Dialogue with Aperture Finance CEO: Intent-driven, AI-assisted, how to create a new paradigm driven by intent?

Questioner: Mia, ChainCatcher
Responder: Julian Zhu, Co-founder and CEO of Aperture Finance

As the trading volume on the Aperture Finance intent platform skyrocketed to $2.3 billion, the number of users also surpassed 240,000, gradually becoming the focus of the intent track in the crypto space.

On April 13, Aperture Finance announced that it helped users rebalance $66 million in liquidity positions in a single day, with over $38 million on the Base network, setting new historical records. Currently, Aperture Finance has assisted users in rebalancing $1.8 billion in liquidity positions.
Aperture Finance stated that it has developed a new type of chatbot based on underlying intent infrastructure, combining AI and intent. This chatbot allows users to "declare their goals" in natural language and achieve more efficient execution and pricing through a solver network.

Recently, ChainCatcher interviewed Julian Zhu, CEO and co-founder of Aperture Finance. Julian Zhu previously served as a senior product manager at Amazon Kindle, Netflix, and AWS. He holds a master's degree in simultaneous interpretation from the Monterey Institute of International Studies and an EMBA from the Haas School of Business at the University of California, Berkeley, where he also serves as the VP of the Blockchain Association for the Asia-Pacific region. In addition to being an entrepreneur, Julian Zhu is a member of the North American Writers Association and the American Translators Association (ATA), with translations including over thirty works such as "Revival" (by Stephen King) and the "Diary of a Wimpy Kid" series.

Perhaps due to his experience with words, Julian Zhu always uses examples during the interview to introduce the novel concept of the intent track in a way that is as simple and understandable as possible.

Julian Zhu describes the intent framework as Doraemon: the intent architecture can be seen as a flexible and open system that automatically selects and executes corresponding solutions based on user needs. This concept is similar to the plot in "Doraemon," where Nobita has various needs, and Doraemon always pulls out the right tool from his pocket to satisfy him. In this metaphor, Doraemon's pocket represents the intent architecture, while the various tools correspond to the solvers within the architecture.

From Translator to Web3 Entrepreneur

ChainCatcher: Your background is quite different from many entrepreneurs. Why did you join the ranks of Web3 entrepreneurs?

Julian Zhu: My experience can be described as a drifting background, transitioning from a liberal arts student to entrepreneurship, which precisely reflects the characteristic of Web3 in lowering the barriers to entrepreneurship. I initially studied simultaneous interpretation and interned at a UN-affiliated agency after graduation. In 2012, I joined Amazon to handle the localization of the Kindle in the Chinese market. I worked at Amazon for three years, during which I had relatively free time and did some translation work on the side, such as translating the best-selling "Diary of a Wimpy Kid" series and Stephen King's horror novel "Revival." Due to pressure from outsourced translation and machine translation, I began learning programming and moved towards product management. At that time, the mainstream was still computer-assisted translation (CAT), where machine translation was only a reference, and human translation was primary, allowing translators and computers to coexist peacefully. Now, ChatGPT is genuinely taking away translators' jobs.

Later, when Netflix was preparing to enter China, I switched to Netflix as the language manager for Greater China, responsible for localization work, and participated in licensing some films to iQIYI, such as "House of Cards" and "The Badlands." In 2018, as Netflix abandoned the mainland China market, I chose to leave.

ChainCatcher: So you transitioned from Web2 to Web3?

Julian Zhu: Strictly speaking, that time was more like a regression from Web2 back to Web1. At the end of 2018, we faced challenges such as GPU failures and rising electricity prices, writing scripts to automate troubleshooting and dynamically adjust mining strategies, and we also dabbled in ASIC miners and FPGAs. However, the sharp drop in cryptocurrency prices and the high electricity costs in California made it difficult for the mining operation to continue. Later, by chance, I teamed up with another friend to set up a Chia mining farm, but it ultimately also could not sustain itself.

Subsequently, I joined AWS as a product manager, responsible for localizing training videos from the US to China, but the slow progress made me lose motivation. According to the speed of entrepreneurship in Web3 projects, two and a half years is enough to take two projects from financing to exit. We only left after securing funding from a US VC.

After leaving, my partner and I discovered a stable profit algorithm strategy and raised funds to establish a private equity fund during the pandemic. However, as the fund size grew, compliance risks became apparent, leading us to dissolve the fund and explore decentralized finance (DeFi), establishing Aperture Finance with a focus on volatility hedging strategies. Although our original intention was to allow more people to share in the cryptocurrency dividends, we still had to operate according to the rules of the crypto circle due to regulatory constraints and user thresholds.

ChainCatcher: After Aperture Finance was established, fundraising must have been a significant challenge, right?

Julian Zhu: For a startup, securing funding is a watershed moment. All three of us partners are elites from big companies, with backgrounds as senior programmers at Google and master's degrees in computer science from Stanford and Cornell, but the road to financing was still full of challenges. It was difficult to make the decision to go full-time into entrepreneurship; in 2021, we were stable in big companies but felt lost about the opportunity cost of entrepreneurship. Our DeFi investment strategy was profitable, but the risks of entrepreneurship remained. So, we made a commitment: if we could raise $2 million before Thanksgiving, we would go full-time into entrepreneurship. Fortunately, we successfully raised the funds, which led us step by step to where Aperture is today.

Entering the Intent Track

ChainCatcher: From starting your business in 2021 to now, you have fully experienced the turmoil of 2022. What was that experience like?

Julian Zhu: It can be described as a rollercoaster experience. We were the first in the entire network to propose a volatility hedging strategy and successfully implemented it on the Terra chain, with our product's TVL exceeding $120 million within three weeks of launch. At that time, Terra was thriving, and our product was highly sought after, with Sequoia Capital and Tiger Global wanting to invest $10 million at a $150 million valuation. However, just as we were about to complete the legal process, the Terra chain collapsed, and we lost that opportunity. Professor Wang Defeng said that if you don't believe in fate at 40, your insight is too poor; perhaps this was destined.

As if that weren't enough, the collapses of 3AC and FTX added two tragic strokes to 2022, and the market completely turned from bull to bear. At the same time, we saw many peers in operational strategies, such as Friktion.fi and RoboVault, announcing bankruptcy, creating a bleak atmosphere.

ChainCatcher: Did you consider shutting down the company at that time?

Julian Zhu: When our company was in trouble, we did not formally discuss shutting it down because we had not exhausted our investment funds and still wanted to try harder. After understanding the industry, we realized we could close old projects and open new ones to return value to investors. In June 2022, our DeFi product was taken offline due to its inapplicability, and we pivoted to building infrastructure to provide automated services for strategies to navigate through bull and bear markets.

ChainCatcher: Is this the prototype of the intent architecture?

Julian Zhu: At that time, we hadn't coined the term "intent" yet; we created a concept called "Composable Automation," which can be loosely translated as "buildable automation." The core idea is the same as intent, which is to build and achieve automation based on user needs.

Whether users want to build staking strategies or liquidity mining strategies, our platform can quickly provide ready-made modules. Initially, this concept seemed avant-garde and unique. It wasn't until June 2023 that "intent architecture" gained widespread recognition in the market. We rapidly adopted it due to Paradigm's promotion and became the first intent network with a product ready a year ahead of others.

Integration of AI and Intent Architecture

ChainCatcher: Speaking of which, what exactly does intent architecture mean?

Julian Zhu: Intent architecture can be seen as a flexible and open system that automatically selects and executes corresponding solutions based on user needs.

This concept is similar to the plot in "Doraemon," where Nobita has various needs, and Doraemon always pulls out the right tool from his pocket to satisfy him. In this metaphor, Doraemon's pocket represents the intent architecture, while the various tools correspond to the solvers within the architecture.

When users express their needs through platform interactions, the intent architecture analyzes these needs and automatically selects the most suitable one or more solvers from the available options to execute. These solvers are like Doraemon's tools, each with specific functions that can solve particular problems. Users do not need to understand the specific workings or details of the solvers; they just need to express their needs and let the architecture handle the rest.

The openness of the intent architecture is one of its key features. This means that new solvers can continuously be added to the architecture to expand its functionality. As the number of solvers increases, the range of user needs that the architecture can meet will also grow, thus achieving a transition from reality to "magic."

In practical applications, intent architecture may involve complex algorithms and data processing technologies to ensure accurate understanding of user needs and selection of the best solutions. Additionally, the design of the architecture must consider factors such as usability, security, and scalability.

ChainCatcher: Are there any well-known intent cases that the public is familiar with?

Julian Zhu: Strictly speaking, intent is not a brand new concept; it's just that the public understands it from a new perspective. For example, when you swap tokens on Uniswap, many times you might see the exchange rate within a reasonable range and confirm the transaction directly. You trust that it has provided you with the optimal exchange method, and it indeed does so. If you click on the details, you can see that your transaction might be split into several smaller transactions. This is an optimization it makes based on the user's default intent (to find the optimal path for the best exchange rate).

This is intent at the token exchange level, and many intent projects have further extended in this direction. However, this remains at the application level, while the greatest potential of intent architecture lies at the infrastructure level, not just in exchanges but in meeting various user intents across the board. For the same vision, Anoma aims to create a new public chain to build a new world and invite all parties to rebuild. We, on the other hand, are relatively conservative, focusing on efficiency, and aim to reintegrate various protocols within the existing ecosystem through intent architecture.

ChainCatcher: Aperture focuses on "AI-empowered intent architecture." What does the specific application look like?

Julian Zhu: AI profoundly influences our product in two dimensions.

First, at the user access level, taking token swapping as an example, basically no one would conduct a full analysis for a single swap, considering various options, including cross-chain, to seek the optimal cost and time solution. Yet this is precisely what intent architecture excels at; users only need to express, "I want to swap 1,000 USDT for ETH at the best rate," and we can compare all options through solvers.

The most direct way to express intent is through natural language, communicating via text chat. We have developed a dedicated IntentsGPT based on ChatGPT, which relies on powerful AI.

The other area where AI is utilized is in solver selection. When our intent network has multiple solvers capable of handling the same intent, we use AI to select the optimal solution. Just like when Nobita wants to travel, and Doraemon has both the Anywhere Door and the Bamboo Copter, the choice can be left to AI.

ChainCatcher: What will the ultimate user experience of Aperture's intent architecture be like?

Julian Zhu: For general users, when claiming Airdrops, they often encounter issues like missing out, phishing risks, or website congestion. On our platform, users only need to connect their wallets and tell IntentsGPT, "Please check and claim all pending Airdrops on my address, ensuring optimal fee operations." Our system will automatically complete all steps for them.

For strategy users, they can simply tell IntentsGPT, "Please analyze and open a liquidity position for the trading pair with the highest daily yield." Our system will provide one-stop service from calculation to execution. Through natural language, users can easily complete various tasks on our platform and enjoy rates superior to manual operations.

Currently, IntentsGPT and the intent architecture have officially launched, and we are not far from realizing this convenient vision.

ChainCatcher: You previously mentioned that Aperture is currently the only project in the intent track with a launched product. What is the current user situation and future development plan?

Julian Zhu: Aperture's current intent tools have launched on nine EVM-compatible chains, and as of now, the total intent trading volume has reached $2 billion. We currently have 200,000 unique users, with daily active users exceeding 8,000, which can be said to have gained recognition from DeFi users.

Each intent network is like an independent e-commerce platform, each with its own characteristics, but fundamentally similar; the competition in e-commerce ultimately comes down to the brands and quality of merchants on the platform. Once the platform is established, the competition is about who can expand faster and attract better brands.

The same goes for intent architecture; it is inherently an open architecture, and adding one more solver means adding one more function. Therefore, our early launch gave us a first-mover advantage, but to maintain this advantage, we need to continuously add new solvers. Relying solely on our internal team for development is too slow, so we collaborate with projects like Propeller Heads and Enso Finance to have them develop solvers for us, while we also open it up to the community, allowing community developers to create solvers themselves.

ChainCatcher: This e-commerce example is quite enlightening, but if brands enter too quickly, quality control may not be well managed. Do you have concerns about this aspect when adding solvers?

Julian Zhu: The complexity and interconnectivity of intent architecture will inevitably impact security to some extent.

We categorize solvers into several types: one type is official solvers, which are developed by Aperture itself, and we have the most confidence in their security.

The second type is second-party solvers, which are provided by partners. We will audit these and provide whitelist permissions.

The third type is third-party solvers, which are provided by community developers. Our tokens will play a core role here. Third parties deploying solvers will need to stake a certain amount of Aperture tokens. For example, if Zhang San wants to deploy a trading-type solver and stakes $5,000 worth of Aperture tokens, their order limit will be $50,000. Here, we apply a different coefficient based on different scenarios, determined by potential maximum losses. If an order fails or malicious behavior is detected, the staked amount will be deducted as a penalty and compensated to users. Conversely, if a solver successfully completes a transaction, it can earn transaction-related fees, thus incentivizing solvers to optimize their algorithms.

ChainCatcher: Besides requiring tokens to be staked for deploying solvers, what other functions do Aperture tokens have?

Julian Zhu: The token accountability mechanism primarily revolves around the deployment and staking of solvers. Additionally, our tokens have multiple uses: they can offset fees, unlock premium features, and participate in governance voting.

The intent architecture aims to provide users with convenience and a high-quality experience, for which we will charge a small fee based on transaction amounts. If users pay with tokens or stake a certain amount, they can enjoy fee discounts. For special features that consume more resources, we have set token usage thresholds.

In terms of development decisions, we not only follow our internal development roadmap but also place great importance on community participation. Community members can propose and vote on how tokens should incentivize developers, such as integrating projects or developing new features, thereby jointly promoting the rapid development of the platform.