A member photographs something they no longer use. Our AI identifies it, checks what that exact model has recently sold for used, grades its condition from the photos and discounts for how quickly things like it actually move. The result is a Share Value, and credits land in the member’s wallet immediately — before any buyer exists. The item stays in their home, committed, until a neighbor claims it.

Those credits buy other members’ things. They cannot be cashed out, refunded or sent to another member. The only thing a Share Credit does is claim goods. We make money on the 5% service fee, charged in dollars at the point of claim, plus sales tax which we collect and remit.

$3–5T

the global stock of idle household goods — our arithmetic, shown in §06

UN population and household data, applied to a US per-household figure

$4,267

of unused goods in the average US household

Mercari Reuse Report (GlobalData), 2023

8.2 billion

people, of whom 57% live in cities — the density this needs

UN World Population Prospects, 2024

9.13M

households in the six California metros we start with

US Census, ACS 1-year, 2024

The thesis. Barter has always failed on the coincidence of wants: Alice needs Bob to want her chair. An AI appraisal priced in a common unit removes that requirement entirely. Alice’s chair becomes buying power, and she spends it on Carol’s bike. What was impossible to coordinate becomes a marketplace — and unlike a marketplace, the seller does not have to wait for the sale.

The market is not American. The United States holds about 4% of the world’s people, and the constraint on this business is density in a neighbourhood rather than a border. Airbnb did not size itself as US travel; the largest players in secondhand today are European and Chinese. We start in one California metro because liquidity is local, not because the opportunity is.

We are raising $2.5M to prove one metro clears, then open five more. The plan below is what we believe, where the numbers came from, and the four things that would tell us we are wrong.

The value is real and it is idle

$4,267 and 161 items per household, $559.8B nationally Mercari Reuse Report, 2023. 12.6% of US households now pay monthly to store what will not fit Self Storage Association demand study, via CRE Daily, 2025 — a market whose existence is evidence for ours: people would rather pay rent on their possessions than deal with selling them.

Selling is a part-time job

Photograph, research a price, write a listing, answer “is this still available?” forty times, get lowballed, arrange a meeting, get stood up. 54% of Americans sold something secondhand last year OfferUp, which means nearly half did not, while holding thousands of dollars of sellable goods.

And it is not safe

41% of US adults have encountered a scam on a social marketplace and 13% were victimised Norton / Gen Digital (Dynata, n=1,000), 2025. Pew puts the share who have actually had an online shopping scam happen to them at 36% Pew Research Center (n=9,397).

The lifecycle

Yours → appraised → in the ShareHouse → reserved → handed off → theirs. A listed item is a commitment: the member keeps it safe and available in the stated condition until it is claimed. Before a claim they can withdraw it by returning the credits or substituting items of equal Share Value. After a claim, it is the claimant’s. Pickup is local, inside 72 hours, with a QR scan and both sides confirming condition.

Or rent it

An item you still want but rarely use is a rental: it stays yours and earns about 7% of its Share Value a week. Most listings offer both, and the first neighbor to act decides which it was.

Listia issued credits for listings and daily activity, and also sold them for cash — a faucet with no goods behind it. Bunz minted BTZ for advertising views, then cut merchant acceptance overnight; the currency was not insolvent, the company had run out of dollars BetaKit, 2019. Yerdle reached 950,000 members and pivoted away from consumer credits entirely Trellis (GreenBiz).

The identity

In steady state, credits minted must equal credits destroyed. Writing a for the share of an appraisal paid at listing, p for the share of listings eventually claimed, B for credits granted and A for appraised value:

The best comparable operator is ThredUp, whose goods are ordinary household items and whose funnel is fully disclosed: 43% / 69% of listings sell within thirty and ninety days ThredUp Form S-1, and 38% of accepted items never sell inside their window at all ThredUp seller terms. That is achieved having physically received, inspected, photographed and priced every item, and rejected a tenth to a third of what arrived.

This changed the product. Our earlier design paid trusted members 100% of an appraisal up front, which the identity says is solvent only if essentially everything clears. The front-load is now capped at 50% for every tier, and tier buys a smaller haircut rather than a bigger advance — the reward lever moves, the solvency lever does not.

Why the haircut lands where it does

Three unrelated industries price exactly this risk and converge on the same band. Pawnbrokers lend about ~60% of value against consumer goods; asset-based lenders advance 80% of net orderly liquidation value FirstCash Form 10-K; OCC; and the Federal Reserve discounts idiosyncratic, slow-to-liquidate collateral by 30–70%. The European Central Bank goes further and charges an additional haircut of up to 13 points purely because a valuation is model-derived rather than market-observed — which every ShareHouse appraisal is.

So liquidity is not a decoration on the appraisal; it is the appraisal. A fast-moving item keeps about 80% of its comp after the liquidity term, a slow one about half. Those are cold-start priors, to be re-estimated weekly per category and metro from our own claim history — moving a few points at a time, published before they take effect, and never applied retroactively.

Credits have to be destroyed, not only circulated

The failure that actually killed these systems was not inflation. It was accumulation: the few participants who accept the currency end up holding all of it, find nothing they want, and start refusing it. One food co-op came to hold 33% of the entire Ithaca HOURS money supply and cut its acceptance rate to 12.5% Krohn & Snyder, International Journal of Community Currency Research, 2008. The currency never recovered.

A circulation loop with no sink is a pure faucet. So a 4% fee on every claim is charged in credits and destroyed — separate from the 5% we charge in dollars. It is one number, adjustable monthly, and it is the primary monetary instrument. It is shown to the claimant as its own line in credits before they confirm — a mandatory fee that is not displayed is exactly what California’s all-in pricing rule forbids, and a sink nobody can see is not a policy instrument, it is a shrinkage. In EVE Online, the most professionally managed virtual economy with public accounts, transaction and broker fees are 37% of everything destroyed CCP Games Monthly Economic Report.

What we publish

Coverage — claimable inventory over credits outstanding — with a target of 125%, a watch level at 115% and a halt at 105%, at which new listings mint on claim only. Alongside it: the clearance ratio (credits destroyed over credits minted), realised clearance by cohort, credits per active member, and a price index. The circuit breakers are pre-committed and published, because a rule announced in advance is monetary policy and the same rule announced during a crisis is a default.

Three things we now know will be attacked

The mint. On marketplaces that paid per-trade rewards, wash trading reached 94.5% of volume, against low single digits where no reward was paid Niu, Li, Peng & Li, arXiv:2403.10361. We mint on listing, so the mint is the reward. The attack is invisible transaction by transaction — the equivalent fraud on Amazon runs on genuine purchases with off-platform rebates — and the only defence with real ground truth behind it is the counterparty graph, which beat text, metadata and image features He, Hollenbeck, Overgoor, Proserpio & Tosyali, PNAS. That instrumentation has to exist before the credits do, because it cannot be reconstructed afterwards.

The appraisal. Zillow wrote down $407.9M for buying homes above what they would later sell for Zillow Group Form 10-K. CarMax, with a physical inspection rather than a photograph, finds that about half of what it buys meets its retail standard CarMax Form 10-K.

The promise. eBay tolerates a 2% transaction-defect rate, and its own definition includes the seller having sold the item to somebody else eBay seller standards policy. Ours is a claim against goods verified by a photograph and held in the counterparty’s home. We should expect a comparable rate and price it.

1. AI can now run the hard part of a credit economy

A credit economy needs three jobs done continuously and cheaply: identify what a thing is, decide what it is worth here, and keep issuing and demand in balance. Until recently each needed a human expert. A vision model now identifies an exact model number from four photographs, grades condition, and does it for roughly five cents Anthropic, 2026. The same class of model writes the listing, generates the marketing imagery, answers the member’s questions and runs the trust queue.

That is the difference between this and every previous attempt. Listia and Yerdle priced goods by asking members what they were worth, which is exactly the input a credit economy must not trust — it is the mechanism by which every one of them inflated. Pricing is the monetary policy of this network, and it is now possible to run it consistently, per item, per neighborhood, and to re-estimate it nightly from what actually cleared.

2. Everyone carries a good camera

The appraisal is only as good as its evidence. A modern phone camera resolves a model label, a serial plate, the wear on a chuck and the scuff on a chair leg — the details a grader would look at. A decade ago the same photograph would not have supported a binding valuation. The supply side of this business is now a two-minute task that anybody can do standing in their garage.

3. Secondhand stopped being a compromise

93% of Americans bought something secondhand last year and the stigma is gone OfferUp. 62% say they prefer transacting locally OfferUp. We are not asking anyone to change their behaviour, only to be paid for it sooner.

4. The credit layer has somewhere to go

We are deliberately launching on a boring, centralised ledger: Postgres, a balance that is the sum of its entries, and no token. That is the right call for year one, because the failure modes of this business are fraud and inflation, and both are easier to fix when we can reverse an entry.

But the architecture is deliberately compatible with more. A Share Credit is already a claim on a basket of real, appraised, committed goods — which is a more honest backing than most digital assets have. Once the network is large enough that the liquidity model is trustworthy and fraud is understood, three things become possible in order: a verifiable public record of issuance and coverage, so members can audit our solvency rather than take our word for it; settlement that does not depend on us being solvent; and, if and only if the market is robust enough to absorb the fraud surface that comes with it, credits that are transferable and eventually tradable.

The stock, not the flow

Most marketplace plans size the flow: what changes hands each year. That understates this one, because ShareHouse is addressed at the stock — the goods already sitting in homes, doing nothing. In the United States that stock is $559.8B across 21.1 billion items Mercari Reuse Report, 2023, or $4,267 in the average household Mercari Reuse Report (GlobalData).

The United States holds about 4% of the world’s people — 8.2 billion live on it UN World Population Prospects, 2024 — and 57% of them live in cities UN World Urbanization Prospects, which is the density this model needs. Sizing ourselves at the US market would be sizing ourselves at a twenty-fifth of the problem.

How we get to trillions, with the arithmetic shown

There is no published figure for the global value of unused household goods, so this is built rather than cited, and every input is marked.

Why this exports, when most marketplaces do not

A resale marketplace normally needs local pricing expertise, local category knowledge and local staff to enter a country. Ours needs a model that already reads a model number in any language, and a liquidity estimate that is per-neighbourhood anyway — San Jose and Seoul are equally unfamiliar to us on day one, and equally learnable from the same signal. There is no warehouse to build and no shipping network to license, because the goods never move more than a few miles.

The evidence that the demand is not American is that the largest players in this category are not American. Vinted turns over €10.8B profitably in Europe Vinted newsroom, 2026; global secondhand apparel alone is forecast at $393B by 2030 ThredUp Resale Report (GlobalData), and secondhand furniture at $47.2B Mordor Intelligence.

Why we still start in one California metro

A global TAM is not a global launch. Liquidity is per-neighbourhood: a member in Rose Garden cares about the two hundred listings within a few miles and nothing else, so the unit of expansion is a neighbourhood however large the market is. California first means one set of sales-tax rules to get right across 13.8M households Census ACS, 2024, of which our six launch metros hold 66%.

We start in San Jose: 693,483 households, 12.7% of whom moved last year Census ACS, 2024. Movers are the wedge everywhere — 39.6M Americans move house each year US Census, ACS 1-year — because they are the only customer who needs both sides of this network in the same month.

Demand is not the question: 93% of Americans bought secondhand last year and 62% say they prefer transacting locally OfferUp.

Dead capital, in the household

Hernando de Soto’s argument was that the poor are not assetless — they own a great deal — but their assets are dead capital: economically inert, because no system gives them a formal, fungible representation that can be exchanged or borrowed against The Mystery of Capital, Basic Books, 2000. His subject was land title in the developing world. The same structure holds, at smaller scale and at every income level, for the objects inside a home.

A drill is worth something. It is not spendable. Between owning it and using its value stands a week of photographing, pricing, listing, answering and meeting strangers — friction so reliably larger than the reward that $4,267 per household Mercari Reuse Report (GlobalData) simply sits there. What ShareHouse does is not primarily a marketplace function. It is a conversion: an illiquid asset into a liquid claim, in about two minutes.

Who this actually reaches

A large minority of American adults report they could not cover a $400 emergency expense with cash or its equivalent Federal Reserve, Survey of Household Economics and Decisionmaking, 2024 — many of them while owning a garage worth several thousand dollars. The gap between those two facts is the product. Liquidity, not wealth, is the binding constraint on most households, and household goods are the one asset almost everyone has.

This is also why the credits must stay closed-loop. The moment somebody can sell their credits for cash, this becomes a lending product aimed at people under financial stress, with everything that follows from it — the pricing pressure, the collections problem, the regulatory surface and the reputational one. Non-transferability is not a limitation we are apologising for. It is what keeps this an exchange rather than a payday advance.

The flow, if the stock moves

§06 puts the global stock of idle household goods at $3–5 trillion by our own arithmetic. A stock only matters if some of it turns. Applying the turnover rates this category actually achieves:

Two effects compound and neither is captured by that table. Activated value is spent, not banked: a credit that can only buy goods has a velocity close to one, so each dollar activated moves through the network rather than sitting in it. And every claim is a purchase that did not need new manufacturing — the environmental case and the economic case point the same direction here, which is rare.

And the honest caveat

De Soto’s thesis has a well-documented empirical problem: land-titling programmes repeatedly failed to produce the credit access they promised, because formal representation turns out to be necessary but nowhere near sufficient — somebody still has to want the asset, and a market still has to exist to clear it. That criticism applies to us with full force. Making a drill liquid does nothing at all unless a neighbour wants a drill. Every number in this section is therefore downstream of the unglamorous work in §04 and §09: clearance rates, liquidity discounts, and density in a single neighbourhood. The macro case is the prize. It is not the plan, and we would be suspicious of anyone who presented it as one.

Why the credit loop is the moat

A credit can only be spent here. Every member who lists becomes a buyer by construction — the closest public analogue is Poshmark, where 48% of sellers spent their earnings back on the platform Poshmark S-1, and that was with cash they could have taken home. Liquidity compounds locally, and a competitor entering a metro faces our density rather than our features.

The honest starting point is the one from the previous section: our appraisal is not proprietary. Meta, eBay and Mercari all ship AI listing and price suggestion, free, to audiences we will not reach for years Meta Newsroom, 2026. Any competent team can build what we have built. Several will.

So the question is not what we own today. It is what gets harder to copy every month we operate. Four things do, and they are the reason this business gets more defensible rather than less as AI commoditises the software layer.

1. The clearing data — the only training set that matters

A general model can tell you what a Herman Miller Aeron sells for. It cannot tell you what one clears for in Willow Glen, in March, in eleven days, at a 0.92 liquidity factor — because that is not on the public internet. It exists only as a byproduct of running this network.

Every appraisal we issue is a labelled example: the item, its condition, its price, and then the outcome — claimed or not, how fast, by whom, at what distance. Within a year of one metro we will hold tens of thousands of these. That dataset is the liquidity term, which is the term that makes a credit economy solvent rather than inflationary. A competitor launching in year three starts with a market-comp API and a guess, exactly where we started — while we are pricing off observed behaviour. The appraisal is not the moat; the correction to the appraisal is.

2. Reputation, which does not travel

A member arrives at their fifth handoff carrying a record: {handoffs completed, as-described rate, reply time, reviews from the neighbours who actually met them}. That record is what lets us advance them 50% of an appraisal instead of 25% — it is not a badge, it is their credit limit.

Reputation is the classic answer to the lemons problem, and it is the one asset a member cannot take with them. Starting again elsewhere means starting at 50% advances and no standing. The same is true in the other direction: item-level reviews accumulate against categories and contributors, so our estimate of what a particular person’s listings are worth improves in a way a newcomer’s cannot.

3. Density, which is per-neighbourhood and must be beaten street by street

A national marketplace is one market. This is not: a member in Rose Garden cares about the two hundred listings within a few miles and nothing else. A competitor cannot enter “the US” — they must beat us in San Jose, then in Willow Glen, then again in Oakland, each time paying the full cost of seeding supply before a single claim is possible.

This is why we launch one metro at a time rather than nationally, and it is the same lesson every survivor learned. thredUP narrowed to a single category to escape the chicken-and-egg problem see §06; Fat Llama’s value concentrated in a small number of deep local lenders EU-Startups. Density is slow to build and equally slow to take.

4. The balance itself

A member holding S$400 has a reason to return that no competitor can match, because the credits are spendable in exactly one place. Poshmark is the closest public evidence that this effect is real even when the money is not locked in: 48% of its sellers spent their earnings back on the platform despite being free to take the cash home Poshmark S-1.

We should be clear-eyed that this is lock-in, and lock-in is only an asset while the thing it locks you into is worth having. A closed loop over a thin marketplace is a complaint, not a moat. It counts as defensibility only in combination with the three above.

Economies of scale, in the literal sense

The loop, stated plainly

More listings make the liquidity estimate better. A better liquidity estimate makes Share Value more accurate, which means fewer credits issued against things that will not move — so the coverage ratio holds, credits stay trustworthy, and members keep spending them. Spending produces handoffs; handoffs produce reputation and reviews; reputation raises advance rates; higher advances make listing more attractive, which produces more listings. Each turn of that loop leaves behind data a competitor cannot buy and trust they cannot transfer.

What would breach it

Honest limits. If Meta decided to issue value against uninspected goods, they would start with distribution we cannot match — though they would also inherit the liability we have spent this plan designing around, which is why they have not. If our liquidity model were wrong for long enough to inflate the currency, no amount of data would save it, because the failure destroys the trust that makes the data worth having. And if a metro never reaches density, none of the four assets begin to accumulate there at all. The moat is real, but it is downstream of getting the economics right first — which is the subject of §04 and the reason it sits earlier in this document than this section does.

The density threshold

A neighborhood is useful at roughly 200 listings. Below that it is a list; above it, a warehouse. We seed each with concierge appraisals — we go to people’s homes and photograph their garages ourselves — then open claims when the map is full enough to be worth browsing.

Three channels, in order

Movers, targeted at address-change and lease-end signals, where the pitch is “photograph thirty things, furnish the new place.” Referral, which works unusually well here because a credit is only spendable if there is something to spend it on, so members recruit their own supply. Neighborhood-scale organic: buy-nothing groups, apartment complexes, university move-out weeks.

Unit economics of one claim

Contribution margin on the fee is 72%, and it is dominated by the fixed part of card processing. That is the reason for a fee floor, and the reason a $20 item is a worse transaction than a $200 one — Poshmark solved the same problem with a flat $2.95 below $15 Poshmark.

Lifetime value

Later revenue, not modelled

A membership that raises the advance rate and adds protection; promoted listings; delivery for furniture; business liquidation supply. None of these are in the forecast; all of them raise the effective take rate without touching the 5% promise.

Revenue reaches $1.9M on $35M of goods exchanged in year five. EBITDA stays negative across the plan: this is a density business, and the marketing and trust costs of opening metro twenty are incurred a year before its members transact. The path to profitability is stopping — at any point we can hold the footprint and the contribution margin covers the fixed base within about four quarters.

We are probably a barter exchange, and that is survivable

Treasury defines one as an organization whose members contract to trade property, where payment is made “by means of a credit on the books” 26 CFR §1.6045-1. That describes us. The consequence is Form 1099-B reporting for members, with credits reportable when issued rather than when spent. We plan to collect tax IDs at listing and issue forms from year one. There is a de minimis floor of 100 exchanges a year §1.6045-1(e)(2)(ii), which covers a pilot and nothing more.

For most members this is paperwork rather than tax: selling personal property for less than it cost produces no taxable gain IRS Publication 525, which is the usual case for household goods. Communicating that clearly is a product requirement.

Sales tax, collected by us

California treats barter as a sale, measured at fair market value R&TC §6006, and the Marketplace Facilitator Act makes the platform the retailer responsible for collecting it CDTFA. We collect and remit at the point of claim. Neither neighbor has to think about it, which is a feature.

What we have designed around

Credits are never sold for money, never transferable and never redeemable, which is the cleanest position under California’s stored-value rules Fin. Code §2003(aa). They never expire, which keeps us out of escheat. The open question is the gift-certificate rule requiring cash redemption of balances under $15 Civ. Code §1749.5 — it applies to certificates that are sold, and ours are earned, but members do give something of value. It is on the list for written counsel.

All-in pricing, which changed the product

California’s Honest Pricing Law makes it unlawful to display a price that omits a mandatory fee; only government taxes and actual shipping may be added afterwards Civ. Code §1770(a)(29), operative 2024. The Attorney General’s guidance is explicit that disclosing the fee later in the flow does not cure it — the displayed price must be the full price California Attorney General.

Our 5% service fee is mandatory and is not a government charge, so it travels with the price everywhere a price appears rather than arriving at checkout. Sales tax is excluded by the statute and is still estimated at checkout, where the jurisdiction is known. This is a product rule, not a terms-of-service clause: no disclaimer cures it, and the exposure is one-way attorney’s fees with no summary judgment available against a class Civ. Code §§1780(e), 1781(c).

One open question for counsel: there is no authority on prices denominated in credits rather than dollars. We have taken the conservative reading — that a credit figure is a displayed price and the fee must accompany it — rather than the convenient one.

Product safety

The CPSC makes it unlawful to resell recalled goods and names the categories to refuse US Consumer Product Safety Commission. Recall screening runs inside the appraisal, and cribs, car seats, helmets, infant sleep products and mattresses are blocked at intake.

Use of funds

What this buys, in milestones

Month 6: San Jose at 2,000 listings and a measured sell-through by category — the first real liquidity discount, replacing our estimate. Month 12: $2.6M of GMV, coverage ratio held above 110%, three metros live. Month 18: the Series A story — 26K active members, repeat rate above 40%, and a credit economy that has run a full year without inflating.