API3's Saša Milić: In Oracles, the Decentralization Bottleneck Is the Data Source

Ashton Addison: What is API3, in a sentence?▶ 00:24

Saša Milić: API3 is the DAO that maintains and manages a network of first-party oracles.

Ashton: What do oracles do, and why does DeFi depend on them?▶ 00:46

Saša: Oracles bring data onto the blockchain. A blockchain is an isolated network for security reasons. Its nodes don't talk to other networks, and that includes APIs, which is how most other computer systems communicate. The only data on the chain is essentially ownership data, meaning who owns which token. There's no notion of an outside world. That's not so much a limitation as a consequence of the high security requirements. Take DeFi. I don't know of any DeFi app that doesn't use price data. It's ironic, because the price of a crypto token isn't on the chain. It's mostly decided by external markets. So pretty much all of DeFi needs price feeds, and DeFi is one category of applications that oracles enable.

Ashton: What does it mean to have a first-party oracle rather than a third-party one?▶ 03:16

Saša: The difference is whether the person who owns the data is also the one who transmits it to the blockchain. A third-party oracle gets information from somewhere else, typically a web API. With crypto prices, it queries a price API and transmits the result to the blockchain. A first-party oracle is the data provider themselves. In our case that's an API provider transmitting its data directly. Our solution centers on the idea that first-party oracles are better in many contexts, mainly because you eliminate the middleman. That reduces security risks and often costs too, since many third-party solutions have multiple third parties querying the same API. We came up with the term in the API3 white paper.

Ashton: Is a first-party oracle more or less decentralized than a third-party one?▶ 08:19

Saša: Decentralization isn't the goal, it's a tool. Once it becomes a goal it gets nebulous, because there are always pockets of centralization somewhere. In this case it's not really about decentralization. Say a certain kind of information, like banking data from one particular bank, comes from only one provider. Adding ten third-party nodes doesn't make it more decentralized. The bottleneck in terms of decentralization is the data source itself. I don't like to pick on other oracle projects, but I've seen data feeds advertise 20 to 30 nodes when there aren't 20 to 30 quality crypto price APIs. Clearly multiple nodes are querying the same API, which creates an illusion of more independent data points than there really are. There are other kinds of data where third-party nodes make sense, but for querying a web API and bringing it onto the blockchain, this is our approach.

Ashton: Why structure API3 as a DAO, and what does that give you?▶ 05:13

Saša: The DAO space is very experimental, with a lot of growing pains. A DAO evolves in ways you might not predict, compared with a traditional corporation. One main reason we have one is that a product is arguably more decentralized when decision-making is decentralized. It's not enough for the protocol to be decentralized. It matters who decides what changes and what gets deployed. A huge benefit is that it enforces transparency. Any time money leaves the treasury, it has to go through a proposal and a vote. You can see who works for the DAO, what they're doing and what they're paid. That matters for data, because it's important to be explicit about where information comes from. With first-party oracles you know the source, and through the DAO you also see how the network is created and maintained.

Ashton: Is your focus still price feeds for DeFi?▶ 11:17

Saša: We're not currently live. Our oracle node is in auditing. Some of the API providers that have joined the network are crypto price APIs, others provide other types of price data, and we do get requests for price feeds. But the team agrees that DeFi isn't going away, and that blockchain will go beyond DeFi. To do that it needs new types of data. Price feeds are important and will be for a long time. The ideal architecture hasn't been figured out. It's a hard engineering problem and a hard statistical problem. But we also care about enabling the next generation of blockchain applications.

Ashton: What are the next milestones for the team?▶ 13:13

Saša: Over the next six months or so, the aim is to get Airnode live in production on Ethereum and other EVM-compatible chains, then move on depending on demand. We want to make sure it's integrated with all the API providers already signed on, and build out data feeds, again depending on demand. It's tricky. You want to serve current demand, but to enable new kinds of blockchain applications you have to foresee future demand. It's a balance between what people want now and what will enable interesting projects in the future.

Ashton: How can people learn more and get involved?▶ 14:33

Saša: Our website is easy to find, and a lot of the conversation happens in our Discord, which is linked from the website. We also put a lot of work into long-form content on Medium, with many articles about our solution and what it can enable. Very few people in crypto seem to have a long attention span, so maybe we shouldn't work so hard on 10-minute articles and should just throw out a tweet now and then. But if you like long-form content, please check out our media.