No Priors: Artificial Intelligence | Technology | Startups
00:29:29 · Sarah Guo, Vlad Tenev
[00:29:29] S1: Wow. [00:29:29] S0: And if you look at the tool calls, the tool calls have just been following this smooth exponential where the last number we announced was something like 30,000,000 tool calls per day. So Agentic Trading, clearly a product market fit. And we started by just having an MCP and a couple weeks ago at our hood summit event, we announced Robinhood Agents, which is basically taking it one step further and the harness and the entire UI is now inside a Robinhood account. So you can kind of think of it as if you're a Claude code or a Codex user, it's almost like a tool like that that's optimized for trading. So you can run loops and say, okay, every day execute a run of my strategy, rebalance my portfolio, look at opportunities. We have skills, so you can you know, do like a portfolio rebalance or all sorts of other things. We have agent apps, so we launched an app store where you can, for example, connect to a source of satellite data and if you wanna create a strategy that like counts cars and parking lots of retailers or something of that nature, we'll make that available. And the way that I think that'll evolve is we're gonna look at all the capabilities that the coding harnesses have and see how we can provide really nice, clean, trader native versions of that that just make it really really easy to be an active trader.
00:31:10 · Sarah Guo
[00:31:10] S1: Amazing. What do you think the impact is of this on either you know, participant mix or financial infrastructure? Just because many of these things that you describe, institutional traders have the ability to do this, but it's expensive and sophisticated, right? And so if that's democratized, what does that mean?
00:31:28 · Vlad Tenev
[00:31:28] S0: Well, that's really the north star of the product. It's can we put the capabilities of a hedge fund or a sophisticated algorithmic trading firm into the pockets of normal individuals. So in the same way that if you were a sophisticated grad student in computer science or something, could write a trading strategy, and if you went through even more effort, could deploy it, you could get rack space in the data center connected to the exchanges. Can we just sort of streamline all aspects of that so that in the same way that it's made it easier for anyone to be a software engineer, Robinhood agents and our MCP can make it easier for more people to be quant traders. And I think the outcome of that is there will just be more trading activity. I think at first it'll concentrated in a relatively small group of sophisticated people, but their trading activity will likely increase quite substantially. And then over time it'll just filter to more and more of the population. Yeah, I think coding is a good analogy. It's like there's lots of people that wanna, there's people that code and for them, you know, they were first to adopt the coding agents and their output multiplied. But then now there's people that just wouldn't have picked up, you know, VS Code or Cursor before, but someone showed them how to use Claude or Codex, and now they can build a website just really easily. So it'll lift, I think it'll lift the tide across the board.
00:33:07 · Sarah Guo, Vlad Tenev
[00:33:07] S1: I believe the desire to make software or automate things is much broader than I did perhaps before any of the coding and agentic coding, encoding AI and agentic coding happened, but wanting to make money off of a belief feels like a much more universal thing. [00:33:24] S0: Yeah, and I think that as people get, as it becomes easier, as you can really, a lot of people have ideas, right? And if you can take that idea and turn it into a strategy with as little friction as possible, If we can get it to approach the friction of just buying a single stock, then I think a lot more people will do it. [00:33:48] S1: There may not be an answer to this question, but is there an obvious way in which financial infrastructure or risk management needs to change in reaction to this? [00:33:56] S0: I don't think so. So I'll tell you the safeguards we have on day one. So right now you have to create a new separate account, which is segregated from your main Robinhood balance. So you can create an MCP account, you can create an Agentic account, you have to move money in there. And then for customers that choose to have trade approvals, by default you have to approve your trades, but then you can turn it off so that you could Dangerously scrapped permissions. Exactly. Well, actually, it's not as dangerous as that because it's not like it can go into your other accounts [00:34:34] S1: It's not my core code money
00:34:34 · Sarah Guo, Vlad Tenev
[00:34:34] S1: It's not my core code money [00:34:35] S0: for you. Yes. So it's sandboxed. It's sandboxed, maybe not quite air gapped, but pretty close. And we built a lot of the, we refined a lot of the infrastructure for the sandboxing and the harness in house at Robinhood. Yeah, so far I think that it's not just a huge percentage of the market. The reason I don't think, I don't see any clear massive overhauls necessary is that high frequency trading has been a thing for a really long time. I mean, markets went fully electronic about twenty years ago. And so everything from the exchanges to a lot of the brokers have sort of conformed to this new reality of most of the workflows being algorithmic trading that happens in data centers in New Jersey, rather than what a lot of people still think happens, which is have paper tickets Sure. Floor traders in a trading pit. So don't think, it's not clear to me that anything fundamental has to change with regulation, but as it gets bigger, which I don't doubt it will, we might run into special cases for which we need more regulatory clarity or... [00:35:53] S1: Yeah, I was more... From an infrastructure perspective, if you just use coding as a proxy, I did not, maybe we all should have, but I think it was not obvious that CICD was going to completely fall over, like, in the face of orders of magnitude growth in code generation. [00:36:14] S0: Oh, yeah. [00:36:15] S1: And so that that was my question, which is like, yes, of course, we've had HFT. Yeah. But if, like, everybody is a high frequency trader, that feels like new load.
00:36:24 · Vlad Tenev
[00:36:24] S0: Yeah. I mean, there's a couple of different layers in the stack. Right? And we we should expect our on chain stack is slightly different, that already has MCPs, but you have our on chain tools, which a lot of it is, you know, I talked about Robinhood Chain being the best chain for real world assets, also it's the best chain for AI agents. So we're seeing a lot of Agentic Trading there. On our CFi, centralized offerings, you know, we have a lot of end to end control over the entire system, right? We have our databases, we have our code that the strategies are running on, the sandboxes, we build a lot of those tools. We have to optimize all of that, but once it gets through our system, you get to places like the Algo, the Market Makers, which are all extremely sophisticated software shops. I mean you're talking Citadel, Jane Street, Jump Trading, these are like low latency, high frequency folks, so I'm not so worried about them scaling to meet the demand. And then you have the exchanges, and the exchanges, because of pressure from HFTs and algo trading, have really optimized their APIs to handle huge volumes. But that said, I think that we're still at a low percentage of the total trading as Agentic. So I think we'll have some warning. Mean, we'll see it increasing and I don't think that we're gonna be at the point imminently where Agentic mediated trades are you know, a double digit percentage of the activity, I think we'll have some time.