A recent post by Gavriel Cohen on X (formerly Twitter) has reignited a spirited conversation among financial professionals and technologists about the accelerating role of artificial intelligence in market analysis, portfolio construction, and the broader transformation of how Wall Street processes information. Cohen’s thread, which quickly gained traction among industry insiders, touches on themes that have been simmering across trading desks and venture capital boardrooms for months — but with a specificity and urgency that suggests the conversation has entered a new phase.
The post, shared via Cohen’s X account, contributes to an ongoing discourse about how AI tools are reshaping the workflows of analysts, traders, and portfolio managers. While the precise contents of Cohen’s thread center on the practical applications and limitations of AI in financial decision-making, the broader implications extend to questions about job displacement, alpha generation, regulatory oversight, and the very nature of what constitutes an edge in modern markets.
AI’s Growing Footprint on Trading Floors and Research Desks
The adoption of AI across the financial services industry has moved well past the experimental stage. According to a May 2025 report from Bloomberg, major banks including JPMorgan Chase, Goldman Sachs, and Morgan Stanley have collectively increased their AI-related technology budgets by more than 35% year-over-year, with much of the spending directed toward large language models (LLMs) that can parse earnings calls, regulatory filings, and macroeconomic data at speeds no human team could match. JPMorgan alone has reportedly deployed an internal AI system capable of analyzing more than 100,000 documents per day, extracting sentiment signals and flagging anomalies for its equity research division.
Goldman Sachs, for its part, has been integrating AI assistants into its investment banking and asset management arms, with the firm’s CEO David Solomon telling investors earlier this year that AI is “not a future initiative — it is a present-tense competitive requirement.” These are not idle statements. The arms race for AI talent on Wall Street has pushed compensation for machine learning engineers and data scientists to levels that rival — and sometimes exceed — those offered to traditional quant researchers, according to reporting from The Financial Times.
What Cohen’s Thread Gets Right About the Current Moment
Cohen’s commentary, as shared on X, resonates because it speaks to a tension that many practitioners feel but few articulate publicly: the gap between AI’s theoretical promise and its day-to-day utility. While marketing materials from fintech vendors and bank press releases often paint a picture of fully autonomous trading systems and omniscient research bots, the reality on the ground is considerably more nuanced. AI tools are powerful accelerants for certain tasks — screening large universes of stocks, summarizing lengthy documents, backtesting strategies across historical data — but they remain prone to hallucination, overfitting, and a kind of confident wrongness that can be more dangerous than simple ignorance.
This is a point that has been echoed by several prominent voices in quantitative finance. Marcos López de Prado, a professor at Cornell and former head of machine learning at AQR Capital Management, has written extensively about the pitfalls of applying AI naively to financial data. In his widely cited work, he warns that the signal-to-noise ratio in financial markets is extraordinarily low, and that many AI models trained on market data end up learning noise rather than signal — a problem that compounds when models are deployed at scale with real capital at risk.
The Regulatory Dimension: Washington Takes Notice
The conversation Cohen has helped amplify also intersects with growing regulatory scrutiny. The Securities and Exchange Commission has been increasingly vocal about the risks posed by AI-driven trading and advisory systems. SEC Chair Gary Gensler, before his departure, had proposed rules that would require broker-dealers and investment advisers to identify and mitigate conflicts of interest arising from the use of predictive data analytics and AI models. While those specific proposals have been revised under the current commission, the underlying concern — that AI systems could systematically disadvantage retail investors or amplify systemic risk — has not gone away.
In Europe, the EU’s AI Act, which began phased implementation in 2025, classifies certain financial AI applications as “high-risk,” subjecting them to mandatory transparency requirements, human oversight provisions, and third-party audits. According to Reuters, several large European asset managers have already begun restructuring their AI governance frameworks in anticipation of enforcement actions. The regulatory environment is creating a two-speed world: firms with the resources to build compliant AI infrastructure are pulling further ahead, while smaller players face the prospect of being priced out of the AI race entirely.
Alpha Generation or Alpha Erosion? The Paradox of Widespread Adoption
One of the more provocative threads running through the current debate — and one that Cohen’s post implicitly addresses — is whether the widespread adoption of similar AI tools across the industry is actually eroding the very alpha they were designed to capture. If every major hedge fund and asset manager is using comparable LLMs to analyze the same earnings transcripts, the same Fed minutes, and the same macroeconomic releases, the informational advantage from doing so approaches zero. The edge, in other words, may not come from having AI, but from having different AI — proprietary models trained on unique datasets, or novel architectures that extract signals others miss.
This is a point that Renaissance Technologies, the famously secretive quantitative hedge fund, has implicitly made for decades. Jim Simons, the firm’s late founder, built his fortune not by using the same tools as everyone else, but by assembling a team of physicists, mathematicians, and computer scientists who approached markets with fundamentally different methodologies. The lesson for the current AI moment is clear: commoditized AI will produce commoditized returns. The firms that outperform will be those that treat AI not as an off-the-shelf solution but as a bespoke capability requiring continuous investment and intellectual differentiation.
The Human Element: Augmentation, Not Replacement
Despite the breathless headlines about AI replacing analysts and traders, the consensus among most senior practitioners remains that AI’s near-term role is augmentation rather than substitution. A 2025 survey conducted by McKinsey & Company found that 78% of asset management executives view AI as a tool that enhances human decision-making, while only 12% believe it will fully replace human portfolio managers within the next decade. The remaining 10% were uncertain.
The reasons for this are both practical and philosophical. Markets are reflexive systems — they respond to the actions of their participants, including the AI systems those participants deploy. A model that works brilliantly in backtesting may fail spectacularly in live trading because its very deployment changes the market dynamics it was trained on. Human judgment, with its capacity for contextual reasoning, ethical consideration, and adaptation to genuinely novel situations, remains an essential complement to algorithmic speed and scale. As one senior portfolio manager at a top-ten global asset manager told The Wall Street Journal earlier this month, “The AI tells me what the data says. I still have to decide what the data means.”
Where the Industry Goes From Here
Cohen’s thread on X arrives at a moment when the financial industry is grappling with the second-order effects of AI adoption. The first wave — deploying AI to automate routine tasks and accelerate research — is largely complete at major firms. The second wave, which is now underway, involves integrating AI into the actual investment decision-making process in ways that are auditable, explainable, and aligned with fiduciary duties. This is a far harder problem, and one that will likely define the competitive hierarchy of the asset management industry for the next decade.
The firms that get this right will not necessarily be the ones with the largest technology budgets. They will be the ones that combine technical sophistication with institutional wisdom — understanding not just what AI can do, but what it should do, and where human oversight remains non-negotiable. The conversation that Cohen and others are driving on platforms like X is a valuable part of this process, bringing transparency and critical thinking to a domain that too often defaults to hype. For industry insiders, the message is clear: the AI transformation of finance is real, it is accelerating, and the winners will be those who approach it with both ambition and intellectual honesty.
As the discourse continues to evolve across social media, academic journals, and boardroom presentations, one thing is certain: the intersection of artificial intelligence and financial markets will remain one of the most consequential and closely watched developments in the global economy. Gavriel Cohen’s contribution to that conversation, however brief, has struck a chord precisely because it reflects the lived experience of practitioners who are building the future of finance in real time — with all the complexity, uncertainty, and opportunity that entails.
