Ordinals’ Second Act: Bitcoin as a Data Layer

The traders left; the builders stayed

Judged by its token, Ordinals looks like a trend that already came and went. ORDI, the BRC-20 flagship that once stood in for the whole experiment, trades near $4.36, with a market value around $91.6 million and a rank in the low 300s, roughly 95% below its March 2024 peak of $95.52, according to CoinGecko. Bitcoin itself sits near $83,927, about a third under the all-time high of $126,080 it printed on 6 October 2025, per CoinGecko. A year on from that top, a milestone we weighed in our look at Bitcoin one year after the record, the easy money in Bitcoin NFTs is a memory, and the crowd that chased it has moved on.

The on-chain record tells a different story. More than 127 million inscriptions had been written to Bitcoin by the middle of 2026, and the clearest signal of a second act landed three weeks before this article went to press: a developer who goes by Patches shipped a free, open-source desktop app that treats the Bitcoin blockchain as something its inventor never intended it to be, a browsable data layer. The speculative phase is over. The infrastructure phase, quieter and far less profitable, is underway. This is the story of how Ordinals stopped being a trade and started being a place to build, and of the base-layer fight that almost stopped it.

What an Ordinal actually is, in one minute

Ordinal theory, the idea at the root of all of this, is simple. Every satoshi, the smallest unit of Bitcoin at one hundred-millionth of a coin, can be numbered in the order it was mined. Once a satoshi has a serial number, you can attach data to it: an image, a song, a web page, a 3D model, even a piece of software. That attached data is an inscription, and the numbered satoshi carrying it is, in the words of creator Casey Rodarmor, a digital artifact.

The mechanics matter for what follows. An inscription is not stored in an OP_RETURN output, the small data field Bitcoin has carried since 2014. It lives inside the Taproot script-path witness, tucked into an unexecuted branch of script that begins with an OP_FALSE OP_IF envelope. Because of the witness discount introduced with SegWit, that witness data is billed at a quarter of the usual weight, which is why inscribing up to roughly four megabytes of content became affordable at all. The upload happens in two transactions, a commit and a reveal, and once revealed the data sits permanently on the base chain, not on a server that can go dark.

Rodarmor released the protocol on 21 January 2023 and refused from the start to call the results NFTs, telling CoinDesk that the term felt tainted and that his real aim was to make Bitcoin fun again. The distinction he insisted on, that the data is the asset rather than a pointer to one, is exactly the property that makes the second act possible. If the picture, the code and the logic all live on-chain, then Bitcoin is not just recording ownership. It is storing the thing itself.

Recursion is the engine of the second act

The feature that turned a collectibles protocol into a building platform arrived in June 2023 and is still the least understood part of Ordinals: recursion. A recursive inscription can reach out and pull in the content of an inscription that already exists, through a small set of whitelisted endpoints documented at docs.ordinals.com. An inscription can request another by ID at a path like /content, or ask the chain for its block height, its children, or its parents. Backwards compatibility is guaranteed, so a library inscribed once can be referenced forever.

That flips the cost model on its head. Instead of embedding a copy of a JavaScript rendering engine in every single artwork, a project inscribes the engine once and every later piece points to it. OnChainMonkey became the standard example, inscribing the Three.js and p5.js graphics libraries on-chain for anyone to use, then building a 3D collection where each piece weighed less than a kilobyte because the heavy code already lived on Bitcoin, as OKX and others documented. Builders started calling the pattern a package manager for Bitcoin, or simply BTC Lego.

The vision was spelled out early. When recursion shipped, the pseudonymous developer Leonidas told The Block that the simple change unlocked many powerful use cases, and that storing art more efficiently could have saved over a million dollars in fees for a single large collection. The same piece floated a stranger possibility: an internal internet inside Bitcoin, where marketplaces would turn into blockchain web browsers. In 2023 that read like a reach. In 2026 someone built it.

A browser for the blockchain

On 22 September 2026, Patches, the developer behind the Turtle Moon project, released Bitcoin Browser v0.1.0 as a public beta, a free desktop application for macOS, Windows and Linux published under an MIT license on GitHub. In an interview with Genfinity in early October, Patches put the thesis plainly: the app costs nothing, and the code sits on GitHub under an MIT license. The point was not to sell anything. It was to prove that Bitcoin can be read as a data layer rather than only a financial ledger.

What the browser actually does is the interesting part. It renders the full range of content types already sitting on-chain, from images and audio to HTML, SVG, PDF, Markdown, fonts and 3D model formats like glTF, GLB and STL. It indexes and displays more than 100 million inscriptions. It reads Runes supply and issuance data. It will install and run a local Bitcoin Core node in one click. And it ships a recursive dependency inspector that traces an inscription back through the layers of other inscriptions it depends on, which is the on-chain equivalent of viewing a web page’s sources. There is even an Ask Bitcoin panel that wires a locally run AI model, through LM Studio or Ollama, to query chain data on the user’s own machine rather than a remote API, an approach that rhymes with the local-first ideas in our coverage of wallets that act on your behalf.

Whether a desktop client finds a mainstream audience is an open question; getting ordinary users to run node software is the same onboarding wall that still trips up self-custody, as we documented in our report on where the wallet funnel leaks. But as a statement of intent, the Bitcoin Browser is the cleanest artifact of the second act yet. It treats the chain the way a web browser treats the internet.

From JPEGs to executable code

If recursion proved you could store software on Bitcoin, the next step was to run it. That is where the data-layer thesis stops being about art. The same witness space that holds a monkey picture can hold program bytecode, and in 2026 that is exactly what some builders started doing.

The most ambitious example is OP_NET, which went live on Bitcoin mainnet in March 2026 and bills itself, per The Defiant, as a deterministic execution layer that runs directly on Bitcoin as it exists today: no soft fork, no new opcodes, no separate chain and no separate token. It introduces Pay-to-OPNet addresses that use Bitcoin’s existing output formats to embed WebAssembly contract bytecode into transactions. Developers write contracts in languages like Rust or an AssemblyScript dialect, compile them to WASM, and deploy them with BTC as the only gas token. A Bitcoin-native DEX called MotoSwap and an OP-20 token standard already run on top of it, with a stablecoin extension on the roadmap.

OP_NET is not Ordinals, and the distinction is worth keeping clear: it is a separate metaprotocol with its own trust assumptions. But it rides the same primitive Ordinals made famous, arbitrary data committed to the witness, and it shares the same worldview. So do the fungible-token cousins that grew out of the inscription boom: BRC-20, which encodes token rules as plain JSON text inscriptions, and Runes, Rodarmor’s leaner follow-up that lives in OP_RETURN. Seen together, they describe a single shift. The block space that the market once valued only for cartoon avatars is being repurposed into a general substrate for data and, increasingly, computation.

The quiet era, by the numbers

The building is real, but so is the slowdown, and the data captures both at once. An empirical study of on-chain activity from late 2022 through mid-2026, The Evolution of Bitcoin Ordinals Inscriptions, counted 127,609,366 inscriptions across roughly 118 million transactions that carried at least one. That is a staggering amount of data for a chain often described as a simple ledger, and it keeps growing even with prices flat.

What changed is the character of the activity. The same study found the median inscription fee rate fell from about 32 satoshis per virtual byte in the early frenzy to just over 1 sat/vB, and the median transaction fee dropped from around 5,076 satoshis to 196. The batching stunts of the boom faded too: the mean number of inscriptions per transaction slipped from 1.31 to 1.09, and the share of multi-inscription transactions fell from 2.11% to 0.44%. In plain terms, inscribing is now cheap and deliberate rather than frantic and speculative. For a builder, a floor of roughly 1 sat/vB means you can write a meaningful piece of content for a few dollars. For Bitcoin’s long-term economics, as a later section explains, that same cheapness is a problem.

The war over what Bitcoin is for

None of the second act is guaranteed, because a loud faction of Bitcoin developers believes it should never have happened. The objection is philosophical before it is technical. To the minimalist camp, Bitcoin is money and only money, and every byte spent on a JPEG or a token is a byte stolen from the monetary network. Luke Dashjr, one of the most senior Bitcoin Core contributors, has called Ordinals a vulnerability that exploits a gap in relay policy and should be filtered out, telling FXStreet that inscriptions bypass the data limits node operators have been able to set since 2013.

The other camp answers with economics. Block space is a market, they argue, and whoever pays the fee has the right to use it, whether the payload is a payment, a token or a picture. This is not a new quarrel. It is the same data-carrier fight that has flared since OP_RETURN was first standardized in 2014, only now it runs at the scale of a hundred million inscriptions, and the outcome decides whether Bitcoin can host a data layer at all. If the base layer turns hostile, every browser and execution engine built on top of it is living on borrowed time.

BIP-110, and the vote that settled it for now

In 2026 the fight came to a head. BIP-110, a proposed temporary soft fork, tried to curb non-financial data by force. As The Block reported, it would have capped most transaction outputs at 34 bytes, restored the old 83-byte OP_RETURN limit, capped certain witness elements at 256 bytes, and temporarily disabled several Taproot features that inscriptions rely on. It was, in effect, a direct attempt to switch off the data layer.

It failed, and not narrowly. Mandatory signaling opened at block 961,632 in August 2026, where BIP-110 nodes began rejecting blocks that did not signal support. The proposal needed 55% of miners and reached roughly 2.6%, per crypto.news. The would-be enforcing chain mined a couple of blocks, fell far behind, and about 99.85% of hashpower simply stayed on the main chain. Michael Saylor, whose firm holds one of the largest corporate Bitcoin treasuries, framed the non-event as proof of the system working: Bitcoin worked exactly as designed, he said, because the proposal was free to fork and the network was free not to follow, telling The Crypto Times that Bitcoin was working as designed.

The contrast with the other direction of travel is sharp. Bitcoin Core v30.0, released on 10 October 2025, moved the opposite way on data policy, raising the default datacarriersize for OP_RETURN from 83 bytes to 100,000, as Decrypt and Unchained covered. That was a relay-policy change, not a consensus rule, and it touches OP_RETURN rather than the Taproot witness Ordinals use, so it matters more to Runes and Stamps than to inscriptions directly. But the signal was unmistakable. The maintainers of the dominant node software loosened the data valve in the same window that the restriction camp tried, and failed, to tighten it.

The shakeout that cleared the tourists

To understand why the market can look dead while development continues, look at who left. The casualties of 2026 were almost all service businesses, the layer of companies that sold convenience on top of the protocol, rather than the protocol itself.

OrdinalsBot, the first inscription service to open after the protocol launched in 2023, wound down in August 2026, selling its brand and tech stack and blaming a shrinking market. Magic Eden, once the dominant Bitcoin NFT marketplace, announced its exit from Bitcoin and EVM chains in late February 2026 and finished winding down its Bitcoin markets in March, with leadership citing a brutal split in which Bitcoin and EVM accounted for most of the cost and a small share of revenue while Solana supplied the overwhelming majority of volume. The Ord.io explorer, run by the same Leonidas who once imagined marketplaces becoming browsers, shut in mid-2026. The irony was not lost on anyone when, months later, a different developer shipped the browser instead.

What remained is leaner and, arguably, healthier. Inscription and trading activity consolidated onto a handful of survivors: UniSat, OKX, the Xverse and Leather wallets, Gamma.io for drops, and Horizon Market, which launched on 30 March 2026 and, per KuCoin, added tooling for recursive inscriptions and BRC-20 tracking that users had been asking for. The service-business layer contracted hard. The protocol layer, the thing that actually stores the data, did not blink.

Where the money still is

The collectibles that defined the first act did not vanish; they just stopped making headlines. Value concentrated into a short list of recognized collections, which traders increasingly quote in BTC rather than dollars so the number is not swamped by Bitcoin’s own swings. NodeMonkes, the first 10,000-piece profile-picture set on Bitcoin and the source of the first Bitcoin NFT sale above a million dollars, remains the reference blue chip. Taproot Wizards, famous for mining a nearly four-megabyte block in Bitcoin’s early inscription days, is less a picture collection than a campaign to change Bitcoin’s base layer. Bitcoin Puppets and the Runestone airdrop round out the set that survivors still watch.

Alongside the pictures sits a quieter market in rare satoshis, where collectors pay up for sats with historical significance: the very first coins mined, satoshis from the 2010 pizza transaction, or vintage 2009 blocks. And on the fungible side, BRC-20 tokens like ORDI and the Runes category, dominated by the DOG token, give the ecosystem a speculative pulse. None of this is the growth story. These are mature, thin markets that trade sideways. The reason to pay attention to Ordinals in 2026 is not the floor price of any collection. It is what the block space is being used for.

What an on-chain data layer means for Bitcoin’s security budget

Here is the question that makes Ordinals a base-layer story and not just a culture-war one. Bitcoin pays miners in two ways: a block subsidy, currently 3.125 BTC and due to halve to about 1.5625 BTC around 2028, and transaction fees. The subsidy shrinks forever by design. Fees are supposed to pick up the slack. In the quiet blocks of 2026, fees are a rounding error, often just 1% or 2% of what a miner earns, which is precisely the pressure we traced through the economics of the big public miners a year after the top.

This is where the two faces of cheap block space collide. A floor near 1 sat/vB is a gift to builders and a warning to miners. The optimistic reading of the second act is that a genuine data and compute layer, apps and not just art, could generate durable, structural demand for block space that keeps paying miners long after the subsidy fades. The pessimistic reading is that today’s activity proves the opposite: there is plenty of data being written, and it is almost free, so it is not paying for security in any meaningful way. Whoever is right, the contest for space is becoming its own market, with its own intermediaries and its own extractable value, a dynamic our piece on the MEV oligopoly explores on the settlement side. A thriving data layer is one of the few credible answers to the security-budget problem. An almost-free one is not an answer at all.

The bear case, and the skeptics who may be right

Enthusiasm should not paper over the holes, and there are several. The first is the one just described: if the data layer cannot command real fees, it is a hobby, not an economy. Demand has to deepen, and there is no guarantee it will.

The second is centralization creeping in through the back door. Ordinal theory, recursion and metaprotocols all depend on off-chain indexers, the software that reads raw Bitcoin and decides what a given inscription or token balance means. The reference indexer, ord, and a few explorers do most of that work. If almost everyone trusts the same handful of indexers and endpoints, the promise of a verifiable system with no centralized API is partly aspirational. OP_NET, for its part, validates its WebAssembly execution off-chain, a real trust assumption that sits outside Bitcoin consensus. A data layer is only as trustless as the clients that read it.

The third is that the truce is fragile. BIP-110 lost, but the restriction camp did not disband, and a future Core release or a shift in miner policy could reopen the fight. The fourth is legal. In the United States the Securities and Exchange Commission largely retreated from NFT enforcement after its early settlements with Impact Theory and Stoner Cats, and a February 2025 staff statement signaled that most meme coins are not securities, which by extension covers much of BRC-20 and Runes. But staff guidance is not law, and a different commission could revisit the question. For builders, the base layer finally feels safe. The regulatory layer is a different matter.

What to watch next

The trade is over. The experiment is not. The next chapter of Ordinals will be written in commits and node counts rather than floor prices, and a few signals will tell you whether the second act is real or just a nice idea.

• Client adoption. Whether the Bitcoin Browser and tools like it attract real users or stay developer toys. Watch GitHub activity and how many people actually run a local node.
• On-chain compute. Whether OP_NET and other metaprotocols find genuine product-market fit, and whether anyone trusts off-chain-validated execution enough to put value in it.
• The next data fight. BIP-110 lost, but the restriction camp is still there. A future Core default or a change in miner policy could reopen the question.
• Fees and the security budget. Does any of this activity start to move the fee needle before the 2028 halving, or does block space stay almost free?
• Marketplace stability. Whether UniSat, OKX, Xverse and Horizon Market hold the line after the exits of Magic Eden and OrdinalsBot.

Casey Rodarmor wanted to make Bitcoin fun again. The market took that literally, bid it up, and left when the fun stopped. What it missed is that somewhere in the noise, Bitcoin picked up a new capability it is unlikely to lose: the ability to be read, not just spent. The tourists are gone. The chain they scribbled on is now a place people build.

Frequently Asked Questions

What is the Bitcoin Browser and is it free?

The Bitcoin Browser is a free, open-source desktop app released by the Turtle Moon developer Patches on 22 September 2026 under an MIT license for macOS, Windows and Linux. It renders inscriptions, reads Runes data, can install a local Bitcoin Core node, and includes a local AI panel, treating the Bitcoin blockchain as a browsable data layer rather than only a payment ledger.

What are recursive inscriptions in plain terms?

A recursive inscription is an inscription that pulls in content from another inscription already on Bitcoin instead of storing its own copy. A project can inscribe a code library or 3D asset once and have many later pieces reference it, which slashes costs and makes larger, interactive on-chain works possible. Developers often compare it to a package manager for Bitcoin.

Why is ORDI down about 95% if Ordinals are still active?

ORDI is a single BRC-20 token that became a proxy for speculation, and its price reflects trader sentiment, not protocol usage. Inscription activity, measured by data written to the chain, kept growing past 127 million inscriptions even as token prices collapsed. The speculative market and the building activity moved in opposite directions.

Did BIP-110 ban Ordinals?

No. BIP-110 was a proposed temporary soft fork that would have restricted the data features Ordinals rely on, but it failed. Mandatory signaling in August 2026 drew only about 2.6% miner support against a 55% threshold, and roughly 99.85% of hashpower stayed on the main chain, so inscriptions continue to work normally.

Can you really run applications on Bitcoin without a Layer 2?

To a limited degree, yes. Projects like OP_NET embed WebAssembly contract code directly in Bitcoin transactions and run a deterministic execution layer on the base chain, with BTC as the gas token and no separate chain or token. The tradeoff is that such systems validate execution off-chain through their own software, which is a trust assumption that sits outside Bitcoin’s own consensus rules.

Marcus Okafor is a senior editor at HOGE Wire covering Bitcoin’s base layer, Ordinals and the economics of block space.