The evolution and challenges of the concept of "intention" from Anoma and TG Bot products

Author: Xuanrui | 0xDragon888, AC Capital Research

TL;DR

• Understanding Intent: Top-down Anoma vs. bottom-up TG Bot
• In-depth Intent: AI is the New UI: Complex Commands vs. Simple Intent
• Navigational Intent: Evolution of the concept of Intent, interaction paradigms, current status, trends
• Guiding Intent: Challenges and prospects of intelligent interactions in Web3

In June, Paradigm's official website published an article titled "Intent-Based Architectures and Their Risks," which brought the concept of Intent to the public for the first time. The development of protocols and infrastructure related to Intent has been rapid, becoming an unavoidable topic in the crypto world at the ETHCC conference in July.

Intent is not a new concept. As Mindao, the founder of DForce, stated: the trend in the crypto space has always been to abstract and automate operations to the extreme. From aggregators, CEX, contract wallets, to cross-chain DeFi, all these efforts are in line with this trend, along with Chainlink and the recent automated middleware, as well as the recent Telegram bot.

However, Intent brings new changes. In the past, the focus was on product interaction, while in the AGI era, the focus shifts to human-computer interaction. AI/LLM (Large Language Models) hold more potential to enhance Crypto interactions.

Top-down Anoma vs. Bottom-up Unibot

In the past six months, protocols, projects, and infrastructures related to Intent have been booming. Here, we do not intend to explain concepts in detail but rather present two excellent examples for a glimpse into the broader picture.

Among the many projects related to Intent, the most prominent is the Anoma Foundation, which completed its third round of financing of $25 million at the end of May 2023. In the context of severe homogenization in Layer 1, Anoma Foundation, with its intent-centric architecture Anoma (full-stack Dapp architecture) and Namada (privacy Layer 1), has raised a total of $57.8 million across three rounds, ranking seventh in Layer 1/Layer 2 financing amounts for projects that have not yet issued tokens.

Reflecting on Anoma's public speech at EthCC, founder Adrian Brink stated that the magic of Anoma lies in the fact that all interactions begin with Intent. Users express their Intent, which is processed through the Black-box Architecture (also known as Magic Happens), enabling core transaction processes based on Intent interaction:

• User Interaction: Users can send transparent, private, or protected Intent to Anoma's Black-box Architecture (intent gossip Layer).

• Intent Collection and Matching: The solver in Anoma's Black-box Architecture collects Intent and performs balancing state transitions to match counterparties.

• Transaction Processing: Matched transactions are submitted to the encrypted mempool, and validators send the blocks packaged by proposers to the execution layer, completing execution and validating effectiveness, ultimately updating the state root.

Source

Behind the new form of interaction, the Anoma team has identified the underlying pain points of blockchain protocols. They reviewed that starting from the first generation of Bitcoin's scriptable settlement, the architectural limitations of the second generation of Ethereum's programmable settlement have led to the current application protocols having at least one Web2 component, which cannot achieve counterparty discovery and resolution. They stated that the intent-centric Anoma architecture represents the third generation of Dapp architecture in the evolution of blockchain protocols, allowing users to define the desired outcome state, achieving efficient and customizable privacy transactions at the intent level. The intent-centric Dapp architecture is ushering in a new era.

Source

With high financing and a new technological architecture, Anoma is currently built by a multidisciplinary team of 37 members from the Heliax development team. Although the Heliax development team has made slow progress over the past two years, a complete innovative ecosystem has formed:

• Anoma: An intent-centric universal Dapp architecture that can be modularly deployed to L1, L1.5, and L2, embedding modular components into various EVM frameworks for invocation;
• Namada: The privacy settlement Layer 1 of the Anoma ecosystem, using a PoS mechanism for asset-agnostic inter-chain privacy, which will be compatible with Ethereum and IBC chains in the future;
• AnomaVM: Simplifying DApp application development on Anoma, using Juvix (an intent-centric programming language) and VampIR (a circuit programming language);
• Typhon: An improvement of the Tendermint consensus based on Cosmos;
• Taiga: Providing a unified execution environment for composable privacy, defining three types of Intent: transparent Intent, shielded Intent, and privacy Intent.

Source

Although many institutions such as CMCC Global, Electric Capital, and Delphi Digital are enthusiastically promoting Anoma and expressing that the intent-centric architecture has unlimited potential for the future, the top-down Anoma has only laid a solid technical foundation without visible exciting applications; on the other hand, TG Bot has opened up the imaginative possibilities of Intent in segmented scenarios using on-chain automated trading tools.

Unibot is a Telegram-based trading bot (TG Bot) that allows users to conduct automated DEX trading through Telegram, providing features such as sniper buying, follow trading, DEX limit orders, privacy, and MEV resistance. Unibot replaces the cumbersome interactions of Uniswap with Intent, offering a convenient DeFi interaction experience. Unibot has also spawned a batch of TG Bot clones, and the total market cap of the TG Bot sector is now close to $200 million.

Top-down Anoma represents innovation, while bottom-up TG Bot represents transformation; the top-down Anoma starts from an intent-based underlying architecture, but the team's progress is slow; while TG Bot brings users a new on-chain interaction interface, it is not intelligent and carries many security risks. Both paths ultimately aim for the same goal—simplifying user interaction experiences and introducing new programmable, customizable user interaction interfaces—User Intent Layer, allowing users to bypass complex on-chain interactions and define transaction states based on user intent.

AI is the New UI: Command Interaction vs. Intent Interaction

The two use cases above provide a basic understanding of Intent. Next, it is time to explore the underlying changes that AI brings to the crypto world—the change in user interaction interfaces.

The AI paradigm is introducing a third user interface paradigm in computer history, shifting towards a new interaction mechanism where users tell computers what they want, rather than how to do it.

• Paradigm One: Batch Processing.

• Paradigm Two: Command-based Interaction Design.

• Paradigm Three: Intent-based Result Specification.

Historically, every revolution in interaction patterns has led to entirely new business models. Generative AI based on LLM has brought disruptive changes to human-computer interaction, directly transforming the original human-computer interaction with various software UIs into interaction with a ChatGPT chat window. This is an unprecedented interaction experience, and LLM + crypto brings a new proposition of intent-based interaction to the crypto world. LLM will also make blockchain interactions potentially smarter by discovering and describing user intents.

Evolution of the Concept of Intent, Interaction Paradigms, Current Status, Trends

After deeply understanding the AI intelligence trend behind Intent, we further explore the evolution of the concept of Intent. Intent is not a new concept; it has existed since the Web2 era. With the rise of search engines like Google, users can input intents, and search engines provide relevant search results to satisfy those intents.

With the rise of e-commerce platforms like Amazon and eBay, the intent-based architecture underwent significant changes. Users can now express their intent to purchase products, while the platform handles logistics. In early 2017, Gartner published a report titled "Innovation Insight: Intent-Based Networking Systems," formally introducing the concept of Intent-based Networking. The key to intent-based networking is presenting users with an interface where they only need to express what they want, while the platform handles the business details.

Intent-based networking is the development trend of network automation and intelligence in the Web2 era. Andrew Lerner, Vice President of Research at Gartner, pointed out in 2017 that intent-based networking would be the next milestone in the field of networking.

Source

In the evolution from Web2 to Web3, the concept of Intent in Web3 is not entirely clear, but there are some common understandings:

• Definition provided by Paradigm: Intent is a set of declarative constraints that allows users to delegate transaction creation to a specialized network of third-party participants while retaining full control over the process.

• Definition provided by the intent-based underlying Dapp architecture Anoma: Intent is a message sent by the user to express custom preferences. Intent is a constraint that defines what the user allows the system to do, rather than a specific execution path.

• Definition provided by the intent-based trading infrastructure Propellerheads: Intent is the part of the transaction that requires assistance from others, simplifying complex transactions by outsourcing the difficult parts to third parties.

In real use cases of intent-based transactions in Web3, users create intents off-chain, outsourcing them to Solvers (decoders) off-chain, encapsulating the complexity of blockchain interactions for the decoders while retaining complete control over the process on-chain, thus lowering the threshold for on-chain interactions.

Source

Intent also brings the LLM architecture concept— intent-centric—an architecture centered around user intent. Currently, protocols and infrastructures related to intent-centric are embedded in the interaction layer between users and Crypto, leveraging LLM to provide a better on-chain interaction experience.

LLM + Crypto brings a new human-computer interaction paradigm, directly converting user intents into smart contract calls. Users only need to express intents like using Apple's Siri, while robots/Ai Agents/third-party decoders help complete complex on-chain operations. In the future, the intelligentization of Web3 interactions based on intent has the potential to greatly reduce the complexity of user on-chain interactions.

AI is the New UI. More and more Web3 projects are exploring the magic of Intent. Whether it is intent-based Dapps or intent-based underlying architectures, they are releasing the potential of Intent from different angles. We attempt to organize this rapidly developing field and categorize it into four major categories from the perspective of interaction, tracking the latest status and trends:

Infrastructure: (Intent architecture layer centered around Intent)

• DappOS: The first intent-centric Web3 operating protocol, building a middle layer between users and public chains, cross-chain bridges, and other crypto infrastructures, allowing Web3 users to interact with products seamlessly and without awareness. Recently, DappOS released its V2 version;

• Flash 2.0 SUAVE: SUAVE is a standalone plug-and-play modular MEV full-chain sorting layer. Preferences (the core concept of SUAVE) are similar to the Intent concept, providing ordinary users with a way to customize transactions, ultimately helping users achieve optimal execution. Flashbots plans to launch SUAVE Centauri in Q4 2023;

• Anoma: An intent-centric universal Dapp architecture that can be modularly deployed to L1, L1.5, and L2, embedding modular components into various EVM frameworks for invocation. The latest update in the Anoma ecosystem is that the privacy settlement Layer 1—Namada—is about to release the v0.22.0 testnet;

• Cow Protocol: CoW Protocol builds an underlying settlement network for traders and decoders, and in July released the Cow Hooks, a new intent system for custom DeFi operations. The latest data from CoW Protocol in August processed a total trading volume of $1.05 billion and generated $5.3 million in profit.

Supporting Facilities: (Supporting infrastructures related to Intent and account abstraction wallets)

• ERC 4337: User Operations in ERC 4337 represent user intents, including the tokens for paying gas, payment choices, and login methods, replacing the current transaction memory pool, thus achieving account abstraction and improving wallet interaction experiences.

• Supporting infrastructures related to account abstraction: Bundler infrastructures (StackUp, Blocknative, Alchemy Rundler, AA-Bundler, Infinitism Bundler), wallet SDKs (Safe, ZeroDev, Biconomy);

• Supporting infrastructures related to Intent: Juvix (an intent-based programming language), Seaport (an open-source NFT trading protocol), Gelato (a Web3 automated execution layer providing automation services for smart contracts), okcontract (a low-level intent automation tool), Delegatable (a smart contract authorization proxy tool), Hyper Oracle (off-chain automation services), symmio (an intent-based on-chain derivatives trading framework).

Enterprise Applications: (Integrating Dapp's intent infrastructure, API, modular intent layers, specific domain decoders)

• Trading Decoder API: PropelleHeads;

• Intent API:** EnsoFinance;

• Intent Trading Infrastructure: Bob the Solver, Brink;

• Composable Intent Tools: Essential.

Consumer Applications: (Intelligent interfaces for user interaction such as Wallets, Dex, Web3 Ai Agents, intelligent search engines)

• Wallets: Safe, Bitconomy, Argent, Ambire, Sequence, Versa, A3SProtocol;

• Dex: CowSwap, Paraswap, Uniswap, 1inch, BananaHq, basedmarkets;

• Intelligent Frontends: Unibot, Dora, Web3 Analytics, kaito.ai.

The intent-based architecture, supporting facilities, and enterprise applications (listed here without elaboration due to space constraints) attempt to address the underlying challenges of implementing intent. However, infrastructure cannot be built overnight; the trend of consumer applications for Intent is what truly excites. The evolution of interaction entry points such as wallets and Dex shows the potential for intelligent interactions.

Intelligent Wallets: ERC-4337 introduces a new (UserOps) user intent layer, where users express intents, and then bundlers convert these intents into executable signed transactions. In recent months, the number of ERC4337 users has significantly increased, with ZeroDev, Biconomy, and Safe modular smart accounts being the most motivated to promote AA + intent. AA + intent has the potential to usher in the arrival of intelligent wallets (SCW).

Source

DEX Architecture Transformation: Intent-based trading aims to improve capital efficiency and user interaction experience. CoW Hooks connects complex actions such as trading, bridging, staking, and depositing. UniswapX has already allowed users to sign intents for off-chain matching and on-chain settlement. Similar intent-based DEXs like BananaHq, Brink, and basedmarkets are gradually increasing, and a new intent-based RFQ narrative (SYMMIO) is unfolding. As more DEXs and aggregators transition to intent-based architectures, the DEX landscape is undergoing significant changes.

Source

Intelligent Interaction Entry Points: Similar to how TG Bot transforms the Web3 front-end interaction interface, interaction entry points focus on understanding user intents and systematically converting them into automated, actionable tasks. Not only TG Bot-like trading robots, but also Web3 Ai Agents and intelligent search engines are making Web3 interactions smarter.

Source

Challenges and Prospects of Intelligent Interactions in Web3

In the past, the interaction experience of Web3 products indeed hindered widespread adoption, as the on-chain wealth creation effect masked many product issues; now, liquidity and user attention are dispersed, and on-chain liquidity is scarce.

At the ETHCC conference in July, developers extensively discussed the future of intent-centric interactions and how to help users complete DAPP interactions more intelligently in Web3. However, the application of intent-centric approaches still faces many challenges:

• Lack of intent-related programming languages (like Juvix)

• Lack of suitable intent architectures (like Anoma)

• Lack of specific domain decoders (like Bob The Solver)

• Intent implementation is centralized and closed (like CowSwap)

• Lack of front-end intent recognition decoders (like Unibot)

• Lack of composable intent implementations (like Brink)

• Intent security risks (as mentioned in the Paradigm article)

There are challenges but also hope. With the emergence of intent-centric protocols, Web3 has the potential to evolve into a friendlier intelligent interaction interface, allowing users to interact seamlessly with mainstream applications like Uniswap, Blur, and AAVE. Interaction could become as simple as hailing a ride with Uber, which represents a true opportunity for Web3 Dapps to surpass Web2 in user experience.

More importantly, combining the visions of Anoma, Flashbots SUAVE, and DappOS for the future of intent-centric interactions, on-chain interactions in Web3 will also become more intelligent, customizable, and fair. At the same time, placing the power back in the hands of users with intent at the center allows ordinary users to empower third-party robots/decoders to complete customized transactions (fees, slippage, privacy, cross-chain, MEV, or other intents). In the future, users will no longer be passive participants on-chain or victims of MEV Searchers and Validators, but the true captains of on-chain activities.

Looking ahead to the future of intelligent interactions in Web3, whether it is account abstraction, chain abstraction, or intent, Web3 is finally paying attention to user experience. The dawn of intelligent interactions is about to illuminate the dark forest.

Acknowledgments

Special thanks to Web3 Analytics, Crypto V, Haotian, Chen Jian Jason, Luke, Grace Deng, SixSix.eth, POOR DAO, #017, armonio.eth, Kiwibig.eth for their valuable discussions, opinions, and feedback on this article.