YC S26 Company Prodigy Research Officially Launches, Claims Its Quantitative Financial Model Achieved Over 100% Real Returns During YC Incubation
Prodigy Claims Its Quantitative Model Doubled Real Returns
YC S26 company Prodigy Research officially launched, claiming that its quantitative financial foundational model achieved over 100% real returns during the YC incubation period, effectively doubling the principal.
The company stated that the strategy employs a Delta-neutral structure and reported no weekly losses throughout the entire period; the benchmark was the S&P 500 and NASDAQ indices, which either fell or remained flat during the same period.
Prodigy claimed that its AI quantitative traders surpassed the 90th percentile traders at Jane Street and outperformed Claude Fable and GPT-5.6 Sol in financial research and quantitative assessments. These comparisons are self-reported by the company, with no public evaluation sets, scoring criteria, or independent verification available.
The company has not disclosed the actual principal, trading targets, leverage, turnover rate, trading costs, capacity limits, maximum drawdown, account statements, or third-party audit reports; therefore, the 100% return cannot be directly translated into replicable, scalable, or risk-adjusted strategy returns.
The founding team consists of brothers Michael Wang and Yuhua Wang. Michael Wang previously worked as a quantitative trader at Jane Street and later participated in cutting-edge model training at Google DeepMind; Yuhua Wang worked on AI systems at Apple and Salesforce.
In terms of market mechanism, Prodigy is currently still a research-oriented AI company supported by YC, rather than a quantitative fund raising externally. Investors are purchasing a growth option that "the model can replace financial research and trading labor"; if the strategy is effective and scalable, funds will flow into model training, data, execution infrastructure, and talent; otherwise, they will be consumed by trading costs, slippage, crowded trades, and risk exposure.
Source: Public Information
ABAB AI Insight
Michael Wang's career path connects trading with model development: his experience at Jane Street provides insights into market microstructure, risk management, and trade execution, while his time at Google DeepMind offers cutting-edge model training capabilities. Yuhua Wang's involvement in large-scale AI platform development at Apple and Salesforce suggests that Prodigy's differentiation is not just in single trade signals, but in integrating models, agents, research processes, and real execution into a closed-loop system.
Prodigy previously focused on predictive market trading agents under the name Prodigy AI, aiming to enable agents to autonomously find information discrepancies and execute trades in markets like Polymarket and Kalshi; now upgraded to a foundational quantitative financial model, this indicates a shift in its capital narrative from single predictive market arbitrage to a broader software and asset management market that serves research, trading, risk control, and financial data analysis.
Historically, Renaissance Technologies' Medallion Fund has long been viewed as an extreme success case in quantitative trading, but its core constraint is capacity: as strategy scale increases, returns can be diluted, which is why the fund has primarily been open to internal employees. Even if Prodigy's short-term returns are valid, it must address the same question—whether Delta-neutral returns can still cover market impacts, slippage, financing costs, and competitive replication at larger capital scales.
This relates to capital concentration. Foundational models allow more participants to gain research and coding capabilities, but sustainable trading returns remain concentrated among a few systems that possess unique data, low-latency execution, risk capital, market access, and capacity management capabilities. Models can lower the cost of discovering opportunities, but they cannot eliminate the market rule that opportunities quickly diminish after being traded; the true moat lies not in "the model can predict," but in "it can execute after capital enters."
ABAB News · Cognitive Laws
Backtesting proves ideas, real trading proves friction
Returns can be replicated, capacity may not be replicable
Models discover opportunities, execution determines who can profit