On 1 February 2023 at 20:38 UTC, the blockchain and earns a reward, which makes rewriting history very expensive.">mining smart contract holding reserves of two or more tokens, supplied by users, that a decentralized exchange uses to execute trades automatically at a price set by a formula.">pool Luxor found Bitcoin block 774,628, and it showed what Bitcoin Ordinals could do to a block. It held only 63 transactions and weighed in at 3,955,272 bytes, a weight of 3,992,769 units against a maximum of 4 million. One transaction of 3,938,383 bytes, nearly all of it witness data, took up almost all of that room to carry a single JPEG picture of a wizard. mempool.space shows that this transaction paid no fee on chain at all. Fourteen months later, on 20 April 2024 at 00:09 UTC, block 840,000 cut the mining subsidy in half and carried 3,050 transactions that paid 37.63 BTC in fees, about twelve times the new 3.125 BTC subsidy, much of it from people rushing to create the first Runes tokens.
Both blocks come from the same idea. Ordinal theory, a numbering scheme published by the developer Casey Rodarmor, gave every satoshi an identity, and the community built collectible inscriptions, a JSON token standard called BRC-20 and a leaner token protocol called Runes on top of it. None of these needed a change to Bitcoin’s rules. Bitcoin stores the data; software outside Bitcoin decides what it means. That split explains most of what followed: the fee spikes, the quarrel over block space, the marketplaces that closed in 2026 and the inscriptions people still lose by mistake.
It follows our step-by-step guide to running a Bitcoin node. That node already checks every inscription and every Runes transaction, because to Bitcoin Core they are ordinary transactions. It will never show you a picture or a token balance, though. For that you need an extra program, an indexer, and almost every risk described further down comes back to that dependence.
Inscriptions, BRC-20 and Runes: three things built on one rule
People lump these projects together as “NFTs and tokens on Bitcoin”, but they sit at different levels. Ordinal theory is a convention for naming individual satoshis. Inscriptions attach a file, any file with a declared content type, to one of those satoshis. BRC-20 is a set of inscriptions that contain small pieces of JSON text, which indexers add up into token balances. Runes is a separate protocol that writes token instructions in a compact message inside an output of the transaction, with no inscription and no numbered satoshi involved.
What they share is a gap in Bitcoin’s design that is in fact a feature. Full nodes check that coins exist, that signatures are valid and that nobody spends the same output twice. They do not interpret the bytes a transaction carries beyond what the script needs. A picture, a JSON string or a token message is just data to them. Whether a particular satoshi is “rare”, whether an address owns 1,000 ORDI or whether a Runes mint is still open are questions answered by indexers, programs that replay the whole chain with an extra set of rules.
How Bitcoin Ordinals give every satoshi a number
A bitcoin is divisible into 100 million satoshis, or sats, and Bitcoin itself never tells them apart. Ordinal theory does. In the words of the ord project’s handbook, “Satoshis are numbered in the order in which they’re mined”. The first sat of the genesis block is sat 0, and each block’s subsidy adds a new range of numbers. When a transaction spends several inputs into several outputs, the numbers flow through it first-in, first-out: the first sats of the first input fill the first output, and so on, with whatever is left over going to the miner as part of the fee.
The handbook insists that ordinal theory needs no separate token, no other blockchain and no change to Bitcoin. That is accurate, and it is also the source of every later difficulty. The numbers exist only in software that applies the first-in, first-out rule from the genesis block onwards. A Bitcoin transaction contains no field that says “this is sat 1,234,567,890”. If you send a transaction with a wallet that knows nothing about the rule, you can move a numbered sat without noticing. The same sats also show up under several notations (integer, decimal, degree, percentile and a name made of letters), which is why explorers can display what looks like four different identifiers for one coin.
The scheme fits neatly onto Bitcoin’s model of coins. If unspent outputs, change and coin selection are hazy, our lesson on the UTXO model covers them; the short version is that you never spend part of an output, you always consume a whole one and create new ones, and the ordinal rule decides which numbered sats end up in which new output.
Rare sats and why collectors hunt them
Ordinal theory also defines a rarity scale tied to Bitcoin’s schedule. A common sat is any sat that is not the first of its block. An uncommon sat is the first sat of a block, a rare sat the first of a difficulty adjustment period (every 2,016 blocks), an epic sat the first of a halving epoch (every 210,000 blocks) and a legendary sat the first of a cycle, the point where a halving and a difficulty adjustment fall on the same block. The only mythic sat is the first sat of the genesis block.
A little arithmetic shows how scarce the top grades are. A cycle lasts six halving epochs, or 1,260,000 blocks, which is also exactly 625 difficulty periods. The first cycle ends at block 1,260,000, so as of 9 November 2026, with the chain at block 975,158 according to mempool.space, no legendary sat has been mined yet. Only four epic sats exist so far, the first sats of blocks 210,000, 420,000, 630,000 and 840,000. “Sat hunting” means scanning your own coins for such sats with the ord software; its guide says this needs a synced node with a transaction index and an index of every sat, a far heavier set-up than a normal node. Whether a rare sat is worth anything is a matter of taste among collectors, nothing in Bitcoin’s protocol.
Inscriptions: a file inside a Taproot witness
An inscription is a sat plus content. The inscriptions chapter of the handbook puts it plainly: “Inscription content is entirely on-chain, stored in taproot script-path spend scripts.” There is no link to an outside server as with many Ethereum NFTs: the bytes of the image, text or program are in the Bitcoin block itself, in the part of the transaction that carries signatures and scripts, the witness.
Commit, reveal and the envelope
Creating an inscription takes two transactions. The first, the commit, sends coins to a Taproot output whose hidden script contains the content. Nothing is visible yet: a Taproot output looks like any other. The second, the reveal, spends that output through the script path, which forces the full script, and so the content, to appear in the witness. That is why every inscription costs two fees.
Inside the script, the content sits in what the handbook calls an envelope: an OP_FALSE OP_IF … OP_ENDIF block. Because OP_FALSE puts a false value on the stack, the code between OP_IF and OP_ENDIF never runs, so Bitcoin skips it while still storing it. The envelope starts with the three letters ord, then a tag announcing the content type (image/png, text/plain and so on), then the body. A single data push in Bitcoin script cannot exceed 520 bytes, so larger files are cut into a series of 520-byte pushes. Unless a pointer field says otherwise, the inscription is attached to the first sat of the reveal transaction’s input. Our lesson on Bitcoin transactions and Script explains the stack and opcodes this trick relies on.
You can check this on any explorer. The transaction that holds inscription 0 (id 6fb976ab…42799) confirmed in block 767,430 on 14 December 2022 at 20:32 UTC. Read its witness on mempool.space and you find the letters ord, the content type image/png and a 100 by 100 pixel PNG file. Version 0.2.0 of the ord software was published on GitHub earlier the same day.
Why a 4 MB block became possible
Inscriptions fit into blocks because of two older upgrades. SegWit, in 2017, replaced the 1 MB size limit with a limit of 4 million weight units, and witness bytes count for one weight unit each while other bytes count for four, so witness data is four times cheaper. Taproot, in November 2021, made script-path spends with large scripts practical. Our lesson on SegWit and Taproot covers both in detail. Nobody designed them for pictures, but together they made a block of almost 4 MB of witness data valid.
Block 774,628 is the extreme case. The block page on mempool.space lists 63 transactions, a size of 3,955,272 bytes and a weight of 3,992,769 units, within a whisker of the 4 million limit. Its large transaction carries one JPEG inside an ord envelope: a Taproot Wizard drawing by the developer Udi Wertheimer, who worked with Luxor’s chief executive Nick Hansen to get it into a Luxor block, CoinDesk reported the next day. A transaction that large is non-standard, so nodes with default settings will not relay it; it had to be handed to the miner directly, which is also how it could pay no fee.
Later inscriptions grew smarter rather than bigger. Recursive inscriptions can call a short list of approved endpoints to read other inscriptions and some chain data, so a generative art piece can load a code library inscribed once instead of copying it every time. The handbook’s chapter on recursion lists these endpoints. Burning follows a strict rule too: an inscription counts as burned only when it is sent to an output that starts with OP_RETURN, not to a so-called burn address.
Numbers that move: cursed inscriptions and the Jubilee
Inscription numbers, the counter collectors like to quote because low numbers sell at a premium, are less stable than they look. Inscriptions written in ways the early ord versions did not recognise, for example several in one transaction, were later indexed as “cursed” and, per the handbook, numbered from minus one downwards. In a blog post of 6 September 2023 titled The Great Renumbering, Rodarmor proposed making inscription numbers permanently unstable and blessing every cursed inscription. The compromise was the Jubilee: from block 824,544, mined on 6 January 2024 at 00:41 UTC, inscriptions that would have been cursed are “vindicated” and get positive numbers. The inscription id, made of the reveal transaction id plus an index, is the identifier that never changes.
BRC-20 tokens: balances written in JSON text
BRC-20 turned inscriptions into fungible tokens with almost no code. On 8 March 2023 at 04:16 UTC, block 779,832 confirmed a reveal transaction (id b61b0172…35735) whose envelope holds a few lines of plain text: the protocol name brc-20, the operation deploy, the ticker ordi, a maximum supply of 21,000,000 and a limit of 1,000 per mint. That inscription created ORDI, the first BRC-20 token, by an anonymous developer known as domo whose GitBook still hosts the BRC-20 experiment page.
The standard has three operations. Deploy creates the ticker and sets the cap, the mint limit and optionally the number of decimals (18 by default). Mint is an inscription that claims an amount up to the limit, credited to whoever first owns it. Transfer is a two-step affair: you inscribe a transfer inscription to your own address, then send that inscription to the recipient, at which point indexers move the balance. Every step is an inscription, so every step pays the two-transaction inscription cost and leaves a small inscribed output on the chain.
Nothing in Bitcoin enforces any of this. An indexer reads every BRC-20 inscription in order, ignores invalid ones (a mint above the limit, a transfer of tokens you do not have) and computes balances in its own database. The author’s own page is blunt: “These will be worthless. Use at your own risk.” It calls BRC-20 a fun experiment and strongly discourages financial decisions based on its design.
Two weaknesses followed from the design. The first is clutter: each transfer leaves dust-sized outputs carrying an inscription, which every full node must keep in its set of unspent outputs. The second is that balances are only as consistent as the indexers. When indexers interpret an edge case differently, two wallets can show different balances for the same address, and there is no on-chain referee to settle it. Choosing which version of the Ordinals rules an indexer follows is therefore a decision with money attached.
Runes: a token protocol designed to leave less clutter
Rodarmor was not a fan of fungible tokens. In his blog post of 25 September 2023 he wrote that “Fungible tokens are 99.9% scams and memes”. His argument was that if people were going to issue tokens on Bitcoin anyway, a simpler design might “serve as harm reduction” compared with BRC-20, which he said had “the undesirable consequence of UTXO proliferation”. The result was Runes: no native token, no off-chain data, balances held by unspent outputs like bitcoin itself.
The Runes specification puts every instruction in a runestone, an output whose script begins with OP_RETURN followed by OP_13 and a series of data pushes decoded into integers. A transaction may carry one runestone. It can etch a new rune (fix its name of 1 to 26 letters from A to Z, optional spacers such as the dots in LOBO•THE•WOLF•PUP, its divisibility, a symbol, an optional premine for the creator and optional open-mint terms), mint units of an existing rune or move balances between outputs. The specification says “A mint is open while all terms of the mint are satisfied, and closed when any of them are not.” Runes that a runestone does not assign go to the first output that is not an OP_RETURN.
Mistakes are punished hard. A malformed runestone becomes a cenotaph: the runes put into that transaction are burned, a rune etched in it can never be minted, and mints in it count towards the cap while the minted units are burned. The protocol was hard-coded to start at the fourth halving: the ord source code sets the first rune height on mainnet at four times the 210,000-block halving interval, which is block 840,000. That explains the crowd at the halving block.
Runes still need an indexer, because only software that replays every runestone since block 840,000 knows which output holds which balance. The difference with BRC-20 is that the balance moves with the output it lives in, so a normal transaction carrying a runestone is enough to transfer it, and no inscribed dust is left behind for each transfer.
Inscriptions, BRC-20 and Runes side by side
| Inscriptions (Ordinals) | BRC-20 | Runes | |
|---|---|---|---|
| What it is | a file attached to one sat | fungible token written as JSON inscriptions | fungible token protocol |
| Where the data sits | witness, inside an ord envelope | witness, inside an ord envelope | an OP_RETURN output (runestone) |
| Transactions per action | two (commit and reveal) | two per operation; transfer needs an extra send | one |
| Where balances live | the sat’s position in an output | an indexer’s database | outputs, computed by an indexer |
| Errors | unrecognised forms indexed as cursed | invalid operations ignored | cenotaph burns the runes involved |
| Start | inscription 0, 14 December 2022 | ORDI deploy, 8 March 2023 | block 840,000, 20 April 2024 |
| Official status | ord project handbook | author calls it an experiment | ord project specification |
Neither BRC-20 tokens nor runes are NFTs. They are fungible: one ORDI is like any other. Only inscriptions are unique items, and Rodarmor himself has long preferred to call them digital artifacts. If the vocabulary is new, see our glossary entry on NFTs.
What your own node sees, and what it does not
A plain Bitcoin Core node accepts and relays all of this as long as each transaction follows consensus and its policy rules. It does not decode envelopes, read JSON or apply first-in, first-out to sats. To see inscriptions and runes yourself, you run ord next to Bitcoin Core. The ord wallet guide says it requires Bitcoin Core’s transaction index (txindex=1) and its REST interface, and that “the Bitcoin Core wallet cannot create inscriptions and does not perform sat control”. In practice most users rely on someone else’s indexer through a marketplace, an explorer or a specialised wallet, which brings back exactly the trust a node was meant to remove.
Timeline: from Taproot to the 2026 shake-out
| Date (UTC) | Event | Block or source |
|---|---|---|
| 14 November 2021 | Taproot activates | block 709,632 |
| 14 December 2022 | Inscription 0, a 100 by 100 PNG; ord 0.2.0 published | block 767,430; GitHub |
| 20 January 2023 | ord 0.4.0 released | GitHub releases |
| 1 February 2023 | 3,955,272-byte block carrying one JPEG inscription | block 774,628 |
| 8 March 2023 | ORDI deployed, first BRC-20 token | block 779,832 |
| 25 September 2023 | Rodarmor publishes the Runes design | rodarmor.com |
| 6 January 2024 | Jubilee: would-be cursed inscriptions get positive numbers | block 824,544 |
| 20 April 2024 | Fourth halving and Runes launch | block 840,000 |
| 10 October 2025 | Bitcoin Core 30.0 lifts the default OP_RETURN limit | bitcoincore.org |
| 27 February 2026 | Magic Eden announces it will stop Bitcoin and Ethereum trading | Decrypt |
| 12 May 2026 | Ord.io and Zap announce they will close on 1 June | The Block |
What Ordinals and Runes did to Bitcoin fees
Inscriptions compete for the same block space as payments, so busy inscription periods push up the price of a confirmation. Our lesson on the mempool and transaction fees explains the auction itself (fee rates in sat/vB, replacement, fee estimation). The numbers below show how far the auction moved around the Runes launch, using mempool.space’s averages of fees per block over windows of roughly half a day.
| Window (UTC) | Average fees per block | Note |
|---|---|---|
| 9 November 2023 | 1.59 BTC | earliest point of the three-year series |
| 19 April 2024 | 0.58 to 0.97 BTC | the day before the halving |
| 20 April 2024 | 5.64 to 13.13 BTC | Runes launch; block 840,000 alone paid 37.63 BTC |
| 21 April 2024 | 2.10 to 2.77 BTC | rush fading |
| 23 April 2024 | 2.30 to 3.00 BTC | still well above pre-halving levels |
| 29 to 30 April 2024 | 0.31 to 0.37 BTC | back below pre-halving levels |
| 8 to 9 November 2026 | 0.021 BTC | 1 sat/vB recommended at every tier |
Block 840,000 itself paid a median fee rate of 200 sat/vB, according to its page on mempool.space. Within ten days the average block was paying less than before the halving. That pattern repeated with every wave of inscriptions and mints in 2023 and 2024: a short spike as people competed to be first, then a quick fall once the novelty was gone.
The calm has lasted. Over the 1,008 blocks to block 975,158 (the week to 9 November 2026), miners earned 3,172.37 BTC, of which 22.37 BTC came from fees, or 0.71%, by our calculation from mempool.space’s reward statistics. That is the part of Ordinals history that matters most for Bitcoin’s long-term security: a fee market that wakes up for speculative bursts and then falls asleep again, while the subsidy keeps halving every four years.
Where the Bitcoin Ordinals market stands in November 2026
By count, inscriptions kept growing long after the excitement faded, largely because inscribing small text is cheap when blocks are empty. Ordiscan’s dashboard showed 127,828,563 inscriptions, 4,465 BTC of inscription fees in total and 34.4 GB of data on 7 November 2026, according to an article by CoinTurk (en.coin-turk.com) republished that day by cryptonews.net. The same report pointed out that inscription totals are not the number of unique owners or users.
The trading side tells a different story. On 27 February 2026, Magic Eden chief executive Jack Lu said the marketplace, once a large venue for Ordinals and Runes, would stop trading Bitcoin and Ethereum-based assets within two weeks and drop them from its wallet at the beginning of April, to focus on Solana and on what he called “the massive opportunity in iGaming”, Decrypt reported. On 12 May 2026, The Block reported that the Ordinals explorer Ord.io and its sister trading app Zap would close on 1 June; their creator Leonidas King said “In the end we ran out of money and don’t see a path forward.”
Token valuations are small by crypto standards. On CoinGecko’s category pages, read on 9 November 2026 around 10:30 UTC, the BRC-20 category had a market capitalisation of about $83.4 million with $10 million of trading over 24 hours, ORDI accounting for $65.3 million of it. The Runes category stood at about $125.2 million, led by DOG at $75.3 million, with only about $448,000 traded in 24 hours. Thin volume means such figures can move a lot on small orders and may overstate what holders could actually sell.
The spam argument and the OP_RETURN change
Ordinals reopened an old Bitcoin argument about what block space is for. One side calls inscriptions spam: data that has nothing to do with payments, that fills blocks, raises fees for ordinary users and grows the burden on node operators, and that slipped in through a loophole in the witness discount. Some node operators filter inscription transactions out of their own mempools. The other side answers that fee-paying transactions that follow the consensus rules are valid by definition, that filters at a minority of nodes cannot stop a miner who receives transactions directly, as block 774,628 showed, and that fees from inscriptions help pay for security as the subsidy shrinks.
The quarrel moved from the witness to OP_RETURN in 2025. For years Bitcoin Core relayed only one small OP_RETURN output per transaction, up to 83 bytes of script. The Bitcoin Core 30.0 release notes (10 October 2025) raised the default -datacarriersize to 100,000, which in practice removes the limit, and state that “Multiple data carrier (OP_RETURN) outputs in a transaction are now permitted for relay and mining.” Operators who disagree can set -datacarriersize=83 to return to the old rule. How that decision was made, and the reactions it drew among node operators, belong to the later lesson on Bitcoin governance and BIPs in this path.
Risks: how people lose inscriptions and tokens
Spending an inscription as a fee by accident
This is the most common loss, and it follows straight from the first-in, first-out rule. An inscription sits on a sat inside an ordinary output, usually holding a few hundred sats. A wallet that does not track sats sees only a small amount of bitcoin and may pick that output to pay for your next transaction. Depending on the order of inputs and outputs, the inscribed sat can end up with the recipient, in change at a new address or in the fee, where it goes to the miner. The ord handbook’s own warning that Bitcoin Core’s wallet “does not perform sat control” is the reason: general-purpose wallets, including many hardware wallet apps, are not built to protect inscriptions unless you freeze those outputs with coin control. There is no undo for a sat that has been paid to a miner.
The same logic applies to runes, which live in outputs too, and to exchanges. Sending an inscription or a rune to a deposit address of an exchange that does not support it usually means the exchange’s systems treat it as a few sats of bitcoin. Recovery then depends entirely on the exchange’s goodwill.
Depending on indexers and marketplaces
Your inscription is on chain for good, but knowing that you own it, seeing it and selling it depend on services. When BRC-20 indexers disagree, balances differ between wallets. When a marketplace or explorer leaves, as Magic Eden and Ord.io did in 2026, users must move to another tool that reads the same rules. Wallets built on one service’s indexer may show nothing at all once that service is gone, even though nothing has changed on the chain.
Fake mints, fake claims and drainers
Hype periods attract copycat mint pages, fake “claim your ORDI” or airdrop sites and links to look-alike marketplaces. They ask you to connect a wallet and sign a transaction or a partially signed transaction that, in reality, hands over your inscriptions or bitcoin. A legitimate mint never needs your seed phrase, and no site should ever ask you to type it. Our guide to avoiding crypto scams goes through the common patterns.
A practical checklist
- Keep collectibles and tokens in a wallet that supports Ordinals and Runes and shows them, and keep everyday bitcoin in a separate wallet.
- Never use the wallet that holds inscriptions to pay for anything unless it lets you freeze the outputs that carry them.
- Check that an exchange supports a specific rune or BRC-20 token before sending it there, and send a small test amount first.
- Read what a signature request does before approving it, and reach marketplaces by typing their address rather than through links in messages.
- Back up the seed phrase offline, as explained in our seed phrase security guide, and never share it or a descriptor containing private keys.
How inscriptions and Bitcoin tokens are taxed in the US and the UK
In the United States, the IRS defines a digital asset as any digital representation of value recorded on a cryptographically secured ledger and names NFTs as an example in its frequently asked questions. Its FAQs say you recognise a gain or loss when you dispose of NFTs, even when no broker form is issued. Buying an inscription or a token with bitcoin is a disposal of the bitcoin, and so is paying an inscription fee in bitcoin, as with any network fee paid in crypto. A broker may leave NFT sales with combined proceeds under $600 in a year off Form 1099-DA; the IRS says the gain or loss must still be reported. Our US crypto tax guide gives the full picture.
Across the Atlantic, HMRC applies UK capital gains tax to disposals of tokens, including swaps and spending. Fungible tokens such as runes or BRC-20 tokens of one type go into their own pool with an average cost, while NFTs are unique and are not pooled, so each inscription keeps its own cost. Our UK crypto tax guide explains pooling and the records HMRC expects. In both countries, low-value collectibles that are hard to price still need a value in dollars or pounds on the day of each disposal.
Ordinals after the hype: data that stays, services that come and go
Nearly four years after inscription 0, the inscriptions themselves are as permanent as any Bitcoin transaction. The meaning attached to them (rare sats, BRC-20 balances, rune ownership) lives in indexers run by a handful of teams and companies, and two well-known services, Magic Eden’s Bitcoin marketplace and Ord.io, announced their exit in 2026. In the week to 9 November 2026, all fees together, inscriptions included, came to less than 1% of what miners earned.
The next lesson in this path turns to Bitcoin layer 2s and sidechains, the other way people have tried to do more with bitcoin than send it. The dates to watch for Ordinals and Runes are less dramatic: the next releases of ord, the indexers that wallets switch to after this year’s closures, and the next halving in 2028, which will cut the subsidy again and make every fee, including the ones inscriptions pay, a larger part of what keeps Bitcoin’s miners going.
