$ALLOX
Fading QuicklySnapshot Window: 2026-07-09 08:20 UTC · ← Back to Crypto Overview
Social Momentum Summary
Total Engagement - Comments: 1, Retweets: 0, Likes: 0, Impressions: 4
Verbatim Community Citations & Social Evidence 1 source posts analyzed
Come and listen to my new podcast with @mrrongxin and @theamyzhao on our recent @initc3org paper, "Paper Agents, Paper Gains: An Empirical Analysis of DeFi Investment Agents." Always great to be part of @WPReadingClub ! [Spoken audio]: I think so, our paper started from this sort of observation that there was a lot of projects out in the broad universe that claimed to have, since 2024, claimed to have these sorts of AI agents trading cryptocurrencies and a bunch of tokens in some capacity. We saw this, for example, with Terminal Truths. We saw examples of this with the AI-16z, Eliza OS. We saw this with some virtuals agents and some of these frameworks out there that all claimed to be trading, basically a bunch of cryptocurrencies and earning supposedly autonomously earning money. And so we wanted to kind of like dig into that claim a little bit more and kind of ask the question of do we actually see what is the actual behavior of DeFi investment agents or all of these agents that we see out there interacting with DeFi protocols and supposedly trading autonomously. And so when we actually come up with, we came up with a couple of research questions here. I think the first one is one of scope. So we wanted to understand it's like, what is the broad landscape of all of these DeFi investment agents? Because as you mentioned, there's close to 2000 projects that we surveyed that was tagged as kind of like AI and crypto. But we wanted to understand what were the different approaches? What were these agents doing and things like that? The second question was kind of these frameworks and design. So how were these sorts of agents actually designed and how do they make decisions? The third was one of performance. Did they make those decisions well? And I think the final aspect is basically like what like a question of like, how does this space like evolve further? So I think the headline result is that it's very interesting. Even though we surveyed close to 2000, we found that the vast majority didn't weren't actually trading agents themselves. They were mostly infrastructure projects or some other. But there was a few agents that kind of spanned the scope of either, for example, giving investment advice or passively holding a bunch of volts or actually actively trading. And two of the most active frameworks were Eliza OS and virtuals. But again, when we dig into the data, as I'm sure Amy can also give us insights there, we realized that a lot of these projects weren't really making users money. And even if like the treasuries themselves were net positive, the users that gave their money to these agents, they were net like losing money as well. So there were a bunch of like interesting kind of like divides and dilemmas that we see throughout that. And finally, we come up the structure for where should agents go? And basically having three pillars there where first one is they needs to be like verifiable that it's actually an agent conducting these trades. They need to be profitable as well. And there needs to be alignment between whoever is giving them money and the actual agent's behavior. Yeah, I think that was a very nice overview of the main parts of the paper. But maybe to go back to also the very start of why we decided to write this and the motivation. One of the things that that we just saw in happening in industry. And this came up as a news bit and we were like, is this too DGEN for research, basically? Is this something that we should actually study? Is it worth writing a paper about? And I think the reason why we ended up writing the paper is because first off, the amount of money that is being handled and also the peak market caps were very significant, right? Like you're talking about 2.7 billion for only one of these projects. That's a lot of funds and actually like a non-insignificant amount of assets that are being managed by these projects.
