#Motivation

I think it’s important for everyone to deeply understand their motivations behind doing a startup. Below are some of my personal motivations, then some of our startup’s motivations:

Sometimes I like to joke to my friends that I was destined for this, that it’s in my blood.

My great grandfather used to play at underground casinos under communist China. Local officers would target him, forcing bribes or “taxes” until the family came to expect them every new year. He used to hide in the forest whenever he won big, waiting out so long he’d get stomach ulcers.

My grandfather was a sharp farmer, one who quickly picked up strong intuition around agriculture who renounced gambling. My dad was a biologist by training, but became drawn to the allure of stock-picking in his 30s. I watched a meteoric rise and fall by him when I was a kid.

I never fully caught the bug until more recently (or so I tell myself) but I remember annoying my parents for a brokerage account at 12. Poker, flips, card games, mahjong, etc were all fun hobbies.

I picked up some other cool experiences when I was younger, like winning international science fairs and running small businesses.

College admissions was actually an enjoyable game for me (though it seems most didn’t share this sentiment!) as it was just a series of compounding trades for probabilistic outcomes. Win something, snowball it into another one, keep iterating, find the feedback loop. While there’s variance in the process and you can’t guarantee a school, you can very aggressively bolster your odds to get into an ivy league. Everybody knows it's a game that can be played, whether you want to admit it or not.

#Every startup is a trading firm

So how did I end up here? Starting a trading firm?

First, I think that every startup is fundamentally a trading firm. Trade attention, expertise, risk, time, or some commodity for other, more favorable items (usually money, sometimes different delayed or alternate outcomes).

Similarly, every person is also a trader. Time with family, career growth, etc. At Harvard I watched lots of friends go into fields they disliked to gain career capital to pivot. That always felt somewhat convoluted to me, but a trade that makes sense for many.

My great grandmother was deeply curious, studying abroad in other countries near China when it was unheard of for women to go to college. She too, was trading (clearly, I thank many people to thank).

So why a monetary trading firm in particular?

Well, money is green and money is fast. It’s also one of the hardest games you can play. My co-founder comes from ex-competitive Fortnite and I can ttell that some of the same dopamine hits that wired that interest wired this. It is likely the most adversarial, well-informed game that one can play.

So, I started trading around when I was 18. Played around in options, equities, and all over the place (though never crypto). I used to wake up early to check my NVDA calls my freshman year at Harvard. More recently in prediction markets, with 90x returns.

A friend once jokingly asked why I was wasting my Harvard degree “bonding” prediction markets (where you buy an outcome at 99% odds and provide exit liquidity when the event has already happened, and hold until it exits at 100%).

I think being this type of person carries a similar archetype of pushing rules to their absolute max to extract every possible edge (fun story: I exploited my high school’s Covid attendance policy so aggressively my senior quote is about not being in class).

Plus, you know how many people give up their first born child for 1% of alpha they can snap up often hourly? It doesn’t scale (and I no longer do it), but man, sweet gig that paid for some sushi (I’m reminiscent of the times we’d virtually connect to radio stations to get faster latency for sports broadcasts or set up transcription bots to trade earnings on Kalshi).

I’ve never blown up any of my accounts, a fact that my dad is still surprised by. Every time I go home, he asks me if the accounts are still alive, and the answer is yes. He says that he would’ve blown up the same amount at the same age several times over.

And many do! I’m sure Leopold Aschenbrenner would have benefitted from having a few extra decades under his belt, but then he wouldn’t have as much situational awareness with such a meteoric rise.

#The long game

Fundamentally, however, this is a long game with long horizons and long returns, with room for many strong actors.

One of my great grandfathers lived to ~107+ (for real! verified!) and his wife and my great grandmother to ~104.

While long, life is still finite.

It is impossible to predict how the world will change 100 years from now (unless Bridgewater’s superforecaster turns out to completely work with no error ever), but capital will track every step of that movement. It is perhaps the smartest, most informed game that one can freely be a part of (unless this trend of companies staying private forever expands, but even then, there’s still opportunity everywhere).

I share the common belief that markets are efficient in the long run, but not over the short run (a similar way to how startups operate with PMF, actually).

So then there are three games a trading firm needs to play.

#Three principles

#An evolutionary trading firm

And of course, the obvious unspoken rule: stay smart and win (get alpha).

I personally like the pod shop structure (essentially a multistrategy hedge fund), as it optimizes for these things in combination (and pods can die or be birthed constantly). There’s a joke in biology that all crustaceans evolve into crabs (carcinization). Maybe the final boss of the fund structure is a pod shop.

What is obvious to everyone is that this is a fantastic breeding ground for AI to play in. And unlike in other industries, where the question of "Why won't OpenAI do this? Why won't Facebook do this?" often has reasonable answers, that isn't really true for trading. Pods are generally independent, under constant pressure, and trading is an eat-what-you-kill environment. If something works, another shop will do it. As one might expect, there are many younger entrants into the space, like Abundance and many stealth startups.

A lot of my friends have told me that when you ask AI about startup ideas, it has a lot of notions around 'legacy systems' and how so many companies are entrenched in the past and need to be 'disrupted.' This is not nearly as true for trading as it is for other industries.

AI is the worst it ever will be, and LLM-based traders aren’t even all that good, but AI/ML can already strengthen with all three points. AI agents can work superhuman hours, monitor a superhuman number of things, and constantly learn from its real or simulated mistakes. While learning compounds everywhere, it is especially strong and immediate in markets. Agents have specific advantages - they can process information without the bias or preconceived notions that often plague people. Trading is a game that most rewards being right when most others are wrong. And this is only what it can do now.

Of course, there are limitations - the robustness of long running agents, memory, learning, context, etc, but these are rapidly being structurally tackled. Context windows have gone from 8,000 to 8 million in the span of two years. Like Leopold predicted in Situational Awareness, you don't need to believe in magic for these problems to become alleviated, just trend lines within orders of magnitudes.

One can even argue that quantitative trading was an early (and still current!) indicator of some of these trends, as ML theory has been embedded there forever and quants functioned as expensive agents.

What is not obvious to people, is how to best utilize and amplify judgement and taste, as markets are still inherently human (which even Jane Street can have a hard time with, partly seen in July as they don’t just purely market make nowadays).

There will be many experiments to carry this out (many of which are already occurring), some from established firms, some from upstarts, and some from individuals amplifying their advantage (those folks should not be scared of AI taking their jobs but rather the mediocre asset managers should probably be sweating by now).

This data on judgement is fragmented, scarce, and the foundation of many people’s careers or even entire firms. Trading will change as intelligence gets cheapened, but some things will stay the same.

Our goal is to systematize the collection of it, and we are actively deploying capital and learning from every trade.

Most AI trading labs pay egregious amounts for their training data. Instead, ours is relatively free and pays us when we succeed! We think this is the ultimate alignment of incentives in one of the largest TAMs in the world.

We can simulate all decisions, reasoning traces, and have ground truth outcomes to learn and compound these benefits. Our model is based on an evolutionary system (quite biological from first principles), with open competition within each pod.

We believe this (alongside data-related benefits) is central to our alpha. While it’s much more nuanced than “Claude, make me money,” we can essentially spin up tons of traders at our disposal (with the help of our token credits and compute). This is an edge that scales, compounds, and is not held by any one person. We purposefully and constantly evolving to avoid decay. Harder to hit a moving target! (and in general, we think edges shifting away from latency to AI/ML we think is in everybody’s best interest)

#Why now

Another question I get is, “well gee, if you have an unlimited money printer, why not just print more money in secret? I’d never tell anyone about it”

Fair, but the answer is quite simple, which is that it’s still brutally hard even when everything goes right. We’re deeply cognizant there is always blood in the water, and this type of neotrading firm requires learning, compounding, compute, capital, talent, research, etc. much like any other technology company. Early alpha decays unless you are constantly at the forefront, and being constantly at the forefront requires some non-trivial level of initial capital for both experimentation and operation. We share these risks with investors who are more than happy to take on these risks with us.

We are already profitable, at the time which should be the worst in the history of our firm. We have a focus on two pods right now, with line of sight on how to scale to hundreds or thousands in the future. In college, I almost always found that interdisciplinary outcomes led to the best results and I suspect a similar thesis will be true with pods.

Our current active pods are semiconductors and biotech. Semiconductors because look at the times we’re living in, and biotech because look at the times we’re living in. We think these are two of the industries that matter most at present (and also require enormous investment and have enormous financial interest). To be honest, we picked these verticals, not our AI traders.

However, unlike a traditional pod shop, we don’t need to hire human analysts to start new pods or adjust for peoples’ bonuses or biases.

Many financial funds are also sophisticated technology companies at their core, and this is what we currently believe the next generation of that looks like.

Why now? Isn’t it obvious?