Rodrigo Coelho: 0:00

An agent thinks of a new business model, goes out, acquires the funds, raises money for it, writes the code for it, starts making money, and just does it all autonomously. Crazy. I'm great, how are you?

Joeri Billast: 0:27

I'm great too. Excited to have you on this podcast, guys. If you don't know Rodrigo. Rodrigo Coelho. He is the CEO of Edge & Node, the code developer of The Graph. And actually, they they recently conveyed Ampersand, which is a dashboard to manage AI agent payments. I'm really excited for this conversation, Rodrigo, as you know we can talk about subjects like AI and Web3 and where it comes together. But before we dive in there, tell me a bit about your background and start where you want to start.

Rodrigo Coelho: 1:00

Well, thanks for having me on. I'm super excited to have this conversation. So, yeah, my background is engineering. Background I started out coding and literally writing code myself. This was in the early days of the internet, kind of like Netscape Internet 3.0 era. But then have been a serial entrepreneur over the years, founded several companies. And then most recently, I was the first hire after the original three founders of the graph and have been there since 2018. So we're now looking at eight years, and took over as CEO of Edge and Node, which is the main core developer team just over a year ago. We've been in the blockchain data indexing space. So the graph, for those that aren't aware, is a method of extracting data from the blockchain, organizing it, and making it available for people that are building applications. And what many people don't understand is the world today is sort of run by copies of the blockchain on databases somewhere in a cloud server. So that's kind of the big secret, maybe, that most people don't know. So when we talk about all of this happening on the blockchain and stable coins coming on-chain and assets coming on-chain, when we're talking about on-chain, we're talking about writing to the blockchain. If we kind of use the analogy of the blockchain as a book, and each page in the book is sort of like a block, and you turn the page, and that's a new block, right? And on each page is a set of transactions that are happening, like in an old school accounting ledger. You're just writing those to the blockchain. But imagine that the book was closed and you didn't have a way to read it. Blockchains don't ship with the read software. When you buy a database to use in the enterprises, you have a way to query it. Blockchains didn't ship with that. Ethereum didn't come with that in the early days, Bitcoin didn't. So reading the transactions out of the blockchain is what was missing. And that's what that was what the graph filled in in 2018.

Joeri Billast: 3:18

So if I understand correctly, the blockchain was always right first, but not read-friendly. So I guess it's fair to say the graph became the search engine layer for Web3 in early days.

Rodrigo Coelho: 3:34

The founders were starting to try to develop applications on top of Ethereum at the time and ran into this problem immediately is okay, if I'm going to build an application that say it's a social media a social app, and I want to run through my list of friends, well, how do I get my list of friends out of the blockchain? So what you'd have to do back then before the graph existed is on your machine locally, you'd have to run through the entire history of the chain to view every time in one of those pages where you added a friend. So if you think you'd have to start from the beginning of the blockchain and then run through and say, oh, Rodrigo added Yodi as a friend there 800 blocks ago. And then Rodrigo added Michelle as another friend. So you have to keep track of that and then say, okay, here's Rodrigo and here's my list of friends to have it in like a table format. Right. So that became a complicated problem because every user would then have to have like a copy of the chain like running on their machine. And you can imagine the complexity. So the graph then said, let's make a copy of the blockchain, organize it into tables and columns and formats, and then the application's developers will tell us the shape and format of the data they need for their application, and then we'll just be the backend for their application, and they can just use GraphQL to query it as an easy way to sort of abstract that whole problem away. And so it solved that problem and it really fulfilled a big gap, and it became immediately kind of the ubiquitous standard. And if you remember in 2021, for those that were around DeFi Summer with protocols like Uniswap, Ave, Compound, et cetera, that was really the graph behind the scenes that was powering all these trading platforms, these DeFi platforms.

Joeri Billast: 5:32

At that moment during DeFi summer, did you realize you are no longer building tooling but critical financial infrastructure?

Rodrigo Coelho: 5:41

So not many people know that. And then the graph was like holding it up. Like that was really true at that time. We had the millions of queries and massive infrastructure holding this whole thing up. So we have years of experience running high volume, high throughput data, a blockchain data infrastructure. And through that experience, we've run trillions of queries over the years, we started to see common patterns and problems that emerged. And we spent about a year and a half in research and development and came up with a new product we called AMP. And we believe that AMP will be the new standard for how people extract and index data from the blockchain. And in a nutshell, it we switched to make it SQL first. So structured query language is kind of the language that everyone speaks to query data. And so we make it, we made it SQL first, we made it local first, so it could be run on a laptop. It's a singular kind of interface. Another problem we have today, there's many different chains. You've got Avalanche, you've got base, you've got Solana. And each of these chains has a different method of how you read the data out of it. Well, AMP gives you a single interface running from your laptop that you can use SQL. We'll have connectors for all the different chains. And now, given that we've got this whole move of traditional finance into using blockchain Rails, AMP is really perfectly positioned for that because we have these large enterprises, banks, et cetera. They have specific needs. Everybody has a different kind of internal systems that they use. AMP can be run on-premise. So they can literally run an instance within their secure environment in their secure cloud. They can bring data in, private data as well, public and private. They can combine it with off-chain data sources. We have the ability to verify data. So offering verifiable proofs that the data sitting in their cloud hasn't been tampered with, that it matches what's on the chain. So if there's like an auditor or a compliance person, you can literally say, okay, the data that we have here matches. And if you're going to take financial decisions upon that data, you want to ensure that that data is accurate and represents the canonical truth on chain. So yeah, we just launched it in November, and we've been in full business development mode and having many, many conversations with large financial institutions to bring it into their data workflows.

Joeri Billast: 8:35

And so what was the signal that told you the market had changed? Was it uh technical pain, enterprise demand, or something bigger happening in finance?

Rodrigo Coelho: 8:48

Aaron Powell You know, obviously everyone was talking about AI, and uh my colleague, our chief marketing officer, her fiance, is very high up at NVIDIA. In a casual conversation, we were talking about AI and Web3 and crypto, and he was explaining how the people at NVIDIA, people high up in AI, just didn't understand crypto. Oh, they weren't into it, they didn't see the need for it. However, they did say, okay, with agents, that seems like the perfect use case for crypto and web three. That was the one thing that they could say, okay, this makes sense because you have low transaction fees and costs, you have the observability and transparency of transactions and which AI agents need rather than going into a black box. And so we had this conversation, and he asked the question so who in Web3 is working on this sort of payment transaction rails? And at the time, this was like January, February of 25, and nothing had emerged yet. And I'm I was thinking, you know, Visa or Stripe, like these people have to be working on this. I mean, we're talking about a multi-trillion dollar movement of money, so they must be working on this. And so we started a bit of research, and then a couple weeks later, X402 came out, which for those that aren't familiar, is some people at Coinbase put out an open standard, an open protocol to allow for micropayments over HTTP, you know, totally free, open source, and it just became a method for these payments to happen in a very easy, streamlined way. And so we saw this and we saw how it reflected our original thoughts a couple years earlier. And so we we jumped in headfirst and s and found how we could contribute and and participate to it. And so we authored this deferred scheme. So that could be like the same batching technique that we developed would be utilized in this new protocol. So we're we're proud of that work. And then we got so we worked closely with Coinbase, with Google and their A2A, working collaboratively and just getting involved in the space. And then what we saw emerging was that you know, if you have all these agents interacting, there wasn't a way for humans to then view and monitor and control what these agents are doing. And so that's the idea of ampersand emerged from that. We launched it in November. But it allows for it's like an observability and control plane for humans to then monitor and control how much agents are spending, what they can and can't do, their wallet controls, and the same batching, monitoring reputation and viewing that soon identity and building on trust as well.

Joeri Billast: 11:51

Yeah, obviously, Rodrigo, you're very passionate about what you've been doing, BitWeb 3, and indeed it makes sense. AI agents, AI payments. I'm curious to hear your opinion about something that I've heard that Jeremy Alert said. Actually, I met him at Paris Boxing Week two years ago. But now I think recently he mentioned that within three or five years we could see billions of AI agents conducting economic activity. And then I'm curious which part of that statement do you agree with? And do you think that maybe people underestimate the complexity?

Rodrigo Coelho: 12:30

Yeah. So I also, as a little sidebar, I met Jeremy Ailaire just recently in Davos and had a great conversation with him. There's an interesting backstory there, is my first startup that I mentioned earlier that was in the late 90s. We were an application development shop building solutions for Fortune 500 companies. And we utilized this software called Cold Fusion. It was sort of like early version of Python or a way to build dynamic content on pages. And that was produced by Alair Software, which was an early company that he eventually sold to Macro Media, which then sold to Adobe, I think. So we had a nice chat about that. He really liked hearing that I was an OG fan of Cold Fusion. So we had a nice laugh about that. But back to his question, I actually think he may be undershooting that number. We saw recently with Claude Bod and Moltbook and this like explosion of agents. Like literally in a matter of 48 hours, there's like, I don't know what the number was, 130,000 agents or 50,000 agents. So we saw how quickly these things just spawn. And you have agents spawning agents spawning. So the numbers can grow exponential. I mean, it could be a trillion agents, for all I know. We were really entering into uncharted waters. And I think this recent, you know, it just happened last weekend that all this activity happened. It really caught me by surprise because I was like, oh wow. You know, we're not even talking about humans interacting with agents. This is how agents interacting with agents and then spawning new things that we haven't even thought of yet, and they have access to money. And then so we saw like they spawned rent a human, for example, right? Like only an agent would think of that and then pay. And that's actually a valid idea, right? Like maybe see there needs to be a marketplace where agents who can't solve a problem will pay a human to like come in if they don't have like, say, the answer from somewhere, like they have a human expert come in and like actually answer the question to help them solve a problem. That's totally plausible. This is an idea that hadn't even been conceived of up until like last weekend. We're seeing just a remarkable amount of an explosion of new business models, new ideas. I saw that we may even have business, like people said, okay, there's going to be the next thing to happen is like a billion-dollar business that has one founder, like one person that builds a business that goes to a billion. Yeah. But now it's like we may have a not a non-human like say a business that an agent spawns that goes to a billion, right? Imagine that. Like that's totally plausible now. An agent thinks of a new business model, goes out, acquires the funds, raises money for it, writes the code for it, starts making money, and just does it all autonomously. I mean, this is wild times. And this is just happening like now. It's crazy.

Joeri Billast: 15:51

I know. It's so crazy. How fast it goes, you know. Um you mentioned that agents they can do a lot. They can transact with stable coins. You talked about rental human. What do you think is the biggest risk if you don't have clear rules, limits, visibility build-in from day one?

Rodrigo Coelho: 16:11

Yeah, this is sort of what we're all about. We think in order for this to flourish, these safeguards and rails have to be put in place prior. Because if you think about it from an enterprise standpoint, at the end of the day, the responsibility for an agent going rogue is going to back up to a human at the end of the day, likely the chief legal officer or somebody at some big corporation or the enterprise itself. And so there's risk involved there. So that's really what we are all about in what we're building with Ampersend is we need to have these ability for humans to then monitor and control what's going on. So we believe that in order for this to really take off and to become a part of everyday life and for enterprises to feel safe in adopting this new technology, is that there need to be sort of a stack of standards and technologies that allow for monitoring, observation, and control. Because at the end of the day, there's going to be some human, it's likely the chief legal officer or the entity itself is going to be responsible for what these agents do on behalf of that enterprise. There needs to be an ability to put guardrails around that. Because if this agent goes rogue and does something that could likely damage the reputation or trust of the enterprise, really, without those safeguards in place, I don't think enterprises are going to widely adopt this. And so that's what we've been focused and working on with Amperson. There are certain initiatives out there. We joined this group called the Advanced AI Society. They have this whole narrative that they're calling proof of control, the proof of control initiative. And their intent is to bring in different stakeholders, enterprises, builders, developers to work on this stack of technologies that need to be in place, be it identity systems, reputation systems, trusted compute systems, and observability. So we're a part of it as well. So once these technologies are in place, and then say there's like a procurement person at an enterprise, they can then have an evaluation checklist to say, okay, all these pieces are in place, and then we can then implement this agentic system for us. So again, it's still early days. It's wild experimentation going on, and enterprises are obviously going to remain cautious before fully adopting, but it's inevitable that they will. We're just putting the foundational pieces in place, we believe, in order for them to feel comfortable and bring these systems into production.

Joeri Billast: 19:09

It's exciting for you building all of that, uh, having these conversations in Davos. I'm sure you you had a lot of insights from there. And I'm wondering now, Rodrigo, what are you now the most excited about, about everything that you heard there, everything that is now going on in the world?

Rodrigo Coelho: 19:27

Yeah, I am really excited about everything that's happening in the space. I feel like we're perfectly positioned with the two products we're working on, with AMP in the enterprise blockchain data space, with Ampersen, with Agentic Commerce and Control there. And really, to me, those are the big two stories of 2026. I heard in Davos that the enterprises like financial institutions moving on chain is the story of 2026. So we're soon to have the Clarity Act, and we'll have the legislation in place for these banks and financial institutions to fully be clear about how they can come online. And we're seeing that it's moving rapidly. And you can go down the list of the largest financial institutions. They all have some kind of internal blockchain proof of concept going on, experimentation. They have dedicated teams working on this transition. And we're at the beginning of this major wave where this will happen behind the scenes, but you know, we will make financial transactions and they'll to us appear as though they're just getting completed. But behind the scenes, they're happening on new rails, which are way more efficient, way more transparent, way more auditable, way more low-cost, and just better. You could think of it like a new rail system, like building high-speed rail for these antiquated systems. For example, this is a fun fact is the CH system in the US. So when you do like an electronic check, this goes through a system that was developed in like the 1960s, still to this day, and it uses like FTP of files once a day. It's a literal text file that gets FTP dual location. We're talking old school stuff. So yeah, this stuff, because it's so entrenched and they're moving trillions and trillions of dollars, it's just too big to fail and they cannot mess it up, right? So they're like, if it ain't broke, don't fix it. But now we've got such a major movement and everybody coming on board and now agreeing, okay, blockchain rails are the way to go. It just facilitates so many new things. You know, tokenization is happening with like stocks, bonds, real estate, even art or other things. So you can represent ownership. As a token which can be traded on blockchain Rails, and then it's all transparent and viewable. Also, what I heard in Davos and what I'm hearing at different conferences, at least in the enterprise side, is privacy. So private data. You know, a financial institution wants to trade with someone, but they don't want to reveal, for example, the entire history of that wallet or their full account details. So they just want to send a transaction and be able to just do that, you know, over blockchain rails without revealing everything. So that's a huge initiative and being talked about a lot. Trust is also a major topic of conversation. The panel I did in Davos was talking about how do we trust the agents. So many people are working on various solutions in terms of like identity, so ways to prove that who developed the agent goes back to human, or I mentioned earlier reputation. There are new standards like ERC 8004, which is like a reputation system, which is open source and decentralized. So agents can then, as they complete work and do transactions, have a whole transaction history, and there's kind of like a feedback mechanism and a rating mechanism. So as agents are going out and finding other agents that can complete tasks for them, they will can utilize this sort of open standard for looking at reputation. And then with Ampersen, we're looking at these open standards as well and how we can fit into this. So one example would be like a skated payment infrastructure. So imagine you have an agent that's going out and buying. You could set rules in there. So like you can't spend more than 50 cents on this task, or you can only spend X, say a dollar per day. Like you're putting limits on it. But even per transaction, you could say only interact with agents that have a reputation score above, say nine out of 10. Only interact with agents with a reputation score of X. And there's going to be other validation methods for agents like staking, for example, or other like ZK proofs. So you could add in kind of criteria. So out of the millions of agents out there, find an agent that has a reputation score that's greater than this, that has staked this amount, and has done this amount. So it allows for a lot of granular control, programmability on different transactions. And so these are the things that we're putting in place now, which are very exciting, which are going to allow for a lot of new and exciting experimentation and use cases that we haven't even thought of yet, frankly.

Joeri Billast: 24:55

Amazing. Thank you so much for sharing all that. By the way, lots of things you just said, I can confirm because in the podcast conversations I've been having, you know, people are talking about these new reels. Also, the word trust, if it's a podcast episode around AI, web three, or vote trust is in every podcast episode topic. Well, Rodrigo, the time went really fast talking to you. If people are now curious about things that you're building and they want to know more about you and about your company, where would you like me to send them?

Rodrigo Coelho: 25:30

Yeah, so you can find us on the web at edgenode.com, edge e a n d n o d e dot com. And we have a separate site for ampersend. So it's amp-er-se-n-d.ai. And we're also on Twitter at edge and node and at ampersand underscore ai. And I can be found at Rod Ventures on Twitter as well. So happy to chat with anybody who's interested.

Joeri Billast: 26:05

I will go and find you there on Twitter. It was really a pleasure to have you here, Rodrigo, as my listeners know. There are always show notes. Every account you mentioned, every website, and also some golden nuggets you mentioned will be found in there. Rodrigo, it was really a pleasure to have you on the show.

Rodrigo Coelho: 26:24

Yeah, you too. Thanks for having me. This was really great. I can't wait to hear it.

Joeri Billast: 26:29

Guys, what an amazing episode full of golden nuggets. I am sure that you know people around you in Web3 AI or maybe with a big interest in these new technologies. Well, be sure to share this episode with them. If you're not yet following the show, this is a really good moment to do this. If you haven't given me a review yet, if you give me these 5 stars, it really goes a long way. And of course, I would love to see you back next time. Take care.