Outlier Venture: 8-Year Journey, L2 for AI Computing by Swan Chain

2024 is a big year for the web3 world, with DePIN and AI as key themes. Swan Chain stands out as a leading EVM Layer 2 solution for decentralized AI computing, grabbing attention with impressive testnet results: 13+ million addresses, 10+ million ETH transfers, and 103+ million transactions.

In a special interview with Outlier Ventures, Charles Cao, Swan Chain’s co-founder and CEO, shared the company’s eight-year journey from its early days to raising $10M in funding.

The Beginning of 8-Year Journey

How did you come to founding what was originally FilSwan and, of course, ultimately Swan Chain?

Before building Swan, I started off in engineering, working at big companies like IBM and Autodesk, and got really into cloud computing when I saw how flexible and scalable it was. I was fascinated by how we could make something similar ourselves, even though it seemed like a huge challenge.

Then, around 2017 and 2018, smart contracts came along, introducing a new way to do transactions without needing to trust the other person — it was all based on code. This opened up a world of possibilities for storage and computing for me.

Around this time, Filecoin finished their public sales and started working on IPFS, which coincided with my growing interest in blockchain technology. I saw an opportunity for a platform focused on data exchange, leveraging Filecoin for data storage, leading to the inception of what was initially FilSwan.

As we progressed, our focus expanded from merely storing data to considering computing solutions. The realization that data storage was only part of the equation led us to explore container-based computing, allowing for diverse applications including AI.

Last year, the advancements in AI and the arrival of stable diffusion models underscored the potential of generative AI. This was a significant moment for us, prompting the decision to develop a new market specifically for AI computing, leading to the creation of a product we called LagrangeDAO.

Images generated using Lagrange’s Stable Diffusion Space by users

LagrangeDAO was designed to leverage the computing power within the Filecoin ecosystem for AI purposes. As we built this out, it became clear that AI computing’s future would demand a more robust infrastructure, especially for high-speed transactions.

Working with Outlier Ventures and their experts, we recognized the need for our own blockchain to meet these challenges. This brought us to OP Stack, upgrading our platform to a faster, more efficient Layer 2 solution for AI computing.

As a founder navigating the space and effectively building out a stack of technologies, how much of the journey was opportunistic or kind of just pragmatically solving problems versus a big vision?

Honestly, this is something we were thinking about from the start. Going back three years, during the Filecoin mainnet launch, they showcased a Jupyter AI computing notebook. That image was ours, highlighting our early tests on AI computing with Filecoin’s infrastructure, signaling our broader vision for the future.

This vision led us to explore multi-chain capabilities, starting with Filecoin. Three years ago, Filecoin seemed like the only project with the global infrastructure to support our computing needs. That’s why it was our first choice. We then expanded to a multi-chain approach to leverage the best features across blockchains, emphasizing cross-chain functionality.

When we worked with Outlier Ventures, our initial plan was to build on the Cosmos SDK. However, its lack of EVM compatibility and the maturity of the EVM Solidity developer community led us to reconsider. Fortunately, finally, we observed the launch of Base chain and opBNB on OP technology, confirming our direction towards the OP stack for our Layer 2 solution.

Our journey is a blend of strategic vision and pragmatic problem-solving. We’ve always had clear goals but remained flexible, integrating new solutions as they became available.

Why Swan Chain for AI Computing?

Deciding to build a Layer 2, could you talk about how you approached this, how you managed the trade-offs, and why you believe computing requires its own Layer 2?

We’ve looked into many technologies, focusing first on storage because it’s essential for AI computing, which needs huge data sizes — terabytes or even petabytes. Filecoin was our best option, as it’s the only one that can handle such massive amounts of data.

However, when we delved into AI computing tasks, considering the cost implications on platforms like Amazon or Google, we realized we couldn’t use a blockchain that was expensive or slow. We needed one that’s affordable and quick.

At one point, we considered Solana for its high throughput and low cost. Despite its advantages, sticking to Solidity and EVM compatibility presented a significant language barrier.

We had to make a choice. Staying with Ethereum or moving to a Layer 2 solution were options, but what really suited us for AI computing was using an application-specific blockchain, or ‘app chain’. This way, we could focus solely on AI computing without the drawbacks of high traffic and gas fees found on more general platforms.

Looking ahead, we’re open to evolving technologies. Developments like Celestia’s modular chains or zero-knowledge proofs could further reduce costs and increase transactions. We have room for growth and improvement in the future.

If you’re focusing on Web3 dApps adhering to decentralized networks, the current market may not be worth your time due to its small size. Thus, your solution should target mainstream organizations and institutions. Why is Swan Chain preferable for typical businesses seeking AI computation?

That’s a very insightful question. People often misunderstand one aspect: they associate decentralization with small scale and centralization with large scale. However, this isn’t necessarily true, and here’s why: they’re like two fundamentally different business models.

The decentralized model can be likened to Uber’s approach and the centralized model can be likened to traditional taxi. With Uber, drivers use their own cars and only really need to worry about their immediate costs, like gas and time. They don’t count the cost of their car or insurance because they’re using what they already have. On the other hand, taxi companies have to buy cars, pay for insurance, and cover all sorts of other costs. So, their expenses are much higher right from the start.

It’s similar when we talk about decentralized computing or storage. Many providers use existing equipment or get it cheaply. They might use servers that are already there but not fully used, or they get special deals on equipment. This way, they can offer their services at a much lower cost. For example, by using space on servers that aren’t being fully used, costs can be up to 70% less than traditional, centralized services.

Also, decentralized systems can work more freely across different places. Unlike big companies that depend on traditional banking and are stuck in specific locations, decentralized services can adapt to local laws anywhere. This means they can work in many more places without being tied down.

Moreover, using blockchain technology, everything is open and clear. Contracts automatically handle payments once a job is done, so there’s no need for a middleman. This is a big change from older systems, where a lot of details can be hidden, and if something goes wrong, evidence might disappear because it’s all stored in one place.

So, the biggest difference between web3 and web2 isn’t just about the technology but about the whole idea behind it. Decentralized systems offer a more flexible, cost-effective way to do things, breaking free from traditional limitations.

The Road Ahead

Do you anticipate people starting to build future products by leveraging the Swan Network?

I believe they already have; it’s just not disclosed yet. This is because I’ve observed people deploying different tokens with names like ZK tokens, USDC, and USDT. These names aren’t coincidental; they’re intentional. I assume there’s a different DAG system waiting for various programs to be initiated. We’re in discussions with them, planning to launch an incentive program and a ‘data grant’ to encourage their involvement.

Thanks so much for joining us and sharing your founder journey. I’m really excited to see what happens with Mainnet. Do you have an ETA, or are you still in the process of finalizing it?

We’re eagerly working towards a Q2 release, which is quickly coming up on the calendar. Currently, we’re in the midst of conducting the final tests on several critical components to ensure a smooth launch.

There’s a likelihood we’ll implement one or two more patches, especially to adapt to recent upgrades like versions 4.8 and 4.4. These updates are key as they’re expected to significantly enhance the Mainnet by lowering gas fees and further improving performance.

View the full talk right away:

About Swan:

Swan is a Layer 2 AI Computing blockchain built with OP Stack technology. Backed by strategic partnerships with industry leaders such as Filecoin, IPFS, Chainlink, Polygon, Microsoft Azure, and Google Startups Cloud Program, Swan is well-positioned to drive the future of Web3 Storage and Computing. With cutting-edge solutions like Lagrange and Multi-Chain Storage, Swan is funded by Binance Labs, Chainlink Labs, SNZ Holding, Outlier Ventures, Chain Capital, and more.

Website: https://www.swanchain.io
Twitter: https://twitter.com/swan_chain
Telegram: https://t.me/swan_chain/
Discord: https://discord.gg/swanchain
LinkedIn: https://www.linkedin.com/company/swancloud