When Shift Happens4 min read
TrueNorth founders say AI trading agents could hit 5% of on-chain volume in 3-4 years
Alex Lee and Willy Chuang of TrueNorth explain why self-custodial AI agents, not centralized apps, will become the main way retail trades crypto.
AI summary of “The Future of Crypto Trading is Here (and AI Agents Are Leading It) | TrueNorth Founders | Drops E48”
Key takeaways
- Alex Lee says under 10% of people trust AI's judgment in financial decisions despite 30% of AI users expressing investment-related intents.
- Willy Chuang expects agentic trading to grow from roughly 0.01% of today's $125 trillion annual trading volume to about 5% within three to four years.
- TrueNorth reports 13,000 users, 55% 30-day retention, and users who talk to its AI are 2.3x more likely to enter a trade.
- Both founders insist AI agents provide beta capture and risk management, not alpha generation, with humans keeping "hands on the wheel."
- They argue agentic trading will concentrate on-chain rather than at traditional brokerages because of self-custody and verifiability.
TrueNorth's founders frame AI as an abstraction layer for finance
Alex Lee and Willy Chuang, founders of the AI trading platform TrueNorth, describe their product as an "agentic brokerage" designed to remove the parts of trading that retail investors don't understand while preserving the excitement of trading itself. Lee explains the core idea as taking "people's colossal," the stressful and hard-to-learn parts of finance, away from users, while leaving the dopamine of watching trades play out.
Chuang says that of the roughly 300 million people using major AI apps, around 30% express prompts or intents related to financial investment, but less than 10% trust AI's judgment enough to act on it. The founders see this gap between intent and trust as the opportunity their platform is built to close.
Why most retail traders lose money, according to the founders
Chuang and Lee point to the "90-90-90" pattern, where 90% of traders lose 90% of their money within 90 days, and attribute this largely to behavioral failures rather than lack of access: emotional decision-making, no trading journal, poor risk control, and lack of reflection on past mistakes. Lee argues AI can close this gap by acting across the full trade lifecycle: sourcing ideas, structuring trade plans against a user's risk preferences, and monitoring positions after execution.
Lee distinguishes between alpha and beta, saying flatly that "AI cannot generate alpha" in the sense of proprietary edge, but it can help capture market beta, for example deciding whether, when and how much to invest in exchange tokens during a bull market.
Human-in-the-loop design, compared to self-driving cars
The founders use an anecdote about driving a Tesla on "Max mode" FSD on the way to the interview to describe their philosophy: agentic trading should keep a human in the loop checking the AI's reasoning at each step, the same way they watched the road while letting the car drive aggressively. Over time, they expect trust to build the way it did "after 40 minutes in the car."
"Money moves at the speed of trust" — a guest (Willy Chuang)
They note that TrueNorth's platform tracks what they call "trader metadata traces," recording a user's reasoning, edits, and outcomes so the AI can flag repeated mistakes or reinforce good patterns, functioning as an automated, far more detailed trading journal than most retail traders keep themselves.
Why the founders believe agentic trading will move on-chain
Chuang argues traditional institutions will resist letting AI agents touch their assets under management (AUM), partly over an "agentic bank run" risk where idle funds get redeployed by agents without institutional consent, and partly over security concerns. On-chain finance, by contrast, is self-custodial and verifiable, which he sees as a structural advantage for agentic trading rails.
Chuang estimates that industry-wide spot and perpetual trading volume across centralized and decentralized venues runs about $125 trillion a year, with agentic trading currently representing "less than 1%, maybe 0.01%" of that. He says TrueNorth expects agentic trading to reach roughly 5% of on-chain trading activity within three to four years, driven by what he calls embedded trust on blockchains rather than convenience alone.
Personalized portfolios as a third generation of brokerage
Chuang lays out a three-era framework for personal finance platforms: first, terminal-based brokers like Charles Schwab and E-Trade; second, mobile, gamified, zero-fee apps like Robinhood; and third, what he calls the "agentic generation," which personalizes portfolios to an individual's goals and risk profile rather than offering generic ETFs or three risk tiers. He gives an example of a user wanting exposure to cybersecurity stocks who could get a custom, auto-rebalancing basket instead of a one-size-fits-all ETF.
On the question of "founder premium," whether AI can price in factors like belief in a founder's long-term capability (citing examples like Elon Musk and Near's Ilia), Lee argues this is "semi-doable already" since AI can mine social platforms like X to assess a founder's track record and signal, though he concedes it is "not quantified to an extent" yet.
Backgrounds, fundraising and competitive position
Lee and Chuang say they are both Taiwanese, educated in the West, and began their careers in China's internet boom, a period they describe as roughly 2015 to 2020, before meeting through a mutual contact at Carlyle right before the FTX crash. Chuang was running a crypto exchange and Lee was investing at Temasek at the time, which froze a potential investment; they reconnected in "early 24" and began building TrueNorth, working on the idea for about a year and a half, including writing a white paper during what they describe as the "AI meme token" era.
On competitive moat, the founders say they deliberately avoid competing at the "general harness level" against models like Claude or OpenAI's products, instead focusing on the gap between a user's intent and an executable, policy-compliant trade, which they view as the core opportunity in agentic finance.
What remains unresolved
The founders leave open exactly how much the 90-90-90 loss pattern can be improved, floating a hypothetical "80% of people losing money" without committing to a figure. They also acknowledge accountability questions around AI-assisted losses are unresolved, noting users remain responsible for trades even with AI guidance, and that the broader industry's infrastructure for agentic trading is "still missing" in places despite their optimism about its trajectory.
Written by AI from the video's transcript. It can compress, misattribute or miss context — the original video is the source. Not investment advice.








