Reflexivity // the firm formerly known as Toggle AI

A trading desk, minus the panic

It is 6:40 a.m. on a Tuesday in Midtown Manhattan. A portfolio manager has roughly twenty minutes before the market opens and a dozen questions that used to take a junior analyst all morning. Which of my holdings are exposed to that overnight supply-chain headline? What did the 10-K actually say about margins? If rates move 50 basis points, what breaks? Historically, the answer to all of these was the same: coffee, spreadsheets, and hope.

That morning now has a different shape. The questions go into Reflexivity in plain English. The answers come back in seconds, with the underlying filings and data points attached so the manager can check the work. No séance with a chatbot that confidently invents a number. No twelve browser tabs. Reflexivity, the New York fintech that used to be called Toggle AI, has quietly become the thing on the second monitor of people who manage serious money.

What it does, in one breath: it takes institutional-grade financial data, wires it into a proprietary Knowledge Graph, and lets large language models answer hard investment questions on top of it - without making things up. That last clause is the entire business.

AI that lies is worse than no AI at all

Everyone in finance wanted to point a large language model at the market and ask it questions. There was one inconvenient detail: language models hallucinate. They produce a wrong number with the same calm confidence they produce a right one. In most industries that is an annoyance. In investing, where a fabricated figure can move real money, it is disqualifying.

The founders had lived the other side of this problem too. As working portfolio managers, they knew the daily friction was not a shortage of data - it was the opposite. Bloomberg here, a data vendor there, a stack of SEC filings nobody had time to read, and the genuinely useful relationships between all of it living only in an analyst's head. The information existed. Connecting it, in time to act, did not.

So the central tension was set, and it has never really changed: give investors the speed of AI without the recklessness of it. Fast and wrong is easy. Fast and right is the whole game.

Two macro guys and a theory named after Soros

Jan Szilagyi and Giuseppe Sette did not arrive from a coding bootcamp. They ran global macro together as co-Chief Investment Officers at the Swiss private bank Lombard Odier. Before that, Szilagyi managed money at Fortress Investment Group and - this is the detail people repeat at dinner parties - reportedly finished a Harvard Economics PhD in about two and a half years, under Ken Rogoff, which is said to be the fastest on record. Sette did time at Brevan Howard and Davidson Kempner, topped his class at Bain Italy, and picked up a Wharton MBA along the way.

Their bet had two parts. First, that the people best placed to build investment AI were investors, not outsiders guessing at the workflow. Second, that the company should be named for an idea, not a gadget. They picked reflexivity - George Soros's notion that investors' perceptions of a market feed back into the market itself. It is a fitting flag for a company arguing that the tools you use to see the market quietly change how you move in it.

The rebrand from Toggle AI was not cosmetic. The original product had drifted toward retail investors; the new name marked a return to the institutions the founders came from, and to the harder, more lucrative problem of serving them.

Six tools, one trustworthy spine

Underneath everything sits the Knowledge Graph: a map, built from the ground up for institutional finance, of how the market's pieces connect. Peers, suppliers, customers, partners, geographic exposure, thematic exposure - the relationships an experienced analyst carries in their head, made machine-readable and kept current as new SEC filings and earnings calls land. Reflexivity argues it is the only such graph built specifically for this job. The language model is not allowed to freelance; it reasons over the graph's verified facts.

The data underneath is institutional by default - integrations with S&P Global, LSEG Datastream, Cboe and Nasdaq - which means customers do not have to assemble a wall of separate data contracts before they can ask a single question. For technology teams there is an API, so the same intelligence can flow into systems a firm already runs.

The receipts

A trust-first pitch is easy to make and hard to back. Reflexivity's evidence is its cap table and its customer list, which overlap in a way that is hard to fake. The October 2024 Series B - $30 million, led by Greycroft and Interactive Brokers - drew checks from Stanley Druckenmiller, Greg Coffey, Thomas Peterffy, General Catalyst and SoftBank's LatAm fund. These are people who evaluate market tools for a living, and they were buying.

The customers tell the same story. Reflexivity counts Millennium and Soros Fund Management among its users - firms that do not adopt research tools for the novelty. And the Interactive Brokers investment is also a distribution deal: IBKR has signaled it will weave Reflexivity's AI into its electronic trading platform, putting the tools where traders already research, monitor and execute.

Trust, at the speed of the market

Strip away the funding and the famous backers and the mission is unglamorous in the best way: give the people who move markets AI they can rely on. Fast, explainable, grounded in verified data rather than guesswork. The team's worldview is that in finance, an answer you cannot check is not an answer - it is a liability with good grammar.

There is a larger ambition underneath. Szilagyi has argued publicly, including on CNBC, that AI will reshape and in places replace parts of the financial advisory role. Read generously, that is the company's real bet: that reliable market intelligence should not be a privilege reserved for the desks that can afford a small army of analysts. Start with the institutions, prove the trust, then let it spread - eventually, through a partner like Interactive Brokers, toward the individual investor too.

Back to 6:40 a.m.

Return to that Midtown morning. The market still opens at 9:30, the headlines still arrive overnight, and the questions are no smaller. What changed is the gap between question and action. The manager who once spent the morning assembling context now spends it deciding what to do with context already assembled - and, crucially, can click straight through to the filing that backs each claim.

That is the quiet shift Reflexivity is betting on. Not AI that thinks for you, but AI that does the gathering and the connecting, then shows its work so you can think faster. If they are right, "I'll have the analyst look into it" stops being a sentence anyone says before lunch. The coffee is still there. The panic, increasingly, is optional.