Twitter founder Jack Dorsey is working on TBD — a whole new internet identity and trust layer, marking one of his first attempts in the cryptocurrency field, aiming to operate entirely on a peer-to-peer architecture while integrating with existing Web2 services.
Many believe TBD sounds a lot like Web3 decentralized identity service ENS, but Dorsey does not see it that way. As a staunch Bitcoin believer (a term that has become synonymous with cryptocurrency skeptics), he feels it is necessary to distinguish himself from the mainstream and ensure TBD is never seen as just another Web3 project.
Specifically, Dorsey realized that Web3 — a vision of a decentralized internet built on open protocols and blockchains, encompassing identity, finance, and social layers — is not the system he seeks.
He realized that there is a fundamental opposition between the reality of Web3 and its vision: it is incompatible with the existing infrastructure of the internet and is committed to completely replacing the latter.
Definition of Web5 (Source: TBD PPT)
As TBD's proclaimed core goal is 'decentralization' without compromise, Dorsey chose to build this system on Bitcoin. In his view, this alone is sufficient to ensure that TBD 'does not belong to Web3,' making it necessary to create a new term for such systems.
To this end, Dorsey half-jokingly proposed the term 'Web5,' both as a jab at Web3 and as a tribute to HTML5. HTML5 is the foundation of today's internet and the last major technological attempt that propelled the evolution of the web 15 years ago.
From TBD's whitepaper, Dorsey views Web5 as a peer-to-peer network based on Bitcoin as the foundational consensus layer and the Lightning Network as the payment network. It revolves around three pillars: 1) self-owned decentralized identifiers, 2) verifiable credentials, and 3) decentralized network nodes for data storage and message forwarding.
'In today's web, identity and personal data have become third-party property. Web5 will bring decentralized identity and data storage to your applications. It allows developers to focus on creating an excellent user experience while returning ownership of data and identity to individuals,' states TBD's official website.
Combining the Twitter Files (a series of internal documents revealing how Twitter was forced to censor sensitive content) with TBD's goals, we have reason to believe that Dorsey essentially wanted to build 'freedom technology.'
His vision for this technology is exemplified in Nostr — an open, decentralized, censorship-resistant messaging protocol designed to address the content moderation and censorship issues of current centralized social media platforms.
For newcomers, Nostr operates similarly to blockchain: each user generates a private key (which must be kept secret) and uses their public key as an identifier. All messages (called 'notes') are encrypted and signed by their creator and can be verified by others.
Nostr does not rely on a single platform to store user data; instead, it stores and forwards messages through independently operated servers (called 'relays') based on simple, open rules. Since users can choose any relay or build their own nodes, no central authority can effectively censor or delete content. The protocol itself is extremely simple, defining only the message format, signature, and publishing method, allowing developers to build additional features on top of this, such as private messaging, image support, etc.
After witnessing how centralized social media companies operate, Dorsey became obsessed with another vision: returning control of network applications from corporations and manipulatable nodes back to users. His interest in and support for Nostr indicate that we have the opportunity to transcend Web2's 'server ownership'; unsettlingly, this model has permeated Web3.
Now, while we cannot know whether Dorsey was being sarcastic when he coined the term 'Web5,' it is certain that he has indeed grasped some key points. Although TBD has never released a product and has ceased operations, Dorsey's insights into Web3's flaws remain precise and prescient.
However, his vision for Web5 need not be limited to Bitcoin and the Lightning Network.
In our view, the significance of Web5 far exceeds its components; it is not merely a 'peer-to-peer network' dedicated to Bitcoin consensus or a framework built around decentralized identifiers.
Furthermore, Web5 is not a semantic 'rebranding' or gimmicky marketing strategy; it is a substantive shift in the internet industry back to its roots.
In our view, Web5 is a mesh structure composed of peer-to-peer networks that connect various PoW (Proof of Work) and UTXO consensus layers, channel networks, and other yet-to-be-conceived systems. More abstractly, Web5 is a thriving decentralized application (dApp) ecosystem built upon this peer-to-peer mesh structure.
The network topology of decentralized and peer-to-peer networks (Source: CKB Eco Fund)
The underlying architecture is the core difference between Web5 and Web3.
Web5 is built upon a truly decentralized, peer-to-peer topology, which is a direct result of adopting PoW consensus and the UTXO model. It does not view blockchain technology as the sole core, but rather envisions a series of open internet protocols enhanced by the latest cryptographic primitives, collectively pushing the internet into a new era.
In contrast, Web3 has failed to deliver on its promises of decentralization, censorship resistance, permissionless access, and self-custody of data and assets, rooted in flaws in its underlying architecture, especially the decision to choose PoS (Proof of Stake) and the Account model.
The current state of Web3
Today's Web3 is a collection of countless 'nominally decentralized' networks. Since the rise of MetaMask and Infura in 2017, these networks have rapidly shifted to a client-server topology.
Despite the arduous research and engineering efforts, we conclude that this outcome is an inevitable product built on the PoS and Account models.
While we respect the principled efforts of many to combat this trend, we do not believe that the flaws in the client-server topology can be fixed. Before analyzing the reasons, let us first examine the current state of Web3.
In February 2009, Satoshi wrote in a post: 'The fundamental problem with traditional currency is the trust required to operate it.'
Observing today's Ethereum, 'trust' seems to be on the rise. Although staking pool operators and block builders are not strictly trusted third parties (TTP), they have evidently become increasingly important privileged roles.
Percentage of ETH staked by each entity of the total stake (Source: dune.com)
The liquidity staking protocol Lido controls about 28% of the total staked ETH, while Coinbase controls about 11%, raising concerns about governance and validation power being concentrated among a few industry giants. Beaverbuild and Titan Builder produce about 89% of Ethereum's blocks, further exacerbating concerns over censorship resistance and maximum extractable value (MEV) control.
Moreover, while the Ethereum base layer is decentralized enough on many metrics — especially compared to most Web3 projects — the horizontal scaling path chosen by its community has led to systems that clearly rely on trust assumptions.
These systems rely on centralized infrastructure providers acting as 'servers,' while users become 'clients' dependent on these servers for network functionality and access. This architecture is indistinguishable from traditional Web2, deviating from the original decentralization goals of Web3.
Taking Rollups as an example, the reliance on centralized sorters creates severe bottlenecks. Ideally, a single entity completely controls transaction sorting and packaging, forcing users to trust its behavior — which goes against the cryptocurrency principle of being 'trustless.' In the worst-case scenario, this entity could completely halt the chain's operation, as seen this year when Ethereum Layer 2 project Linea paused sorting due to an attack on a decentralized exchange within its ecosystem.
Worse still, Linea is not an exception. Almost all Ethereum Rollups operate in a centralized manner, allowing operators to censor transactions or indefinitely halt the chain's operation. If a chain can be paused at will, what is its purpose? Traditional centralized databases clearly perform better; why run a chain?
Even if we ignore these risks and naively assume that the trusted third parties of current Web3 infrastructure are reliable, we cannot escape the fact that, as Nick Szabo pointed out years ago, these third parties are essentially security vulnerabilities, and countless security incidents have repeatedly confirmed this.
For example, in July 2023, the cross-chain protocol Multichain lost over $125 million due to suspected insider embezzlement. The root of the vulnerability lay in its CEO, Zhao Jun, controlling most of the platform's multi-party computation (MPC) keys, and he was arrested by the Chinese police. A year prior, a similar incident occurred with Axie Infinity's Ronin Bridge, where North Korean hacker group Lazarus stole over $600 million in user funds by controlling 5/9 validator private keys.
In addition to trust and security issues, horizontal scaling (i.e., executing transactions through side chains) has also led to severe liquidity fragmentation and infrastructure cost issues. Currently, there are dozens of Ethereum Layer 2s, most of which have become ghost chains due to their inability to attract sufficient liquidity.
The total TVL (Total Value Locked) of the top two Layer 2 projects, Arbitrum and Base, exceeds the total TVL of the remaining 18 Layer 2 projects combined ($32.12 billion vs. $11.43 billion). (Source: L2Beat.com)
Liquidity attracts traders, trading volume generates liquidity, and the combination of the two attracts dApp developers. Liquidity fragmentation traps Layer 2 in a network effect dilemma: chains that break through the critical point continue to grow, while others gradually wither, ultimately concentrating liquidity and user activity among a few winners.
Although these systems are referred to as Rollups, they remain blockchains with scarce block space. This means that successful Layer 2s will still face the same scalability and fee volatility issues as the underlying chain, leading to an increasingly complex demand for Layer 3 security assumptions.
The increase in the number of chains means higher infrastructure costs — after all, someone needs to maintain all the Rollups. Even after Ethereum's EIP-4844 upgrade introduced data blobs and reduced Layer 1 data availability (DA) costs by 100 times, the average monthly cost to run a Rollup still reaches $10,000 to $16,000 (assuming 2 million transactions per month).
Under the same assumptions, the cost of Layer 1 alone reaches $25,000, while using alternative DA layers like Celestia or EigenDA is cheaper by several orders of magnitude. Unfortunately, for many Layer 2s, the fees paid by users are insufficient to cover infrastructure costs, meaning 'server' operators must bear the expenses themselves. This financial burden raises the entry barrier for new participants, giving an advantage to well-funded entities and further exacerbating centralization.
In contrast, PoW+UTXO chains achieve scalability through vertical scaling (adding payment channels or state channels on top of the base layer). Validation remains low-cost and accessible, allowing users to run full nodes or light client nodes on ordinary hardware, ensuring broad participation in the network. By managing state through UTXO, users only need to validate transactions relevant to themselves without relying on centralized intermediaries.
Protocols like the Lightning Network, Ark, and RGB++ are exemplars of this path. Users can directly establish payment channels, with security anchored in the underlying layer's PoW consensus. There is no need for cross-chain bridges or centralized sorters that could become points of failure, maintaining the network's peer-to-peer topology and ensuring true decentralization and censorship resistance.
How did Web3 get to this point?
To understand why we are building Web5, we first need to clarify where Web3 went wrong. The best way to do this is to examine the design choices in Ethereum's history.
First, we must clarify that we hold no prejudice against Ethereum (or any other chain). Instead, we use it as an example to analyze the flaws of the PoS + Account model.
In this category, Ethereum is the most decentralized chain at the technical, ideological, and community levels, and it is also the origin and main construction platform of the Web3 narrative. Criticizing Web3 by citing other chains is clearly unfair. Furthermore, we believe that the Ethereum community's efforts to achieve the goals of Web3 are sincere, and its failures stem from decisions made a decade ago.
Ethereum's first error
Ethereum's first error stemmed from its initial attempt to turn the blockchain into a 'world computer.' In this article, we deeply explain why this is fundamentally a bad idea, so here we only present the conclusion: blockchains are for validation, not computation.
When Bitcoin developer Gregory Maxwell pointed this out more than nine years ago, Vitalik Buterin vehemently refuted it.
Looking at the current state of Ethereum, it seems the argument of 'everything on-chain' has been abandoned. Any and all attempts to expand the world computer are through 'expanding on another chain,' i.e., the more widely known Rollup-centric roadmap.
In other words, the Ethereum community has abandoned its original ideals in favor of a more technically conservative 'modular blockchain' path. Today, the base layer is used for validation and final settlement, while adjacent chains or Layer 2 are responsible for transaction processing.
Ethereum's second error
However, this shift failed to establish a peer-to-peer network, rooted in Ethereum's second architectural flaw: abandoning Bitcoin's UTXO model in favor of the Account model.
At the time, Vitalik presented two arguments to justify this transition: 1) 'UTXO is theoretically complex, even more so in implementation'; 2) 'UTXO is stateless, making it difficult to support complex applications that require state management (such as various smart contracts).'
Although these arguments may have been valid at the time and seen as important innovations, the industry has since made significant progress. Statefulness — the maintenance and updating of the blockchain's 'state' or the collection of all current data, balances, and conditions generated by past transactions — is indeed necessary for computation, but the Account model is not the only path to achieving statefulness.
In 2017, Cardano launched the extended UTXO (eUTXO) model; in 2019, Nervos proposed the Cell model — a stateful general-purpose UTXO model; recently, developers of BitVM even achieved state computation on Bitcoin through Taproot.
Looking back, choosing the Account model over the UTXO model seems to have been a short-term decision: although it made it easier for developers to quickly build dApps, it sacrificed many inherent advantages of the UTXO model.
Crucially, the UTXO model allows for true ownership of assets and data — precisely the core goal claimed by both Web3 and Web5.
The UTXO model does not have accounts in the traditional sense, but tracks asset ownership and transfers through addresses and unspent transaction outputs (UTXO).
UTXO is a unit of cryptocurrency that has been received but not yet spent, associated with the address that specifies who can spend it. In this model, users manage the funds corresponding to UTXOs by controlling private keys. The sum of these UTXOs constitutes the user's available funds, without the need for traditional accounts.
In contrast to the Account model, accounts are divided into external accounts (EOA, controlled by private keys and capable of initiating transactions) and contract accounts (CA, i.e., smart contracts that cannot initiate transactions proactively and consist of code and data). The problem is that in the Account model, all non-native assets (tokens other than ETH in Ethereum) are managed by CAs. This means that non-native assets are second-class citizens in this model. The token balances displayed in user wallets do not represent actual ownership; these tokens are managed by the CAs controlled by the EOAs that created them.
Real-world cases illustrate the severity of this issue. Brian Pellegrino, co-founder and CEO of LayerZero, recently pointed out in a tweet that there is a severe vulnerability in the token contract of the cross-chain interoperability protocol Across: a function in the token contract allows the contract owner to transfer tokens from any wallet at any time. In short, the Across team can steal tokens from any user holding these tokens.
Worse, such cases are not isolated. Many token contracts contain similar functions that allow contract owners to issue or destroy tokens at will, or censor and confiscate user assets.
Centralized stablecoin issuers default to such functions (as necessary compliance measures), allowing them to confiscate tokens suspected of being illegally obtained (e.g., through vulnerabilities or theft).
In the UTXO model, all assets are directly controlled by the user's private key, making them first-class citizens. Taking Nervos CKB, which employs a stateful UTXO model, as an example, token contracts only define token logic (e.g., 'total supply of 1 million' or '50 tokens issued per block'), while the asset data recording user's balances (e.g., 'Alice holds 100 tokens') is stored in Cells directly controlled by the user (which can be viewed as stateful UTXO). This means that even if the token contract is attacked, hackers cannot steal user assets.
Ethereum's third error
Ethereum's third error is abandoning PoW in favor of PoS. The justifications for this decision include 'PoS has significant advantages in security, reducing centralization risks, and energy efficiency' and 'higher security at the same cost.' However, for many readers, it is now evident: PoS cannot replace PoW. For those who still doubt, refer to (Why Follow Satoshi) or (Why the World Needs Miners).
Moreover, time has provided evidence to rebut these arguments. Last year, Vitalik himself wrote a lengthy article warning of the inherent centralization risks of PoS. The following excerpt summarizes its core points:
One of the biggest risks of Ethereum L1 is that PoS may become centralized due to economic pressures. If there are economies of scale in participating in the core PoS mechanism, large stakers will naturally dominate the network, while small stakers will drop out and join larger pools. This will lead to increased risks of 51% attacks and transaction censorship. Besides centralization risks, there is also the risk of value extraction: a small group may seize value that should belong to Ethereum users.
Although Vitalik proposed several Ethereum-specific solutions in his text, we believe this is futile. The reliance on centralized power and trusted third parties is an inherent attribute of PoS + Account model blockchains.
Moreover, adopting PoS consensus and the Account model triggers a series of cascading effects, ultimately leading these networks to form client-server topologies, moving closer to a fully centralized Web2 system rather than the ideal form of Web3.
Therefore, the only way to achieve true decentralization, censorship resistance, permissionless access, and self-custody of data assets (the goals of Web3) is to construct a peer-to-peer network based on a PoW+UTXO system (Web5). To understand this, we must analyze the core differences between PoS+Account blockchains and PoW+UTXO blockchains.
PoS+Account vs. PoW+UTXO
There are significant differences between PoS+Account systems and PoW+UTXO systems, and the secondary effects of their implementation are even more profound. Some seemingly subtle design choices may ultimately lead to vastly different forms of the chain.
We will validate the following hypothesis through several dimensions: chains that choose PoS or the Account model can never form a flat, truly peer-to-peer network.
Differences in statefulness
The first dimension supporting our hypothesis is the difference in statefulness assumptions between PoW+UTXO and PoS+Account chains.
For example, in UTXO-based systems, transactions are stateful, containing inputs and outputs. Each transaction explicitly consumes which UTXOs and generates new UTXOs, carrying all the state information needed to update the ledger. However, the on-chain environment is essentially stateless, as transactions can only affect the UTXOs they directly reference and cannot modify other parts of the ledger.
In contrast, in Account-based systems, transactions are stateless — containing only operational instructions (i.e., the actions or method calls expected to be executed) without needing to explicitly state the current state of the relevant accounts. The on-chain environment is stateful, allowing any transaction to modify the state of any account or contract. For example, a smart contract can interact with multiple accounts and change various state variables, leading to a highly interconnected system state.
In UTXO-based systems, user-created transactions explicitly specify the changes to the ledger; in Account-based systems, users rely on blockchain nodes to compute these changes.
In terms of consensus mechanisms, PoS consensus is stateful. Validating consensus requires access to on-chain data, particularly the current set of validators, their staking status, and random numbers. As the validator set changes dynamically, nodes must continuously track these states to validate new blocks.
In contrast, PoW consensus is essentially stateless: nodes only need to verify the proof of work in the block header to confirm the blockchain's validity, without additional on-chain state information.
The differences in these stateful assumptions mean that in a PoS+Account model, users must track global state to validate transactions, which requires running a full node.
However, the statefulness of the PoS+Account model significantly increases the storage and computational burden on full nodes. Nodes must independently execute all smart contracts to validate transactions, track changes in the validator set and their stakes, and handle authentication, proposals, and other data related to block validation. This leads to nodes needing to store and compute additional state information.
Data comparisons can visually illustrate the differences: the minimum requirements to run an Ethereum full node are a 2TB SSD, 16GB of RAM, and a seventh-generation or higher Intel processor; while running a Bitcoin full node only requires a low-end CPU, 2GB of RAM, and at least 15GB of available disk space. Moreover, Ethereum is facing a state explosion issue — its state is growing at 3.5 times the rate of Bitcoin, and cannot prune old state data, meaning there is no limit to state growth.
Due to the high hardware requirements for running full nodes on PoS+Account chains, there are very few actual operators. At the same time, the complexity of implementing the PoS+Account model and its security trade-offs result in a near absence of truly trustless light clients (refer to our in-depth article on this issue), forcing users to rely on centralized RPC services like Alchemy and Infura to access the blockchain.
In short, the PoS consensus and Account model make running full nodes difficult and render trust-minimized light clients unfeasible, leading users to have no choice but to rely on a few centralized RPCs and APIs to read and update state. This dependency breeds a client-server network topology, akin to the centralized model of Web2.
Thus, 'Web3' has recreated the Web2 issues it intended to solve: a lack of security, privacy, and censorship resistance. The RPC providers serving most Web3 users can censor transactions, as evidenced in the OFAC sanctions against TornadoCash.
These RPC providers also collect user data, including blockchain addresses and IP addresses. Moreover, since most users' traffic relies on these providers, if their centralized infrastructure encounters problems or goes offline, the entire user base (especially 'mass adopters') will be unable to access the blockchain, as seen in the 2018 Infura service disruption caused by CryptoKitties congestion.
In contrast, the PoW+UTXO system makes full nodes, SPV, and light clients easy to implement, allowing users to verify transactions without relying on trusted third parties. This promotes a more direct (and therefore more private) way of participating in the blockchain and a peer-to-peer network topology, achieving true decentralization.
Deterministic differences
Blockchains are essentially replicated deterministic state machines, making them universally recognized 'single sources of truth.'
The deterministic performance modes of PoW+UTXO and PoS+Account systems lead to differences in network topology, particularly reflected in validator elections, block times, and finality.
In PoS systems, validator elections are deterministic — validators take turns producing blocks according to preset rules. While this approach improves efficiency, it introduces vulnerabilities: validator IP addresses are public, allowing attackers to target specific validators with DDoS attacks, leading to network paralysis during their block production periods. Moreover, validators must know and cooperate with each other, as the network's health depends on this.
More importantly, deterministic block production puts validators in a privileged position, allowing them to extract economic rent from users. Specialized companies leverage resources and income to expand staking scales, continually earning block rewards and MEV, creating a positive feedback loop of 'the rich get richer.' The centralization of MEV supply chains and block builders exacerbates this trend.
In contrast, on PoW chains, validator elections are non-deterministic. Before a block is mined, no one knows who will generate the next block, which promotes equality among nodes in the peer-to-peer network. The consensus set is also non-deterministic; miners can freely join or leave the network, and any node can generate blocks; no miner is indispensable for the continuation of the chain. This is not possible in PoS, as the consensus set is deterministic, requiring the presence of certain validators to advance the chain's development.
As a result, PoW networks are more robust, with no node having a privileged position, and no node guaranteed an opportunity to exploit users for profit.
Possible future of Web5
A network composed of PoW+UTXO chains still seems like a pipe dream to many. Web3 has become an industrial machine, continuously producing new systems to address the issues arising from Ethereum derivatives. While some thinkers are beginning to understand the intricacies of PoW+UTXO, Web3 remains entirely built on the PoS+Account model.
Although Jack Dorsey did not lead the TBD project to the promised land, ironically, the future of Web5 is indeed TBD (to be determined).
Even Satoshi imagined a world composed of a massive blockchain and industrialized nodes/miners. Today's Web3 universe includes these; however, we are always thinking about the chain: which RPC does MetaMask point to? To which chain are assets bridged? Does the address format comply with standards? And so on.
In this industry where every technical detail seems to have tokens and teams, viewing the blockchain as a truly committed layer outside the system is a whole new territory. Fortunately, our vision for Web5 is already in progress.
Despite initial controversy (perhaps 'rgbp2p' is a better name), RGB++ is leading the Web5 wave, integrating Bitcoin and Nervos CKB in a trustless manner, without cross-chain bridges or dubious security mechanisms. Support for Dogecoin is in development, with future plans to connect with PoW+UTXO chains like Kaspa and Ergo.
The progress of channel networks is even more exciting. For most of 2024, the Fiber Network project is dedicated to implementing Lightning Network compatibility on CKB and is rapidly approaching mainnet. Although the Lightning Network is known for its high difficulty, CKB provides a new computational foundation to test various improvements to the Lightning Network without requiring consensus changes.
The Polycrypt team has been working on state channel networks for nearly seven years and recently released integrated functionality for cross-account and UTXO models, supporting eight chains, including Ethereum, Polkadot, Dfinity, Cardano, Cosmos, Stellar, Fabric, and CKB.
Amid the BitVM and Bitcoin revival wave, with the maturation of Taproot Assets, an off-chain renaissance is also unfolding. Teams like Ark and Mercury are exploring new possibilities for Bitcoin-native off-chain computation.
Conclusion
The only path to achieving true decentralization, censorship resistance, permissionless access, and self-custody of data assets as claimed by Web3 is to build a network with a peer-to-peer topology. So far, only PoW+UTXO systems are possible.
In the PoW universe, the success rate of blockchains is extremely low; they are more like fleeting meteors than convenience stores. These miracles or beautiful accidents are only used for consensus and final settlement, and everyone can participate in running them. Validation remains low-cost and accessible, allowing users to run full nodes or light clients on ordinary hardware.
Through vertical scaling (adding payment channels or state channels on the base layer) to improve throughput. State is managed through UTXO, allowing users to verify only transactions relevant to them without relying on centralized intermediaries.
The path of innovation is always filled with uncertainty, and the future of Web5 is no exception. But as Nervos and Nostr client developer Retric stated in this article: 'This is a vibrant community driven by values such as freedom, decentralization, and open communication. It is not just technology — it is a movement.'
After a decade of observing Web3, we have witnessed few surprises and are now ready to break free of these shackles. We are ready to embrace uncertainty. We hope you are too.
Special thanks to Jan, Aijan, and Neon for reviewing this article and providing valuable feedback.
Follow the author on X @radicalizedpleb, @matt_bitcoin
