Three Founders Say Their AI Beat Bridgewater Last Year

A small team says it had the best year in the industry

In January, Vertus, a company of three founders based on the Isle of Man, published its 2025 results through PR Newswire: a 51.15% return, a Sharpe ratio of 2.13, and a day in late November when its models moved over $1 billion in trading volume. In the same release Vertus put the S&P 500 at about 17% for the year, Bridgewater at 34% and D.E. Shaw at 28.2%. The numbers are the company's own, checked by an outside accounting firm at its request, with the report available on request.

Read that paragraph again with the size of the team in mind. Three people. Julius Franck, a German quant. Alex Foster, who built algorithmic trading infrastructure in the UK. Michal Prywata, a Polish-Canadian engineer who co-founded Bionik Laboratories. Vertus says it is self-funded and has taken no venture capital. The largest funds in the world spend hundreds of millions a year on quant research. A three-person team posting a year like that, on its own money, is the interesting part of this story.

The architecture is the story, more than the return

ChatGPT, Claude and Gemini all sit on the same foundation: the Transformer, published by Google researchers in 2017. Every improvement since has been a bigger, better-trained version of the same idea. Vertus went a different direction. Its models are built on what it calls brain topology, a structure modeled on how the human brain routes, connects and decides, with millions of artificial neurons wired into configurations shaped by the problem in front of them. The product names give the scale away: an 8 Meganeuron general model, a 16 Meganeuron coding model.

Vertus is not alone on this road. Brain-inspired and non-Transformer architectures are one of the most active research directions in AI right now: liquid networks, state-space models, spiking networks. What makes Vertus different is where it chose to prove the idea. Markets are the hardest benchmark there is. They do not care about your paper. They pay you or they do not. Choosing that as the first arena is a bold call, and the claimed result is why people are paying attention.

Why finance is where new AI shows up first

Markets are the ideal proving ground for a new kind of intelligence. The data is enormous, it updates every second, and the scoring is instant and beyond argument. A model that reads price, news, flow and sentiment across every major market at once and turns that into trades is doing, at machine speed, what a floor of analysts does at human speed. For decades this was locked inside a handful of firms with the budget to build it. Renaissance, Citadel, Two Sigma.

What is changing is who can build it. A quant, an infrastructure engineer and a hardware founder can now stand up a model that competes at that level, and they can do it from a small island rather than a Manhattan tower. Vertus describes itself as the technology layer underneath funds, family offices and professional investors, rather than a fund itself. That is a new shape for the industry: the intelligence as a product, sold to whoever runs the capital. If it holds, the edge that used to come from headcount comes from architecture instead.

What opening the models would mean

The most forward-looking part of Vertus is the plan, more than the trading. The company already lists a chat interface and a developer API on its site, and the reel from Curious Dre that put this in front of me had the state of play right: interest has been high enough that signups are paused and waitlisted, and there is no confirmed release date. Nothing official yet.

Consider what it means if it happens. Today the best AI models in the world learned from the internet and are judged on exams and coding puzzles. A model that learned by being scored against live markets, every day, for a year, has been trained on a very different kind of feedback. Whether that makes it a better general reasoner is an open question, and a genuinely exciting one. It would be the first time a model that earned its keep in the hardest arena there is was handed to the public to try on their own problems.

How I am reading it

I run my business on AI every day, and I have said for a long time that the real prize is predictability: knowing what happens next before you spend the money. Trading is predictability in its purest form, and it is where new approaches get proven or discarded fastest. A team of three, a new architecture, and a year that, by its own account, beat the biggest names in the industry is exactly the kind of signal I watch for.

The sensible move today is simple. Join the waitlist. It costs nothing. When the models open, put them on your own problems and measure. The returns are the company's claim until a public track record sits behind them; the architecture is worth studying either way. Whatever the final numbers turn out to be, the direction is set. Intelligence that learns from consequences, built by small teams, sold as a layer instead of locked inside a fund. AI is entering investing from an angle nobody planned for, and I want to be watching when it lands.

Build the Cystem. Watch it work.

Sources: Vertus press release on PR Newswire, January 2026, and the Vertus website. The reel that surfaced it is from Curious Dre on Instagram. The Transformer paper is Attention Is All You Need, Vaswani et al., 2017.