Pilot interviews

Wills Harris and the Polaron team on reading the news with AI

Wills Harris and teammates at Polaron, a team of physicists, on using language models to read energy sector news for The Energy Portfolio on Autopilot.

Autopilot7 min read

Wills Harris is a physicist at Polaron, a small team that has spent three or four years building trading strategies with large language models. Wills is pursuing a PhD in physics and working at a national lab. Two teammates, Adam and Eli, both physics PhD students, also sat down with us; each answer below is marked with the name of the person who gave it.

Open it in the Autopilot app. https://link.joinautopilot.com/eXKr?c=wills-harris-interview&af_sub1=intro

Polaron built a knowledge graph that ingests a large volume of news articles a day, assesses their likely impact on energy sector companies, and uses that to inform The Energy Portfolio on Autopilot. The team is careful to draw a line: the language models are a tool the team built and directs, not an investor making decisions on its own.

We talked about why a team of physicists ended up building for finance, what a knowledge graph actually is, and how the team thinks about trusting a system like this with your money.

This interview comes from the transcript of our conversation with Wills Harris and two Polaron teammates, Adam and Eli, in June 2026. It has been edited and condensed.

Who are you, and what is Polaron?

Wills: My name is Wills, and I've been working with some friends for maybe three or four years on building quantitative and qualitative strategies using large language models. We've been doing it sort of bare bones by ourselves, experimenting, trying a bunch of different things. We learned about Autopilot and we thought it was a cool opportunity for us. We've created a portfolio that analyzes the energy sector, gathers an enormous amount of news, processes the news in an intelligent way with various LLM prompts, and uses that to generate bets on different companies. And it's updated periodically.

What do you believe the edge of the Polaron team and your strategy is?

Wills: We analyze an insane amount of information, and the way that it's analyzed is unique to our algorithm. We have a graphical structure where different information is processed locally, and then that information is passed on to the next part in the graph. This improves the quality of the analysis, but also reduces the LLM cost. So we're able to process much more information at a lower cost.

What's the background of you and your team?

Wills: We all come from a physics background. My undergrad is in physics, and now I'm pursuing a PhD in physics and working at a national lab. When you're going through a physics program, you learn a lot about processing information in an intelligent way, and that aspect in particular has been really helpful in designing these algorithms.

How did a team of physicists end up building for finance?

Wills: It came somewhat organically. When ChatGPT was announced, we were all obsessed with it, and we got together and said, we want to create something that involves LLMs. Finance seemed like a natural area, partly because you can think of it as a zero sum game, so it's the ultimate test of how you design something. If you're using LLMs, the natural thing to do is to process textual information, to process news. At the time there were no ways of doing that, there were no standards, there weren't many packages designed for using LLMs. So it was sort of the Wild West, and it involved a lot of creative engineering.

Why have LLMs analyze news and sentiment rather than something more numerical, like fundamentals?

Wills: Our thesis is that basically all quantitative information is impossible to compete with. Quantitative information is very easy to process in a sense, there's an ideal way of processing it, and the people who do it the best are the people who will get the edge. Qualitative information is much more subjective, so it's natural to think that the edge will be distributed over many different investors, and it depends a lot on the way that you design it, which is again subjective.

What makes what you built different from someone just asking a chatbot what stocks to buy?

Wills: AI has been proliferating finance, and everyone's algorithm is a bit different, you can look at published papers and see how they're designed. Ours is designed in a different way that uses a knowledge graph.

Adam: We believe in the capabilities of LLMs, but we don't necessarily believe in the idea that an LLM is by itself a financial investor that you can just ask questions. We feel like we need to provide the context to the LLMs so that they can do the more nuanced analysis of a given company.

What is a knowledge graph, in plain terms?

Adam: The basis of the way that we store and process information is this knowledge graph, which is a way to represent complex information, for example the energy sector in the financial world, by representing companies and institutions and various factors that affect these companies through a graphical model. You can connect a company to its competitors, or its business relations, or its regulators. The knowledge graph represents all that information in a structured way, so that you can infer things about the companies through it. And when news events happen, you can encode them into the knowledge graph, the way that a news event about a particular company is affecting other companies and other institutions.

Why the energy sector?

Wills: Personally, I'm a huge fan of the energy space, with a physics background I'm obsessed with nuclear, and I'm also interested in green tech. It's especially sensitive to news that doesn't just affect a company specifically, but the broader market and geopolitical news. So for our type of algorithm, the energy sector is ideal.

How often does the Portfolio actually change?

Adam: With our Autopilot portfolio, we don't update it in real time as events are happening. We aggregate information over a longer period of time, because we don't update our portfolio every single time an event comes in. We update it on a period of a week or two.

How should someone who likes the idea, but is nervous about trusting a system with their money, think about it?

Eli: We prioritize not veering off too far from the market. We constantly track what our current holdings are relative to a baseline, and we make sure that these aren't too far off. We make sure that not one sector of the energy market is too heavily weighted.

If someone follows the Energy Portfolio on Autopilot, what are they betting on?

Eli: They're betting on two things. One is that while we're focused on the energy sector, the energy sector can track the broader market. And then on top of that, they're betting on our ability to interpret the news and the predictions that come along with that.

“"Finance seemed like a natural area, partly because you can think of it as a zero sum game, so it's the ultimate test of how you design something."”

Follow Wills Harris's Portfolios on Autopilot

Wills Harris and the Polaron team publish The Energy Portfolio on Autopilot, an investing app that lets you follow Portfolios from investors, AI models, and public filings in your own brokerage account. See the Wills Harris Pilot page for the full list. https://link.joinautopilot.com/eXKr?c=wills-harris-interview&af_sub1=body

  • You follow in the brokerage account you already have, and the money stays there.
  • On Premium Tier, when the Portfolio changes, Autopilot Advisers sends the orders to your broker. On Basic Tier you get an alert.
  • Review the Portfolio's method, holdings and any available performance record before following it. New Portfolios show return figures after 30 days.
  • The Pilot sets the subscription price, there's a 0.00% AUM fee, and you allocate at least $500.

You're never trading at the same moment as the Pilot, and your account won't match the Portfolio exactly.

Open it in the Autopilot app. https://link.joinautopilot.com/eXKr?c=wills-harris-interview&af_sub1=end

"Autopilot" refers to Autopilot Holdings Corp and Autopilot Advisers, LLC, an SEC-registered investment adviser. The Energy Portfolio is managed by Polaron, an independent Third-Party Pilot unaffiliated with Autopilot. Autopilot doesn't manage, validate, or guarantee this Pilot's portfolio. This interview reflects their own views as of June 2026, edited and condensed, and isn't a recommendation to buy or sell any security or to follow any Portfolio. Concentrated and thematic strategies carry heightened risk of loss. Where a Pilot uses an AI tool in their own stock-selection process, that tool does not advise clients, select investments for clients, or place trades. Frequent trading in your account may result in short-term capital gains, which are generally taxed at higher ordinary income tax rates. Investing carries risk-including possible loss of principal-and past performance is not indicative of future results. Figures current as of 10/09/26. View our Customer Relationship Summary and other important info at joinautopilot.com/disclaimer.

The performance data presented reflects the aggregated results of actual client portfolios managed by the Registered Investment Adviser (RIA), based on the trading activity in the securities contained in the relevant portfolio over the specified period. If fewer than 10 clients are invested in the portfolio, the data represents the aggregate performance of all clients following this strategy. Performance figures are shown net of advisory fees, calculated using the applicable subscription fee and an assumed $50,000 account size. Because subscription fees are flat rather than asset-based, accounts smaller than $50,000 would experience a higher effective fee percentage and lower net performance, while larger accounts would experience a lower effective fee percentage. Actual investor outcomes can vary due to factors like brokerage support for fractional shares, account size, trade timing, fees, and taxes, and past performance should not be interpreted as a guarantee of future results or typical of all portfolios.

Ask about this Portfolio in ChatGPT