Exclusive: AI Finance Is Now “Institutional Grade”, Says True Trading Co-Founder Igor Stadnyk

Just like in any other industry, GenAI in financial markets is showing plenty of promise. However, much of that conversation remains clouded in hype, vague promises, and misunderstood capabilities. 

To really understand the transformative potential of AI in finance, we talked to Igor Stadnyk, Co-Founder and AI Lead at True Trading. For more than 15 years, Igor has been designing systems that make autonomy transparent and useful, not just for machines, but for people, as well. 

Read more below to find out why AI won’t beat top quant firms anytime soon, where it actually adds value for traders, and how it could reshape trust, execution, and risk in financial markets.

AI-powered trading should be more than automation

It’s Stadnyk’s opinion that AI will not be displacing the dominant players among top-tier quantitative firms.

“AI does not outperform Jane Street, Jump Trading, or Two Sigma on signal quality or execution speed,” he says. “Those firms have structural advantages in infrastructure, talent, and capital that no AI layer can close. Anyone claiming otherwise is just selling.”

That’s why, for him, it’s not about AI outperforming the “creme de la creme”, but about using the tool to meaningfully change outcomes. Stadnyk suggests the answer lies not in raw performance metrics like speed or signal precision, but in how decisions are made under uncertainty.

“The real question is different. Where does AI actually win? In ambiguity.”  

Traditional quantitative models follow fixed rules and perform well when market conditions are stable and predictable. However, they tend to struggle when signals conflict, conditions shift quickly, or the market behaves in ways they weren’t designed to handle.

This is where AI introduces a different set of capabilities. Rather than relying on static rules, AI systems can process broader context, identify non-obvious correlations, and adapt to changing conditions in real time. 

“AI can onboard someone into a complex strategy in minutes, surface non-obvious correlations, flag when a liquidity pool isn’t worth entering, and dynamically recalibrate position sizing based on live account state and market conditions,” Stadnyk says.

He hints that AI is not the one to create new sources of alpha, but it will reduce the friction between strategy and execution. This will be felt, at its greatest, by non-institutional participants, whose performance could drastically improve.

“Bloomberg with AI beats Bloomberg without AI. A retail trader with AI does not beat Bloomberg. That’s not the point,” the True Trading co-founder claims. “The point is that a serious retail trader with AI now operates closer to institutional quality than at any point in history. Not because AI invents new alpha, but because it removes the execution and cognitive gaps that used to be insurmountable.”

This distinction – between generating alpha and enabling its consistent capture – is something Stadnyk emphasized multiple times throughout the interview. It also informs his perspective on one of the most discussed developments in modern finance: the shift toward on-chain, non-custodial systems.

In traditional financial markets, trust relies on layers of intermediaries, regulation, and legal frameworks. Banks, brokers, clearinghouses, and custodians work together to ensure transactions settle correctly and counterparties fulfill their obligations. On-chain systems replace this model entirely.

Verifiability At The Execution Layer: Not A Nice-To-Have

“On-chain, trust shifts from institutional counterparties to cryptographic verifiability,” Stadnyk says. “Smart contracts replace intermediaries. The logic is public, auditable, and executes deterministically. Nobody changes the rules mid-game.”

This represents a different form of trust, one based on transparency and code rather than institutional reputation, and for Stadnyk, that model is stronger than traditional finance. 

Execution logic cannot be altered arbitrarily, and every transaction can be independently verified. At the same time, he acknowledges that the system is still evolving and has not yet reached the level of maturity required for broad institutional adoption.

“For institutional players engaging on-chain today, the practical path still runs through familiar wrappers: audited contracts, legal opinions, custody infrastructure,” he explains. “The regulatory framework isn’t mature yet. But the direction is clear. Verifiability at the execution layer is not a nice-to-have. It’s the only trust model that scales without a counterparty in the middle.”

The idea that trust can be embedded directly into the execution layer has significant implications for how financial systems are designed, he argues. It also intersects with another rapidly developing area: autonomous AI agents operating within markets.

Be careful what you wish for

The concept of self-directed agents executing trades, managing portfolios, or even competing in markets is quite a popular one on X and Reddit these days. But Stadnyk is skeptical about these quickly becoming a dominant force. 

He identified three main roadblocks that are both fundamental and persistent:

The first is data. “Every model has a training cutoff,” he notes. In financial markets, where conditions evolve continuously, historical data alone is insufficient. What matters is the ability to incorporate real-time information while discarding outdated context. This is not simply a matter of increasing data volume. It is a problem of managing relevance.

“Feeding fresh data to an agent while continuously pruning stale context is a genuinely unsolved engineering problem,” Stadnyk says. “Keeping a decade-old news event in working memory while processing today’s price action is at best wasteful, at worst dangerous.”

So the challenge is about choosing what information is kept, what should be discarded, and how fast. Without effective solutions, AI agents risk making decisions based on irrelevant or misleading inputs.

The second constraint is economic. AI systems are not free to operate, since compute, data processing, and infrastructure all cost money. “Models are optimized to be helpful, which means they produce an answer, even when the correct answer is ‘I don’t know,’” Stadnyk explains. In a trading context, this tendency toward overconfidence can translate directly into financial losses.

“If your win rate doesn’t compensate for inference costs, API calls, and operational overhead, the math doesn’t work,” he says. “I’ve watched teams push enormous amounts of on-chain data through context windows and burn through tokens with nothing to show for it. The economics have to close on every single trade.”

Focusing on unit economics adds a level of discipline often overlooked in discussions about AI in finance. A system cannot succeed by producing plausible results alone – it must generate them efficiently, reliably, and in a way that produces real profit.

The third constraint is regulatory. Financial markets operate within clearly defined legal frameworks, and autonomous agents do not yet fit neatly into those structures. “There are no clear rules for what an autonomous agent can or cannot do in financial markets,” Stadnyk says.

He explains that current regulations were created before agent-based systems existed. Consequently, there is no clear framework for handling liability, accountability, or market manipulation when an algorithm is the actor. “There’s no framework for agent liability, no definition of what counts as manipulation when an algorithm acts on model output,” he notes.

Until these questions are resolved, institutional players deploying fully autonomous agents at scale face significant legal and regulatory risks.

Problems: Value Versus Architecture

Perhaps the most critical (and often misunderstood) aspect of AI in financial markets is alignment. As systems become more autonomous, making sure they operate within intended boundaries becomes essential, and Stadnyk approaches this problem from a technical rather than philosophical standpoint.

“Most people asking this question are thinking about it at the wrong level. Alignment in trading systems isn’t a values problem. It’s an architecture problem.”

He says current AI models are designed to be helpful, not necessarily accurate or risk-aware. This creates vulnerabilities, particularly in adversarial environments, and the risks are very, very practical:

“The real threat isn’t a rogue AI deciding to drain your account,” Stadnyk says. “It’s prompt injection, supply chain attacks on tools and skills, and narrative manipulation through poisoned data.”

These attack vectors highlight the importance of controlling how AI systems interact with external inputs and tools. In response, Stadnyk advocates for tightly constrained architectures that limit what agents can access and how they operate.

No Silver Bullet

Stadnyk is no AI-skeptic, but he also doesn’t seem to be overly optimistic. For him, AI is neither a silver bullet that will outperform the most advanced trading firms, nor a marginal tool with limited impact. Instead, it is an enabling technology, one that reshapes how decisions are made, how systems are designed, and how trust is established.

Its influence is already visible in the gradual convergence between institutional and non-institutional capabilities, the emergence of verifiable execution models, and the growing importance of infrastructure-level innovation. At the same time, significant challenges remain, particularly in areas such as data management, economic viability, and regulatory clarity.

For financial institutions, the takeaway is clear: AI does not transform the competitive landscape on its own, but through effective integration into existing systems and processes. Firms that use it as an infrastructure layer rather than a standalone tool are likely to benefit most.

Looking forward, Stadnyk expects steady, incremental improvements, instead of “dramatic” breakthroughs. 

More Must-Reads:

This AI Just Learned How to Use a Crypto Wallet

The First Regulated Tokenized Equity Trade Is Now Live in the U.S.

Exclusive: Crypto Trader Josh Rager Says Bitcoin’s 4-Year Cycle “Isn’t What It Used To Be”