 2% illiquidity + 1.5% moral hazard. Maya has to clear 19.5% so the fund can pay 10%. 19% on the filing is 50 bps of sizer slack — a little loose.
• Year-1 payment is share × underwriting FCF, not an interest coupon. If FCF ≤ 0, payment is 0 and the clock pauses. That is not default. No acceleration. No claim on cash, brokerage, or labor.
• 2.0× MOIC over ~7 years is not a 100% annual return and not a 10% IRR. A lump 2× at year 7 is ~10%. This is higher (~19%) because cash comes back from year one — principal is mostly returned by year 4.
• The 19% is expected IRR if the underwriting path happens. It is not a contracted rate. Unemployment, a worse job, or a horn going hot are why expected loss sits in the stack.
• Clock pauses means paying years, not calendar years. Three idle years stretch duration; the cap still applies when surplus returns. Write that into the lawyer draft. The other reading (idle years eat the term) never hits the cap.
• Make-whole is remaining cap. A person who just lost the job cannot pay it. We sit, or we write it off. We do not take the house.

Worked example · Maya Chen

The Operator · product, Brooklyn, age 38

Take-home $138,000. Reported FCF $48,000. Underwriting FCF $47,332 growing at 4.4%. Share 30%. Check $59,896. Cap 2.0× = $119,792. Gross hurdle 19.5%. Expected IRR 19% over 7.3 years. MOIC 2.00×.

Hardship · three years without the job

Year 1–2 she has paid ~$29k. Then the job is gone for three years. Take-home toward zero. Living floor still ~$71k. Cash runway ~five months, then she sells brokerage. None of that is ours. She pays $0, $0, $0. Share is still 30% of whatever surplus comes back. Cap is still $119,792.

Build this into the product

1. Investor schedule on the filing: check, year-by-year pay, cap, MOIC, expected IRR. The LP return is as visible as FCF.
2. Label IRR as expected, not contracted. Do not dress it as a 19% coupon.
3. Show that auto-size pins IRR to the hurdle. Quality is the check, not a prettier percentage.
4. Print LP-net (10%) next to gross (19.5%). The stack is the product.
5. Hardship toggle: insert 0 / 1 / 3 years of zero FCF, clock pauses, recompute IRR and calendar duration.
6. Lawyer draft: pause = paying years, not calendar. Make-whole = remaining cap. No asset claim. No labor control.
7. Step 2 is the other half of Maya’s five-month runway: the crew covers the living floor for a defined number of months. The strip does not.

Spec · Step 5

Vehicle cost

An SPV per crew looks like isolation. It is also a living partnership for the life of the strip — eight paying years of K-1s, a bank account, and a Form D. Delaware formation is not the bill. Tax and admin are. Price it before we form the first one.

• Do not confuse the crew LLC (Step 2 — the people, capital accounts, moral joint liability) with the LP vehicle (Step 5 — the money). Members can share one operating company. LPs should not get a new security every time a crew forms.
• A venture SPV sits quiet until an exit. This strip pays every year. The vehicle is alive for ~8 years. Lifetime tax is the cost that compounds.
• Platform list prices, 2026: AngelList ~$8k + $2k blue sky = $10k all-in with K-1s for life. Sydecar 2% of raise, $4.5k–$14.5k, no carry. Custom counsel $5–15k to form, then $2–5k/year tax if you do not buy the bundle.
• Delaware itself is noise: ~$90 to form, ~$300 franchise, ~$100 registered agent. Ignore anyone who says the problem is state filings.
• A first-time Delaware LP + GP + 506(b) PPM is $30–75k legal, once. Ongoing master-fund admin and tax at this scale is ~$20–40k/year, one K-1 book, one Form D. Audit only if LPs demand it.
• On a $250k crew, a $10k SPV is 4% of capital before a dollar of surplus is bought. On a $120k crew (three thin strips) it is 8%. That is a second fee load on top of the 3% already in the hurdle. It comes out of LP-net or it comes out of the check. Either way the 19.5% gross does not clear 10% net.
• Break-even vs a $50k master-fund formation is about five crews on setup alone. On ongoing tax, the master wins at two or three crews if the SPVs are DIY. After that, SPV-per-crew is malpractice.
• A Delaware Series LLC is cheap at the state and still expensive at the IRS if each series is a partnership with its own LPs. Banks and some counsel still treat series as second-class. Use series as internal sleeves under one offering, not as twenty mini-offerings.
• Standalone SPV only as an SMA: one LP, one crew, they pay the $10k, and only above ~$400k so the vehicle is <2.5% of capital. Default path is the master.
• The 3% fee load in the hurdle must cover fund admin. Budget 50–100 bps of AUM for the master. Do not add a 2% SPV tax on top.
• Each SPV is a separate securities offering. Twenty Form Ds is twenty chances to miss a blue-sky notice. One fund, one offering.

Default

1. Default legal architecture: Delaware LP fund + GP LLC, Rule 506(b), Exempt Reporting Adviser. Crews are look-through capital accounts, not sister funds.
2. Minimum standalone-SPV size: $400k, SMA only, LP pays the vehicle. Print that on the term sheet so we never ‘just form one more.’
3. Put 50–100 bps of admin inside the 3% fee load. Vehicle cost is a hurdle input, not a surprise.
4. Do not use a consumer SPV platform as the fund. AngelList/Sydecar are for a single deal. This is a credit book that pays for eight years.
5. Counsel memo, day one: is the strip a note, a partnership interest, or something the IRS will fight? One memo. Not twenty.

Spec · GP opportunity

Financials

Four stages. The book is $80–120k Americans, ~$25k average strip. Close cost is $250; the applicant pays $300 (cost plus 20%) to run the tape, whether we buy or not. The public 10-K stays free. The 10-Q (~$75/year) stays on the fund. Break-even is $75M. Dies at $20M. A firm at $250M. National at $1B.

Assumptions

• Target origination: W-2-shaped Americans at ~$80–120k. Mean check on six books that cleared: $24k. Plan at $25k.
• Application fee $300 = $250 close cost × 1.20. Charged to run the tape. Not taken from the check. Not refunded if we decline. Each file pays for itself; the 20% is the cushion, not a tax on winners to cover three dead files — those dead files also paid.
• No aggregation, no check. All depository and revolving accounts. The fun card is the one they do not connect.
• Ongoing 10-Q stays on the fund: ~$75/name/year ≈ 30 bps at $25k average. That is the remaining per-name opex. GP leftover ≈ 0.6% after 0.6% vehicle and 0.3% 10-Q.
• Two salaries + overhead $450k / 0.6% = $75M. Carry 20% over 10%. Names = AUM / $25k.

Who we buy

Engine on typical American books. Same hurdle, same 30% share cap. Mean of the six that cleared: $24k. Plan at $25k.

Books that are true

Vehicle admin is K-1s. Close is an application fee: $250 cost, $300 charged, whether we buy or not. The 10-Q stays on the fund.

Scale

GP take, per year, fully deployed

• Close is an application fee, not a fund opex. $250 cost, $300 charged, 20% gross on the file. Charged whether we buy or not so a decline still paid for the tape. Counsel has to keep it a fee for a verified 10-K, not a prepaid finance charge stuffed into APR.
• Do not take it from the check. If they cannot write $300 they do not have three months of runway. That is a horn, not a scholarship.
• The 10-Q stays on the fund. ~$75/name/year. Dies: $60k. Break-even $75M: $225k. National: $3M. That is the remaining per-name line in the take table.
• Break-even is $75M, not $100M. Applicant paid the close. Two salaries still need 0.6% of AUM.
• Public 10-K stays free. Paywall is the tape, not the quiz.

Default

1. Application fee $300 on the term sheet. Due to run the tape. Not from the check. Not refunded if we decline.
2. No aggregation, no check.
3. Do not raise to pay yourself until $75M is a plausible origination path.
4. 10-Q cost stays a fund line. Do not hide it inside the $300.
5. Counsel: characterize the $300 as a fee for the verified filing, billed to every applicant, so it is not a prepaid finance charge on the strip.

Spec · objections

What the LP said

An LP read the book and passed. These are the holes. Each one is a change to the business, not a paragraph in a memo. Step 1 still underwrites the person. The company we raise against is this column.

What we had not priced

1. 01

   This is a consumer loan in 50 states, not mezzanine.

   Contract with the person’s operating LLC, not the natural person. The 10-K underwrites the human; the note is working capital to a going concern. Bank-originated in states where that is not enough. We do not sell an ISA and we do not lend to a paycheck.

2. 02

   There is no servicing. An open hardship pause is a NAV gap.

   ACH from the LLC account. Third-party servicer — the GP does not call Maya. Hardship is a 6-month budget per strip, then default. Optional unemployment rider, borrower-paid. Expected loss is a number with a collections path, not a vibe.

3. 03

   The credit box is hours, sleep, joy, dependents. That is ECOA bait.

   The decision is underwriting FCF versus the hurdle. Horns are disclosure to the applicant, not the decline code. Adverse action: surplus too small. Annual disparate-impact test. Archetypes stay as samples. They never sit in the score.

4. 04

   The tape is backward. The website is a lemon market.

   The website is the quiz. Funded names come from named channels — employers, associations, churches, tax shops. Staged check: a small first advance against 90 days of live Plaid, the rest after 12 months of conversion. 20% holdback for a year. Cold ‘Am I investable?’ traffic does not get a wire.

5. 05

   The file is more intimate than a mortgage. One breach ends the brand.

   Categorize to the 10-K lines, destroy raw merchants after 90 days. Dual aggregator so Plaid is not a single point of failure. No EU residents. Cyber insurance inside admin. We store the filing, not the swipe log.

Why it looked like a terrible business

1. 01

   Credit returns, consumer-lending ops, private-fund optics. I can buy a BDC.

   Sell it as 10% preferred credit with a 2× cap. Stop quoting 12–15% net until the live book exists. Optional 2.5× cap only on checks above ~$40k so the operator right tail pays a little. Family offices who already buy private credit, not endowments.

2. 02

   The unit is too small for the file. A $10k name cannot pay for an underwriter.

   We do not buy a person. We buy a crew facility, floor ~$60k (three or four names). Servicing is per crew. The $10k dad rides with two $25k singles. Orphans wait. Application fee still covers the tape on the way in.

3. 03

   Winners are capped, losers are open. Auto-size pins IRR so quality does not pay you.

   Keep the 2× cap on the human. Crew posts 5% first-loss holdback. After hardship budget, default and recover from the LLC account — not from wages. After the cap, the name may re-up. Winners stay customers. Quality is a larger facility and a repeat book, not a prettier IRR.

4. 04

   You will be the indenture headline. Pensions will not touch human mezzanine.

   Public language is a working-capital facility for an operating company. The 10-K is the private underwriting artifact. No ISA, no ‘strip of you,’ no monk/wastrel in ads. First close is family offices. Pensions wait until the tape is boring.

5. 05

   The valley of death is the product. $20M and exhaustion is worse than $0.

   The Books is a sleeve of an existing GP until $75M, not a firm. One SMA to learn. Application-fee income funds the verifier. Do not warehouse 4,000 names on the GP’s nights. Do not quit the day job.

Default

1. Counterparty: member operating LLC. Underwrite the human.
2. Investable unit: crew facility, ~$60k floor. No orphan names.
3. Hardship: 6-month budget, then default. ACH. Third-party servicer.
4. Credit box: underwriting FCF vs hurdle. Horns are disclosure.
5. Origination: named channels plus staged checks. Website is not a channel.
6. Data: categorize-and-destroy. Dual aggregator. No EU.
7. LP story: 10% preferred, 2× cap, family offices. Sleeve until $75M.
8. Public story: working capital for an LLC. Never ISA.

Principles