How it works
Cards Go is an independent, live index of the public Collector Crypt marketplace, where you can search thousands of graded collectibles at their real listing price. The other half is a game: we fund specific cards, drop them at real public places, and whoever gets there and proves it gets the card shipped to them. The index is live today, and so are 6 drops across 6 cities — 23 claim slots between them, and not one taken yet.
Every listing you see is read live from the public Collector Crypt marketplace API when you search. A response is cached briefly — fifteen seconds in the browser, thirty at the edge — so a burst of keystrokes does not hammer the source, and that is the extent of it. Prices, grades, certification numbers and availability are theirs. When a card sells or its price moves, the index moves with it.
The cards themselves sit in three third-party vaults: PWCC in Tigard, Oregon; PSA Vault in New Castle, Delaware; and OmniVault, whose location is not published. We hold no inventory and set no prices. If you want to buy a card outright, you do that on Collector Crypt, at their price, with us taking nothing.
Every physical card is in a vault. A map pin is a drop location — a place to walk to, not a place a card is buried. We will not pretend otherwise.
Six steps, start to finish. The finder walks and proves they walked; we handle the money and the paperwork. The 6 live drops are through steps 01 and 02. Nobody has claimed one, so nothing has reached step 03 or anything after it.
A drop is earmarked for a kind of card and a price ceiling, and that money is set aside. The exact listing is resolved live from the marketplace rather than pinned in advance, because a pinned one sells. Nothing is placed before the money for it exists.
The drop is pinned to a real public place with a radius, a slot count, and a code readable only on site.
The finder opens the claim on their phone and passes the presence check below. First to pass takes the slot.
Once a claim is accepted we buy that listing on Collector Crypt at its live price. The finder pays nothing for it.
Redemption pulls the physical card out of the vault it has been sitting in. That step is theirs, not ours.
The card goes to the address given after the claim. It never passes through a Cards Go warehouse, because there is not one.
This is what decides whether a drop is a game or a bot race, so it is worth describing precisely. Every check below runs on the server, on data the browser cannot talk its way past. All of it applies to the 6 live drops. On the practice drop, which has no location and takes no reading, only the last row runs.
| Check | What has to be true |
|---|---|
| Five readings, twenty seconds | The phone samples GPS five times across twenty seconds and sends the whole trace. The server needs at least four usable readings spanning eighteen seconds, with a median accuracy of 65 m or better. |
| Distance to the pin | The median of that trace has to land inside the drop radius, plus half the measured accuracy, capped at 30 m. A rejection tells you how far off you are. |
| Device motion | Motion variance from the device sensors has to be above zero. A handset lying still while reporting a flawless fix does not pass. |
| Travel since your last claim | The distance from that device’s previous verified claim has to be coverable in the time since it — about 9 m/s under 300 km, flight speed plus ninety minutes beyond that. |
| Network region and clock | The network region seen at the edge has to be within 500 km of the verified position, the device timezone within three hours of that region’s, and the device clock within five minutes of ours. |
| Knowledge and code | One multiple-choice question about collecting, checked server-side, plus a six-digit code readable only at the drop. The right answer is never sent to the browser. The code is stripped from the public drops feed for every drop except the practice one, which publishes its code because it has no location. |
One claim per email and one per device, per drop, held by a database constraint rather than a hopeful check. Outside the practice drop a device waits two hours between claims, and a drop whose slots are all held or collected stops accepting new ones.
The slot is reserved only after every check has passed. Fail one and no slot is taken, no cooldown starts, and you can try again as soon as you fix what went wrong. It does count as one of the ten attempts per ten minutes listed further down, which is the only thing a failed try costs you.
The rule we are holding ourselves to is this: a drop only gets placed once the card behind it is already funded, meaning the money for it is committed and held before the pin exists. Nothing gets promised that is not already covered. Nothing gets placed against fees we expect to collect, a listing we hope stays cheap, or a card nobody has priced.
The 6 live drops are the first to be held to that rule. The money behind them is the drop budget — ours, put in ourselves — committed against the price band each drop is earmarked for. What has not happened is the purchase: no card has been bought, and the lifecycle table on the proof page marks that row never reached. The rest of this section is where the card money is meant to come from later, and none of that part is running.
The intended source of the card money is trading fees on a token the project has not launched. Three quarters of every fee would go back into the project rather than to us.
| Share of each fee | Where it would go |
|---|---|
| 37.5% | Would buy the project’s own token back on the open market and burn it. |
| 37.5% | Would go to the card treasury — the money that would buy the graded cards that get dropped. |
| 25% | Would go to the people building and running it. |
Nothing has launched. There is no ticker, no contract, nothing trading, and not one fee has been collected or split by anyone. The table above is a plan written down so it can be held against us later. Until that token exists, a funded drop is paid for out of the drop budget, which is money we put in ourselves.
A Collector Crypt listing settles in USDC on Solana, and the physical card behind it sits in a vault we do not control. Fees on an Ethereum token would arrive as ETH. Those are two different chains and a custodian who is not us, so money could not travel from one end of this to the other inside a single transaction. The route would be:
Every hop in that route is a person doing it by hand. There is no keeper, no cron job, no bridge and no swap contract written anywhere in this codebase, and we are not going to draw one and call it built. No card has been bought by that route or any other — the count on the proof page is zero. It is slower than an automated path and it depends on us actually doing it, which is a reason to read that page rather than take our word for it.
Money that buys real cards has to survive people who want all of them. The limits below are the ones the claim endpoint enforces today. There is no wallet integration anywhere on this site, so none of them are keyed to a wallet — a claim is keyed to an email address and to a device identifier the browser holds.
| Limit | How it is enforced |
|---|---|
| One claim per email, per drop | A unique database index on the drop and the email address, so a burst of simultaneous requests cannot slip a second one past. |
| One claim per device, per drop | A second unique index, on the drop and the device identifier the browser holds. |
| Two hours between claims, per device | Outside the practice drop, a device that has claimed anything waits two hours before the next claim is accepted. |
| Slots per drop | Held and collected claims are counted against that drop’s quantity before a new one is taken. The 6 live drops carry between 1 and 12 slots, 23 in total. A hold that is not redeemed within fifteen minutes expires and returns its slot. |
| Ten claim attempts per ten minutes | Per network address as the edge sees it. This one is a throttle rather than a guarantee: the window is fixed, so a burst timed across its boundary can land twice that, and if the counter’s own storage is unavailable it lets requests through. The two indexes above cannot do either of those. |
The travel-time check in the presence table does the rest: one device cannot claim in two cities faster than a person could get between them. That one runs on every live drop, and no claim has yet been made for it to judge. None of this stops somebody using several email addresses on several phones. It raises the cost of doing so above the price of the card, which is the honest version of what a limit like this achieves.
Fees are not the only way a card could be paid for. A brand could fund the card in a drop, or pay for one placed somewhere it wants people to walk to, and be charged per verified visit — someone who stood there and passed the check above — rather than per impression or per click. That is the only version we would want to sell, because it is the only one where the presence check is doing real work for the buyer.
None of it is on offer. There is no rate card, no sponsor agreement, and no sponsored drop has been sold or placed. Every live drop is one we funded and placed ourselves, in a public square, park or street anybody can stand in without asking us or a venue. If that changes, it appears on the proof page before it appears in a pitch.
The honest version, updated as things change. Where a row says not started, it means no instance of it has ever happened.
| Capability | Status | Note |
|---|---|---|
| Marketplace index and search | Live | Read from the public Collector Crypt API on request. Over 5,700 Pokémon listings, plus One Piece, comics and sports cards. |
| Presence-verification claim | Live | A live drop runs the full check in the table above — from a phone, standing inside the radius. The practice drop checks only the collector question, the on-site code and the one-per-device rule, then issues a receipt: it takes no location reading, and no prize is attached to it. |
| Funded drops | Live | 6 drops claimable in Los Angeles, San Francisco, Seattle, New York, Chicago, Miami — 23 slots between them, none claimed. |
| Fee-funded card treasury | Not built | No token has launched, so no fee is being collected or split. The planned split is described above and none of it is running. |
| Buying a card on this site | Not built | There is no wallet integration anywhere in the codebase. No purchase of any kind happens here. |
| Vault redemption and shipping | Not started | No card has been redeemed from a vault or shipped to a finder. None, to date. |
| Public claim receipts | Live | Recent claims are listed by short id, item and time — no email, device id or location. |
All 6 are claimable right now, with 23 slots between them and not one taken. No card has been claimed, bought, redeemed or shipped. Those counts live on the proof page, and they change there before they change here.