In a nondescript office somewhere in New York, a single founder is attempting something that would have seemed like science fiction just a few years ago: running a hedge fund where artificial intelligence doesn't merely assist with trades—it executes them. No human analyst burning midnight oil over earnings reports. No portfolio manager overruling the algorithm when conviction wavers. Just AI, making the calls.
Standard Signal, a Y Combinator-backed venture, bills itself as "the first hedge fund that researches and executes trades purely with AI." Bold claim. Perhaps even brash. But founder Michael Royzen isn't exactly new to ambitious declarations about artificial intelligence. He previously built Phind, an AI search engine that reportedly processed over 150 million searches and developed what the company described as "the best-performing coding LLMs" of its era.
Now Royzen is wagering that the same large language model technology can uncover what Standard Signal's website calls "new fundamental truths" in financial markets. The fund claims a market-neutral strategy with a Sharpe ratio above 2.0—a level of risk-adjusted returns that, if achieved, would place it among the elite performers in the industry. Trades are "fully explainable," the marketing materials insist. Autonomous financial LLMs operating "within strict risk parameters." Minimum investment? $100,000.
These are unverified marketing claims, not audited results. Worth noting. But they signal something larger happening across the hedge fund industry: the migration from AI as a useful tool to AI as the strategy itself.
The Illusion of Ubiquity
By late 2025, when the Alternative Investment Management Association published its latest survey, the headline number seemed definitive: 95% of hedge fund managers reported using generative AI, up from 86% two years prior. Among 150 managers representing approximately $788 billion in assets under management, 58% expected increased use of AI in alpha generation—nearly triple the 20% who said the same in 2023.
But these numbers mask a messier reality. Bloomberg's 2025 European Institutional Equity Trading Study found that AI adoption on trading desks "remains limited and mostly in testing." McKinsey's State of AI report from November 2025 revealed that while 88% of organizations use AI in at least one function, only 23% have scaled agentic AI—the autonomous kind that can act independently—in any part of their business. Another 39% are still experimenting.
The gap between "using AI" and building an entire fund around it? Vast.
Most firms deploying AI today are using it to parse earnings calls faster, screen stocks more efficiently, optimize execution algorithms. That's a far cry from handing the entire research and trading process over to a machine. The difference between assistance and abdication.
The Autonomous Cohort
Standard Signal isn't alone in testing the fully autonomous thesis. Its Y Combinator batch-mate AYVID also emerged positioning itself as the "first fully autonomous AI hedge fund," deploying what it describes as "LLM agent swarms" across six asset classes and more than 300 instruments. The firm showcases walk-forward backtests spanning 2018 to 2025 and currently trades on Interactive Brokers in paper mode while pursuing seed conversations, according to its materials.
Then there's AIDGE, which brands itself as an "Autonomous AI Hedge Fund" but provides scant public detail about its approach or progress.
These are startups, mostly. Marketing sites. Demo consoles and pitch decks. None have established track records, and their performance claims remain unverified by independent audits. But they represent a philosophical break from the past: the idea that AI can be trusted not just to assist but to lead. To replace, even.
The counter-narrative comes from a different corner of the AI-native investing world. Numerai, founded in 2015 as a crowdsourced machine learning hedge fund, announced a $30 million Series C round in November 2025 at a $500 million valuation, according to the company. More tellingly, JPMorgan Asset Management committed capacity of up to $500 million to the fund in August 2025, per Numerai's announcement. The firm reported a net return of 25.45% for its Meta Model in 2024, with only one down month—though these performance figures, like the funding details, are company-reported and not independently audited.
Numerai's model isn't fully autonomous in the Standard Signal sense. It aggregates predictions from thousands of data scientists competing to build the best models—a crowd-sourced approach rather than a solitary AI agent. But its validation by a Wall Street titan suggests institutional appetite for AI-native approaches exists when governance, transparency, and track records align.
When, perhaps, the humans haven't been entirely removed from the equation.
The Founder's Wager
Michael Royzen's résumé reads like a highlight reel of the LLM era. Before Standard Signal, he co-founded and led Phind, which built conversational AI for developers. Before that, he worked on machine learning at Lyft, Cloudflare, and Microsoft. He graduated as a Turing Scholar from the University of Texas at Austin—a program that produces some of the field's top computer scientists.
In an August 2024 podcast on building LLM search engines, Royzen detailed the technical architecture behind Phind's systems, including how the company trained coding models that reportedly outperformed established benchmarks. The discussion offered a window into his engineering philosophy: iterative improvement, rigorous testing, a focus on reliability over hype.
That same philosophy appears to inform Standard Signal's approach. The fund claims its models operate "within strict risk parameters" and produce "fully explainable" trades—two assertions that, if true, would address the black-box concerns that have plagued earlier generations of algorithmic trading.
But skepticism is warranted here. The SEC's March 2024 enforcement actions against Delphia (USA) Inc. and Global Predictions Inc. for misleading AI claims established a precedent: overstating AI capabilities can trigger regulatory consequences. Real ones.
Standard Signal's public-facing materials carefully thread the line between confidence and caution. The website emphasizes explainability and risk management. But without audited performance data or regulatory filings, allocators evaluating the fund are left to weigh the founder's technical credentials against the inherent uncertainty of a first-time investment manager deploying untested technology in live markets. It's a bet on potential, not proof.
Ghosts of Funds Past
The history of fully autonomous AI trading is short. And sobering.
In January 2016, Aidyia launched an AI-driven hedge fund with the explicit promise of "no human intervention." The firm attracted attention for its ambition. Two years later, coverage vanished. The firm, too.
Sentient Investment Management, which spun out of Sentient Technologies with considerable fanfare, shut down in 2018 after less than two years of operation. The reasons remain opaque, but the outcome is clear: building an AI hedge fund that survives contact with real markets is harder than the pitch decks suggest. Much harder.
On the retail side, AI-managed ETFs offer a longer but mixed track record. The AI Powered Equity ETF (AIEQ), launched in October 2017 using IBM Watson, has struggled to consistently outperform the S&P 500 over multi-year periods. Qraft's suite of AI-enhanced ETFs—including AMOM and QRFT—have produced what the company calls performance "milestones," though investors should verify claims with sponsor data rather than marketing materials. Always.
These cautionary tales don't invalidate the autonomous thesis. They do suggest that early-stage AI funds face formidable challenges: overfitting to historical data, underestimating regime changes, failing to account for market microstructure, and—perhaps most critically—running out of capital before the technology matures. Or before investors lose patience.
The Compliance Minefield
Regulators on both sides of the Atlantic are watching the autonomous AI trend with a mix of interest and suspicion.
In the United States, the SEC's Division of Examinations named AI use, automated tools, and algorithms as examination priorities for 2026. The agency's April 2024 sweep against five advisers for Marketing Rule violations underscored the risks of displaying hypothetical or backtested performance on public websites without compliant disclosures and recordkeeping. A December 2025 Risk Alert reiterated these concerns, if anything more forcefully.
FINRA's 2026 Annual Regulatory Oversight Report included, for the first time, a dedicated section on generative AI and explicit discussion of "AI agents." The guidance emphasizes governance, testing, and monitoring—baseline expectations that apply to broker-dealers but signal broader supervisory trends across the industry.
In Europe, the regulatory landscape is even more complex. The EU AI Act's core obligations are expected to take effect in August 2026, though proposals under the Digital Omnibus may shift specific high-risk application timelines into 2027. Hedge funds operating as EU investment firms must continue meeting MiFID II Article 17 and RTS 6 requirements for algorithmic trading, which now intersect with AI Act obligations for any systems classified as high-risk.
In February 2026, ESMA published a supervisory briefing on algorithmic trading that explicitly addressed the risks of advanced AI—including deep learning, reinforcement learning, and generative AI—in signal generation. The briefing stressed robust governance, pre-trade controls, kill switches, conformance testing, and oversight of outsourced software. Compliance staff, ESMA noted, must understand how AI-driven algorithms operate. Not just how to monitor their output.
For a one-person shop like Standard Signal, meeting these expectations at scale will be a test of operational maturity as much as technical prowess. Maybe more.
What Success Looks Like
If autonomous AI hedge funds are to move beyond the graveyard of failed experiments, they'll need to demonstrate several things simultaneously. Not sequentially—simultaneously.
First, performance that survives regime changes. Walk-forward backtests are necessary but not sufficient. Allocators will demand evidence that models can navigate unexpected shocks—interest rate reversals, geopolitical crises, flash crashes—without human intervention producing catastrophic losses. The kind of losses that end careers.
Second, explainability that satisfies both investors and regulators. "The AI made this trade" is not an explanation; it's an abdication. Firms claiming full autonomy must be able to reconstruct decision paths, identify which signals drove conviction, and document that risk parameters were respected in real time. AIMA's 2025 survey found that investors increasingly reward managers who pair AI capability with clear governance and human oversight frameworks, even when those frameworks are lighter-touch than traditional approaches.
Third, compliance infrastructure that scales with ambition. Managing model risk, maintaining audit trails, substantiating marketing claims, implementing kill switches, ensuring data protection—all of this will require investment in systems and personnel. The one-person team structure that works for a beta product may not survive regulatory scrutiny at scale. Almost certainly won't.
Fourth, differentiation beyond the technology itself. As FactSet rolls out AI-enabled document search to 85,000 users and Bloomberg integrates its finance-tuned LLM across platforms, the data and infrastructure advantages that early AI funds might have enjoyed are eroding quickly. Standard Signal's claim of discovering "new fundamental truths" implies a proprietary model architecture or training approach. Whether that translates to sustainable alpha remains to be seen. The market has a way of humbling bold claims.
The Allocator's Dilemma
For institutional investors evaluating these funds, the calculus is uncomfortable.
The upside—gaining early access to a genuinely transformative strategy—competes with the downside of backing unproven technology managed by first-time operators in a regulatory environment that's tightening by the quarter. It's a bet most large allocators aren't built to make.
JPMorgan's half-billion-dollar capacity commitment to Numerai offers one data point: large allocators are willing to bet on AI-native approaches when the firm has a track record, institutional-grade operations, and transparent governance. But Numerai had nearly a decade to build that credibility. The cohort of fully autonomous funds launching now is asking investors to make that leap in months, not years.
AIMA's research suggests that allocators are increasingly differentiating between "AI feature" funds—those that use AI as one tool among many—and evidence-backed, well-governed AI-native strategies. The former may get meetings; the latter will get capital. The question is how long firms like Standard Signal, AYVID, and AIDGE have to prove which category they belong in before the market moves on. Or before the capital runs out.
Meanwhile, concentration risk around AI mega-cap stocks complicates the hedging calculus. PivotalPath data cited by Axios in October 2025 showed rising correlation and crowding around companies like Nvidia, increasing beta exposure even for funds claiming market-neutral strategies. If AI-native hedge funds are themselves directionally exposed to the AI trade through their holdings, the diversification benefits for allocators diminish. The irony isn't lost on anyone.
Agents at the Gate
The technology underlying autonomous hedge funds—large language models, reinforcement learning, agentic workflows—is advancing faster than regulatory frameworks can adapt. McKinsey's data showing 23% of organizations scaling agentic AI in at least one function suggests the broader enterprise world is beginning to trust agents with consequential decisions. Finance, with its regulated, high-stakes environment, will likely lag that curve.
But perhaps not by much.
By late 2026 and into 2027, the regulatory environment should be clearer. The EU AI Act's initial enforcement wave will offer case studies on what "high-risk AI system" obligations mean in practice. The SEC's examination priorities will have produced enforcement actions—or their absence—that signal how aggressively U.S. regulators intend to police AI governance gaps. FINRA's guidance on AI agents will have evolved from principles to expectations, from suggestions to requirements.
For Standard Signal and its cohort, the window to establish track records before that scrutiny intensifies is narrow. The fund launched recently. If it can demonstrate live, audited performance through multiple market regimes while maintaining regulatory compliance and capital inflows, it may yet validate the fully autonomous thesis.
Or it may join Aidyia and Sentient in the archive of ambitious experiments that proved the technology wasn't ready—or the markets weren't ready for the technology. The difference, this time, might be that the underlying models are finally sophisticated enough to close the gap between promise and execution.
Might be.
Either way, the industry is watching. Ninety-five percent of hedge funds may be using AI in some capacity. But the real question is whether any of them are ready to let it take the wheel entirely. To trust the machine not just with analysis but with capital. With livelihoods. With the future.
The answer, for now, remains unwritten. But the experiments are live, the money is moving, and the founders are betting their reputations that this time—unlike Aidyia, unlike Sentient—the technology is finally ready for the audition.
Whether the audience is ready for the performance is another matter entirely.
