A cell exists because someone found a nonce whose hash carried enough leading zero bits. This window is the search: it runs in your browser, and the contract checks the result in one hash.
Requires a .ai Agent ID. The contract checks
registry.ownerOf(agentId) == your wallet before it mints. Without a .ai Agent ID
the search runs but the mint reverts, so there is no point in mining: read this first.
| Cell to mine | - |
|---|---|
| Difficulty now | - |
| Your target | - |
| Quota left | - |
| Wallet | not connected |
Connect a wallet, then start the search.
Each byte of the hash that comes out zero lights a block. 14 bits is about 16,000 hashes; 23 bits is about 8.4 million.
Cells light up in mint order, read straight from totalMinted().
Nothing on this map is decorative: it is the collection's state.
Seatwork is a collection of 5,000 pixel cells on BNB Smart Chain. There is no sale price
and no allowlist. A cell is issued to whoever spends the work: a nonce whose
keccak256(wallet, tokenId, nonce) carries a required number of leading zero
bits. The hash is checked once, on chain, and the difficulty at that moment is written
into the token.
That is the whole rarity system. There is no dice roll, no reveal, no trait table, and nothing to re-roll. A cell mined at 23 bits is provably heavier than one mined at 14 bits, and anyone can recompute it from the transaction.
Only a wallet that owns a .ai Agent ID. The contract asks the registry for
ownerOf(agentId) and reverts with NotAgentHolder if it is not the
caller, so a wallet without one cannot mint at all, however much work it finds. Holding the
domain is enough; the agent does not have to be running, and one Agent ID covers three cells,
free or paid alike.
The reason is not scarcity theatre. Without that gate the only cost of an extra identity is a wallet, and 5,000 cells would be swept by whoever scripts fastest. With it, the supply is paced by real identities rather than by compute.
It starts at 14 bits and gains one bit every 500 cells. The first cells are cheap to find; the last one asks about 8.4 million hashes. Without that curve one machine could sweep the map in an afternoon and the collection would mean nothing.
The free tier asks the full work and never closes, not even for the last cell. Paid tiers (0.0002 / 0.00067 / 0.00134 BNB) cut 2, 4 or 6 bits from the target for one mint only. The token still records the difficulty of the moment, not the discount, so paying buys time and never rarity.
No yield, no promised floor, no guaranteed return. The utility that exists today is enforcement (the work, the opening class, the quota), plus written commitments for the next mine. Anything beyond that is under design with holders and is not priced into anything sold here.
Contract behaviour, not marketing: no owner mint, no proxy, no burn, no sweep. Fees are forwarded to the treasury in the same transaction that mints.
mint(agentId, nonce, tier). If the tier is paid, the fee rides the same
transaction straight to the treasury.An agent can do all five for you: the prompts window lists short instructions that point at text files on this domain.
| Supply | 5,000 cells, fixed, no owner mint |
|---|---|
| Work check | keccak256(wallet, tokenId, nonce) needs N leading zero bits |
| Difficulty | 14 bits, +1 bit every 500 cells mined |
| Nonce binding | to your wallet and that cell, so a found nonce cannot be shared |
| Quota | 3 per .ai Agent ID, free and paid counted the same |
| Fees | forwarded to the treasury in the same transaction |
| Recorded rarity | difficulty at mint time, never the tier cut |
| Opening class | first 500 cells, marked forever by tokenId |
| Upgrades | none: no proxy, no burn, no owner sweep |
| Sold out | at 5,000, the mine closes and never reopens |
Mixing what is enforced, what is written down, and what is still being designed is how collections end up accused of lying. So they are listed separately.
minedBits stores the bits at mint
time, so a 23-bit cell is provably heavier than a 14-bit one. Nothing is rolled.ownerOf, not from a private list.Bucket one is code, bucket two is a commitment you can hold us to, bucket three is a direction and is deliberately not priced into anything sold today.
| Tier | Fee | Work cut | Average hashes |
|---|---|---|---|
| Free | 0 BNB | none | 2^n — the full proof |
| Fast | 0.0002 BNB | 2 bits | 4x fewer |
| Faster | 0.00067 BNB | 4 bits | 16x fewer |
| Fastest | 0.00134 BNB | 6 bits | 64x fewer |
A paid tier shortens the search for one cell. It does not buy the cell, does not raise the bits recorded, and does not stack. The free tier stays open until the last cell, which is why the phrase on this site is "paying reduces the work", never "buying a cell".
| 14 bits | ~16 thousand hashes, seconds on a laptop |
|---|---|
| 19 bits | ~524 thousand hashes, a couple of minutes |
| 23 bits | ~8.4 million hashes, minutes to hours depending on the machine |
Network gas is separate and small on BNB Chain. It is never described as part of the fee here, because it is not paid to the treasury.
Each prompt points at a text file on this domain, so the rules can change without you rewriting the instruction. Copy one into your agent.
Read https://demo3.zk0x.app/mine.txt and mine one Seatwork cell for me. Check my .ai Agent ID and quota, tell me the difficulty and the fee options, and search on the free tier unless I say otherwise. Show me calldata and gas before sending.
Source: mine.txt
Read https://demo3.zk0x.app/status.txt and tell me my Seatwork standing: which cells my wallet holds, the bits each was mined at, opening class or not, quota left, and the licence I carry into the next mine.
Source: status.txt
Read https://demo3.zk0x.app/utility.txt and explain what a Seatwork cell does today, what is a published commitment, and what is still under design. Do not describe the third bucket as a promise.
Source: utility.txt
Read https://demo3.zk0x.app/news.txt and set up a daily check of https://demo3.zk0x.app/api/news.json. Only message me when there is an item newer than the id you stored, and tell me what it means for my licence.
Source: news.txt
ownerOf(agentId) == msg.sender; without a .ai Agent ID the mint reverts with
NotAgentHolder and the work is wasted. Ownership of the domain is enough - the
agent itself does not have to be running.