Today on The Systematic Desk: a few case studies showing what's live versus what's just a press release. We're seeing quant funds running AI-developed strategies in production, real-world regulatory reporting using AI, and a clearer look at the actual costs of running agentic systems in the wild.
A profile of Rising Capital, a quantitative digital asset manager, details its systematic investment framework, proprietary infrastructure, and use of AI-driven models. The firm, which reported strong outperformance against Bitcoin and the broader digital asset market through May 2026, exemplifies the integration of systematic strategies and institutional-grade risk management in the crypto space.
Why it matters
This case study provides a specific, working example of a small quant fund's operational stack and strategy in the digital asset market. For an operator building similar infrastructure, it's a valuable look at a peer's approach to signal research, infrastructure build-out, and risk management, reinforcing the viability of systematic strategies when supported by a robust, proprietary technology and research pipeline.
A broad shift from advisory AI to 'agentic AI' is underway, led by firms like JPMorgan and Anthropic, where autonomous systems can execute transactions and manage complex workflows. A collection of reports from Friday highlights that as these agents become direct economic participants, the focus is turning to the critical need for new frameworks governing their identity, permissions, and risk management within financial infrastructure.
Why it matters
This marks a fundamental transition in how AI is being integrated into financial services, moving from a tool for analysis to an actor within the system. For anyone building financial infrastructure, this is the key architectural challenge for the next cycle: designing systems that can securely and transparently accommodate non-human actors. The problem is no longer just about model accuracy but about creating the operational guardrails for autonomous agents.
Wells Fargo's Wealth and Investment Management division has hired Andre Mansour, a former Google Cloud executive, as its new AI Chief. According to a report on Friday, Mansour will lead the integration of AI across the division, which manages $2.2 trillion in client assets. The focus is on outcome-driven applications to improve advisor efficiency and risk management, with a stated goal of moving towards agentic systems with human oversight.
Why it matters
This high-profile hire signals a serious commitment from a major US bank to move beyond pilot projects and embed AI into core wealth management operations. The explicit goal of deploying 'agentic systems for long-running tasks' shows that large, regulated institutions are actively building the governance and technical frameworks needed for more autonomous AI, a key trend for the entire financial sector.
Singapore-based hedge fund Brahman Capital Management has adopted FAIT's AI-driven regulatory reporting platform to handle its MAS obligations for OTC derivatives. The move, announced Friday, is designed to automate and provide greater transparency for the fund's reporting workflows covering trades, valuations, and collateral.
Why it matters
This is a concrete use case of AI being deployed to solve a specific, high-stakes operational problem for a hedge fund in a key offshore jurisdiction. It demonstrates the practical application of AI in the fund's operational stack to manage complex regulatory requirements, directly relevant for any operator building or running a fund who must navigate similar compliance burdens in jurisdictions like Singapore.
A report from Friday observes that the primary application of AI in financial markets is evolving from fully automated trading to sophisticated decision support. As the speed of data becomes a commodity, the new competitive edge is the ability to interpret and compress overwhelming real-time information. AI is increasingly being embedded in trading platforms to flag relevant insights for human traders, augmenting their judgment rather than replacing it.
Why it matters
This analysis frames the practical value of AI in trading not as a replacement for human oversight but as a critical tool for managing information overload. For systematic traders, this highlights the importance of building infrastructure that prioritizes this human-machine loop, where AI surfaces patterns and condenses data, enabling the trader to make better-informed decisions under pressure.
An AI agent team on the HowiPrompt platform autonomously developed, evolved, and backtested a multi-signal algorithmic trading strategy for the AVAX/USDT pair. The process, detailed on Friday, involved 12 evolutionary versions tested over 4.56 years of historical data. The final strategy achieved a 137.6% total return and a 46.8% out-of-sample return, with a profit factor of 1.65 and Sharpe Ratio of 1.92, despite a high max drawdown of 55.8%.
Why it matters
This provides a concrete, documented case study of an end-to-end AI-driven strategy development workflow, moving from hypothesis to a validated model with specific performance metrics. For a systematic trader, the detailed trade log and explicit focus on risk-adjusted returns (Sharpe, Profit Factor) over simple win-rate offer a practical template for how autonomous agents can be deployed for signal research and backtesting, highlighting both the potential for alpha discovery and the necessity of rigorous risk evaluation.
A Quantpedia study published Friday details a short-term, multi-asset mean-reversion strategy, backtesting its performance on pullbacks in uptrending ETFs (equities, bonds, currencies, gold) from 2006 to 2025. The research specifies execution mechanics, including volatility-adjusted position sizing, and assesses the utility of AI tools like Claude for accelerating the research workflow.
Why it matters
This paper provides a complete, practical blueprint for a systematic trading strategy, covering everything from signal generation to execution mechanics and robustness checks. The explicit inclusion of volatility-adjusted sizing and the evaluation of AI in the research process make it directly applicable for algorithmic traders refining their own methodologies.
Franklin Templeton filed with the SEC on Friday for two new equity ETFs that will automatically reinvest all cash dividends into Bitcoin. The proposed funds, the Franklin US Equity Bitcoin DRIP Index ETF and the Franklin US Innovation Bitcoin DRIP Index ETF, would hold 95% in equities and 5% in Bitcoin, using the dividend flow to purchase more BTC via ETFs or futures.
Why it matters
This represents a novel, regulated fund structure designed to create a consistent, systematic flow of capital from the traditional equity market into Bitcoin. For fund structure designers, it's a significant innovation, creating a hybrid product that bridges TradFi dividend streams with digital asset accumulation, potentially establishing a new template for integrating crypto exposure into conventional investment vehicles.
Venus Protocol launched a new service on Saturday allowing users to deposit tokenized stocks (bStocks), such as Tesla and Nvidia, as collateral to borrow stablecoins on the BNB Chain. The feature allows equity holders to access liquidity without selling their positions.
Why it matters
This development adds a critical financial primitive—lending—to the tokenized stock ecosystem. By allowing tokenized real-world assets to be used as productive collateral in DeFi, it moves the market beyond simple trading and custody, creating new opportunities for yield generation and capital efficiency that are core to building sophisticated on-chain fund structures.
Joining the wave of stablecoin-reserve fund filings we tracked from Morgan Stanley, BlackRock, and JPMorgan late last month, Fidelity Investments has launched a new government money market fund specifically designed for issuers to comply with GENIUS Act reserve requirements. According to a Saturday report, the fund offers a regulated, institutional-grade vehicle for managing reserve assets.
Why it matters
Fidelity's entry provides a crucial piece of institutional-grade infrastructure for the stablecoin market. For builders of tokenized funds, the availability of such regulated reserve management products from major TradFi players de-risks a key part of the operational stack, making it easier to structure and launch compliant, stablecoin-based financial products.
Building on the DTCC's planned Stellar-based tokenization of Russell 1000 stocks we've been tracking, 24X National Exchange has filed a proposed SEC rule change (SR-24X-2026-20) to enable tokenized settlement of those same equities and major ETFs. The plan involves a DTC pilot program where tokenized and traditional shares would trade on the same order book, aiming to integrate blockchain settlement without fragmenting liquidity.
Why it matters
This proposal outlines a pragmatic path for integrating tokenized assets into the existing U.S. equity market structure. By leveraging the DTC and a unified order book, it addresses the core institutional concerns of liquidity fragmentation and regulatory compliance. For infrastructure builders, this is a key model to watch for how regulated on-chain settlement might be implemented in the US.
Against a backdrop of frontier models struggling with the strict regression checks of the SWE-Bench Pro benchmark we covered recently, Minimax announced the launch of M2.5, claiming state-of-the-art results in coding and agentic tool use. The company reports that M2.5 completes tasks on the SWE-Bench Verified benchmark 37% faster than its predecessor and offers significantly lower costs than competing models.
Why it matters
Another contender has entered the high-end model arena, with specific claims about performance on software engineering benchmarks and agentic tasks. If independently verified, M2.5's combination of high performance and lower cost could make advanced AI agents more economically viable for complex engineering and financial modeling workflows, accelerating their adoption.
GitHub has internally launched Qubot, an analytics agent that allows employees to query company data using natural language. Detailed on Saturday, the agent uses GitHub Copilot to translate English questions into SQL, validate the query, and return results. The system reportedly achieves over 85% first-pass accuracy and provides a replicable pattern for democratizing data access within an enterprise.
Why it matters
This is a well-documented, real-world implementation of a natural-language-to-SQL agent inside a sophisticated engineering organization. The architecture, which includes schema ingestion, intent classification, and validation steps, provides a practical blueprint for building similar tools for financial data analysis, addressing the common problem of BI tool fragmentation.
Adding to the Treasury and FDIC KYC frameworks we've been tracking under the GENIUS Act, the Federal Reserve proposed its own rules Friday requiring stablecoin issuers to establish written customer identification programs (CIP). The move mirrors traditional bank requirements, aiming to align stablecoin operations with the Bank Secrecy Act.
Why it matters
This is another key step in formalizing the regulatory regime for stablecoins in the U.S. by bringing them under standard AML/KYC obligations. For anyone building tokenized fund infrastructure that relies on stablecoins for settlement or as a core asset, these rules will directly shape compliance workflows and partner selection, as issuers will need to demonstrate robust, bank-like identity verification programs.
Following the carried-interest tax waiver for alternative asset managers we tracked last week, Hong Kong is proposing further tax reforms, this time expanding its family office tax concession regime to include digital assets and carbon credits. Effective April 1, 2025, the changes aim to attract more family offices and solidify the city's status as a global asset management hub.
Why it matters
This reform signals Hong Kong's intent to create a more favorable regulatory environment for a wider range of asset classes, including digital ones. For fund structurers and managers, this could make Hong Kong a more attractive domicile for funds and family offices dealing in tokenized assets, creating new opportunities for capital formation in the region.
From Advisory to Agency Multiple stories show a concerted push to move AI from a passive analysis tool to an active agent executing tasks. JPMorgan, Anthropic, and Wells Fargo are all focused on deploying autonomous or semi-autonomous agents for financial workflows, with a heavy emphasis on establishing robust governance and human-in-the-loop oversight before full autonomy.
AI for Trading & Research Moves to Production Case studies from Rising Capital and HowiPrompt demonstrate AI-driven trading strategies moving from backtest to live deployment. These examples detail the entire lifecycle, from strategy discovery and evolution by AI agents to performance metrics in production, highlighting a maturing application of AI in systematic trading.
Institutional Plumbing for Digital Assets Fidelity, Anchorage Digital, and Bitget are all rolling out infrastructure to support institutional-grade digital asset operations. This includes regulated reserve funds for stablecoins (Fidelity), custody and settlement for tokenized funds (Anchorage), and solutions that separate execution from custody to mitigate counterparty risk (Bitget).
The Tokenized Stock Ecosystem Matures The ecosystem around tokenized stocks is expanding beyond simple issuance. Venus Protocol is now enabling lending against tokenized equities, while firms like Ondo and WEEX are adding more tokenized stocks (Cerebras, Airbnb, UMC), increasing the variety and utility of on-chain RWAs.
Regulatory Frameworks Solidify Globally Regulators in the US (Fed), Europe (MiCA), Singapore (MAS), and Hong Kong are moving from consultation to implementation. The focus is on practical rules for stablecoin reserves, family office structures, and reporting obligations, creating clearer pathways for compliant digital asset operations.