Key takeaways
- KUVI AI built an agentic finance operating system that uses LLMs to interpret user intent and match it to executable financial functions on blockchain.
- The company targets the $10-30 trillion investment and brokerage market rather than just the DeFi trading niche, positioning AI agents as replacements for traditional wealth management.
- LLMs became reliable enough about a year ago to validly understand and interpret human intent, enabling the core technology behind agentic finance systems.
- KUVI's vision extends beyond algorithmic trading to comprehensive wealth management, allowing users to manage and coordinate their wealth through natural language commands.
- Traditional financial intermediaries like BlackRock may become unnecessary if agentic frameworks can directly execute user intent without intermediaries or high management fees.
From DeFi trading tool to wealth manager
Ashton Addison: Let's start at a high level. What has Kuvi AI actually built?▶ 00:47
Dylan Dewdney: Essentially, a tool set that uses LLMs for intent analysis. That's the pluggable part of our solution. About a year ago, LLMs got good enough to understand a person's intent reliably, and that made a light bulb go off for us. If you can interpret the intent and match it with a function, such as "Kuvi, I want to swap BTC for Ethereum," you can go from the simplest operations to more and more complex ones as we integrate more data sources and triggers. In our context it's all financial, which is why we call ourselves an agentic finance operating system, or AFOS. I'm really old, so I was thinking of MS-DOS.
Ashton: Who are you ultimately building this for?▶ 03:49
Dylan: We start with traders, crypto and degenerate traders maybe, but we want to end up with something more comprehensive and mass-oriented. We have a big bet that people will use this the way they use ChatGPT, for training plans or even as a therapist, once it's connected to actually accomplishing things. We want people to use it to manage and coordinate their wealth. We're not going after the five or ten billion market of DeFi and maniac traders. We're going after the 10 to 30 trillion investment brokerage market. Do we actually need BlackRock in the context of super-genius agentic frameworks, with people behind them just wanting to get their intent across?
Ashton: Most people lose up to 2 or 3% a year to wealth managers. Did you always plan to go after that mainstream market?▶ 07:20
Dylan: Not at all. We didn't have as big a vision when we began. We were much narrower and thought we might make money on transaction fees from trading volume. Then, through market research and talking to advisers, investors and our community, we realized people aren't engaging with AI plus crypto to trade much better or run automations, at least for now, although I think that will be a big deal. They're more interested in getting information and analysis back. For example: "Kuvi, what are the top influencers I follow on X saying right now? Are they all mentioning the same thing in this 24-hour period?" X was one of our first API integrations. They may go somewhere else for the trade, but we don't care about that as much. We want people to get used to using Kuvi as the way they understand how to optimize their wealth management.
Do we actually need BlackRock in the context of super-genius agentic frameworks, with people behind them just wanting to get their intent across?
Why a general chatbot can't do the job
Ashton: Can the major LLMs like ChatGPT just implement this themselves? How do you compete?▶ 12:40
Dylan: The answer is no. About 30% of what we do is intent analysis. The other 70% is hard-coded functions. We're not in the training business. We don't have billions of dollars, though we'll eventually fine-tune our own model. If you ask ChatGPT to swap BTC for Ethereum, it's not going to happen. It'll return some links or tell you how to do it and hold your hand, but it won't execute. And as far as I can tell, they're not focused on that. We're optimizing our data sets and integrations specifically for wealth management, putting together Lego blocks of different trading algorithms and triggers. For us, those models are just a tool.
Ashton: Last week's tariff-driven crash liquidated a lot of traders. Can agentic finance help in a situation like that?▶ 15:16
Dylan: Absolutely. Say you have $10,000 on an exchange, $5,000 in BTC and $5,000 idle in stablecoins. You could tell Kuvi that if BTC drops a certain percentage in a 24-hour, six-hour or two-hour period, it should notify you with a message on your phone. Even better, you set up an automation that does something if it drops. There are stop-loss orders on centralized exchanges, but I don't think they're as intuitive as just talking to Kuvi. So yes, it could be prophylactic. You could protect yourself from massive drawdowns.
Strategies for everyone
Ashton: Today you can get an alert, but you still have to go and press buy yourself. How is this different?▶ 18:28
Dylan: Until now you'd have had to write out the bots and the code, and engineer all of it. Some people have engineered bots like that, but we're democratizing the ability of really anybody to deploy engineered strategy. That's a powerful opportunity over the next year or two. It's a bit of a zero-sum game, since not everybody can deploy a great strategy. But if we widen the net, people will stumble upon things. It's like enough monkeys writing on typewriters: someone will create a Shakespearean sonnet. I'm not comparing our future users to monkeys, but if enough people put together different algorithmic trading strategies, someone will strike on something interesting.
We're democratizing the ability of really anybody to deploy engineered strategy.
Token launch and the road to 2026
Ashton: The token launch is close. What can the community expect?▶ 22:12
Dylan: We're coming close to TGE. We've put it off once already. Why do a token at all? It's a valid question. If you hold Kuvi, you get a reduction in transaction fees when you transact through Kuvi. We're also thinking of a subscription model once the product is out of beta, where holding a certain amount of Kuvi means you don't have to pay. It also helps build a bit of a moat: if people hold Kuvi and transact with it, hopefully that leads them back to our platform. We're not doing a massive launch. We're not a massively capitalized project, so we can't spend millions on a big centralized exchange listing. We'll open on MEXC, which was a great ROI for us. The beta comes out soon after TGE.
Ashton: What do you hope to achieve in 2026?▶ 26:53
Dylan: The main thing is the product. We're getting friendly testers now, but soon we'll do an invite-only beta, probably free. I want people to tell us how great it is or how much it sucks, probably the latter at the start. We'll try not to look at the token chart at all. If we get more momentum and capital, we'll buy between eight and 10 H200s to fine-tune our model in house, for about 20 to 30% more efficiency. We'll also add a team member or two on the technical side and on the BD side. And we'll keep adding integrations. We've spoken to Messari and Blockworks about integrating their data sets.
This interview has been edited for length and clarity. Recorded November 2025.

