Prodigy

The AI-Native Trading Lab: A New Species in Quant Finance

Prodigy Research is not your typical fintech startup. It's an AI research lab that trains foundation models specifically for quantitative trading, and it's already trading live in financial markets. The bold claim: its proprietary model and agent harness outperform leading AI models like Claude Fable and GPT-5.6 Sol on trading and financial-reasoning benchmarks. Even more striking, the company says its AI quant trader beats a 90th-percentile Jane Street trader—a human benchmark that would have seemed unthinkable a few years ago.

This is a new category: AI-native trading, where the model isn't just an assistant but the primary decision-maker. Prodigy is backed by Y Combinator, which gives it early credibility, but the real story is in the details of its performance claims and the team's pedigree.

Live Performance: Up 108% in Two Months—But What Does It Prove?

The centerpiece of Prodigy's website is a performance chart: the firm claims its live systems more than doubled capital over two months, up 108%, while the S&P 500 was up just 0.6% and the NASDAQ was down 3.5%. The chart is striking, but it raises immediate questions. Two months is a short window, and such extraordinary returns are rare in institutional trading. The company also states it has 'never had a down week,' which is an extraordinary claim that invites skepticism.

Prodigy attributes this performance to 'rigorous and comprehensive risk controls' and delta-neutral strategies, which aim to be market-neutral. But without audited statements or third-party verification, these numbers remain self-reported. For a startup, such performance claims can be a double-edged sword: they attract attention but also scrutiny.

The Benchmarks That Matter: Beating Claude Fable and GPT-5.6 Sol

Prodigy claims its model outperforms Claude Fable and GPT-5.6 Sol on trading and financial-reasoning benchmarks. These are likely hypothetical or internal benchmarks, as no public details are provided. The company does not specify the exact metrics or datasets, making it difficult to assess the validity. However, the claim suggests a focus on domain-specific reasoning rather than general intelligence, which is a smart positioning for a niche AI lab.

The mention of 'agent harness' implies a system that combines the model with tools for data retrieval, execution, and risk management. This is consistent with the trend toward agentic AI, but again, specifics are lacking.

The Team Behind the Model: From DeepMind to Jane Street

Prodigy's co-founders bring a rare combination of AI research and trading expertise. Michael Wang, CEO, trained state-of-the-art foundation models at Google DeepMind, including Gemini 3.1 Pro, and previously worked as a quant and Series 57-licensed trader at Jane Street, where he built strategies that generated billions in PnL. Yuhua Wang, CTO, built AI at Apple and scaled Salesforce's AI platform to millions of queries per day.

This pedigree is crucial for credibility. DeepMind and Jane Street are both top-tier institutions, and the combination of these skills is rare. It suggests that Prodigy understands both the technical and the domain-specific challenges of trading.

The TAM Argument: Why Prodigy Targets the Largest Market on Earth

Prodigy's 'Why now' section argues that trading firms generate hundreds of billions in profit annually, with Jane Street alone on track for a $64 billion run rate in 2026. Global equities markets exceed $150 trillion in capitalization, and U.S. equity average daily volume is growing more than 40% annually. These figures are from the website and are not independently verified, but they illustrate the massive opportunity.

The TAM is indeed enormous, but it's also a crowded space with established players like Renaissance Technologies and Two Sigma. Prodigy's edge would need to be its AI-native approach, which could potentially find alpha in ways that human-driven quant models cannot.

The Risks and Open Questions: Trust, Verification, and the 'Never a Down Week' Claim

Prodigy's claims are bold, but they come with significant risks. First, the lack of third-party verification for performance and benchmarks is a major concern. Second, the 'never had a down week' claim is extraordinary and could be a red flag if it's not substantiated. Third, the regulatory landscape for AI-driven trading is still evolving, and Prodigy's approach may face scrutiny.

Moreover, the website includes a disclaimer that it does not constitute an offer to sell securities, and past performance is not indicative of future results. This is standard, but it underscores the speculative nature of the venture.

For potential partners or investors, the key questions are: Can Prodigy sustain its performance over a longer period? Will it publish audited results? How will it handle market regime changes? These are open questions that only time can answer.

Prodigy Research is a fascinating experiment at the intersection of AI and finance. If its claims hold up, it could redefine quantitative trading. If not, it will be a cautionary tale about the hype cycle in AI. Either way, it's a company worth watching.