The Research Team We Never Hired
On Friday we launched our first ETF, ticker AVOS. It is the highest-conviction, long-only, active global equity portfolio we can build. More info at https://avosglobalequities.com/.
Last week I made three claims about the conflict in Iran, prefaced by the disclaimer that “it could age like milk.” Claim #1 and #2 have already curdled. Luckily for me (and our investors), claim #3 was that almost every trader should sit this one out, since “in a scenario where I highly doubt that even the central participants know what they are going to do, it’s a little hard to imagine anyone having a real edge.” In our commodity fund we hedged away as much of our exposure to the war as was practical, and we are up a bit for the month so far.
So this week I decided to take a breather from being wrong about the war, and instead share our firm’s evolution on AI. I share it to pull back the curtain on what we are doing as fiduciaries, but also (hopefully) to be helpful to you. With the exception of James Wang, I suspect no one reading this is using AI nearly as effectively as they could be, in their business or in their life.
Last summer, we raised some GP capital from a few existing clients and showed them a buildout plan. Multiple research hires at different seniority levels, and all expensive. We planned to scale our investment team the conventional way, because even though we were using AI at the time, it was no substitute for humanity.
We never made those hires.
By October, AI research capabilities were such that I went on stage at a conference and said that I was struggling to see why we would hire any junior investment people. For example, that month I wrote a piece where I went back through every Chinese five-year plan and examined whether they did what they said they were going to do. AI did the immense grunt work of parsing all of the plans, separating out testable promises, and then checking what happened subsequently. That would have been a junior person’s job for a couple of weeks, and instead it took me a few hours to do the entire study and write it up.
The improvements continued to compound. Having displaced the junior person, we then went after the gray hair.
Earlier this year we built our first AI-powered system that was designed to be a peer to me and our team in trading one of our most important markets. The engagement model was completely different from “ask a factual question, get back a factual answer.” Instead, we fed it raw data and what we believed were our edges in trading that market. It then went off and did 3 three things:
It pressure-tested our assumptions. It came back and told us where our edges were real and where we were kidding ourselves. That kind of honest audit is nearly impossible to get from an employee who’s trying not to offend their boss, and it’s hard to do on yourself when you’ve been trading a market for years and your biases are ossified.
It proposed new ideas. Ideas that we had either just missed or were blind to because of our aforementioned biases.
It freed us up to do other things. It scans markets and news flow on a regular schedule, checking what’s happening against our framework and flagging when something looks off. We were already doing this work, but now it gets done for us. We can instead focus on other things…like systemizing and launching our equity ETF.
I want to be clear about what I’m claiming and what I’m not. I don’t trust the system
enough to give it trading permissions, and I may never. I own the firm’s PnL and it doesn’t make sense to me that I would be disintermediated. Any output that it produces must be checked before using…just like human output needs to be checked. But it’s getting hard to argue that any single human on our team has a better handle on that market than the system does. We are now rolling this out across every market we trade.
None of the researchers we thought we needed are getting hired. Some of that cost savings we’re keeping as a buffer against the inevitable AI price hikes. Some accrues to our shareholders. And the compounding improvement in our research and positioning accrues to our clients — including, as of Friday, anyone who buys the AVOS ETF.