Julia Chirkova
Richie Sater, a Certified Public Accountant (CPA) without a formal mathematics degree, has achieved a significant breakthrough by leveraging artificial intelligence to solve a complex mathematical problem that had remained unsolved since 1986. The challenge, known as Problem 10.34 from the Kourovka Notebook, a renowned collection of unsolved mathematical problems originating from the Soviet era, was tackled by Sater’s custom-built system of AI agents. This achievement, however, was met with initial skepticism from the established mathematical community, underscoring a critical challenge: how to validate and trust AI-generated technical work at scale. Sater’s experience offers a compelling case for the evolving role of CPAs in an era where AI is rapidly accelerating the pace of technical output, potentially outpacing the capacity for human expert review.
The resolution of Problem 10.34 marks a notable event in the realm of recreational mathematics, a field that has historically fostered intellectual curiosity and collaboration among mathematicians. The Kourovka Notebook, first compiled in 1963, has served as a crucible for mathematical inquiry, with many of its problems inspiring decades of research. Problem 10.34, posed by Viktor Monakhov, presented a particularly enduring enigma, eluding the efforts of numerous mathematicians over nearly four decades. Sater’s unexpected success, particularly given his non-traditional background in mathematics, initially prompted disbelief among experts.
Sater’s approach involved constructing a sophisticated system of AI agents, operating from his personal MacBook. This "small army" of AI workers was designed with specialized roles: one agent was tasked with scouring existing academic literature for relevant precedents and theorems, another focused on generating specific examples to test hypotheses, a third was programmed to identify potential flaws and counterexamples in the work of other agents, and a fourth managed the intricate task of verifying and citing sources. This multi-agent architecture, described by Sater as "messy and noisy and dangerously effective," allowed for a parallel and iterative approach to problem-solving that far surpassed traditional linear methods.
The journey to solving Problem 10.34 began with Sater’s proactive communication with Viktor Monakhov, the original problem poser. Upon informing Monakhov of his proposed solution, Sater was presented with a new challenge: another problem from the Kourovka Notebook, this time with the explicit intent of testing his AI system’s capabilities. Within a mere 12 hours, Sater’s system reportedly generated a counterexample that definitively answered this subsequent problem. This rapid turnaround time, coupled with Sater’s lack of formal credentials in advanced mathematics like group theory, naturally raised eyebrows within the mathematical community.
The skepticism was palpable. As Sater recounts, "We have significant doubts." This reaction is understandable within a scientific discipline that relies heavily on peer review and rigorous validation. The very notion of a non-specialist, utilizing AI, resolving a problem that had vexed generations of mathematicians was, by conventional standards, an extraordinary claim.
This is precisely where Sater’s professional background as a CPA became instrumental. The principles of auditing and assurance, which are at the core of accounting, provided him with a framework for responding to skepticism. In auditing, when a financial figure or claim appears exceptionally positive or unexpected, auditors do not simply accept it at face value. Instead, they engage in a process of tracing the origin of the data, rebuilding calculations, and designing specific tests to verify the accuracy and integrity of the information. Sater applied this same methodical, evidence-based approach to his AI-generated solution.
He developed what he terms an "audit system" for his mathematical findings. This system was designed to provide an explicit audit trail, meticulously documenting every step of the AI agents’ work. This included verifying every citation, generating standalone executable code that others could run to replicate the results, and conducting adversarial checks to probe for weaknesses. The goal was to furnish the reviewers with "verified evidence, not slop," making it exceptionally easy for them to test and trust the outcome. The effectiveness of this approach was confirmed when Monakhov and his collaborator, Irina Sokhor, responded, "The GAP file together with the output transcripts have convinced us." The System for Computational Discrete Algebra (GAP) is a widely used open-source computer algebra system, and the provision of its output alongside transcripts served as the verifiable "audit evidence" that satisfied the mathematicians.
The implications of Sater’s experience extend far beyond the realm of mathematics, pointing towards a fundamental shift in how technical work will be produced and validated across various industries. The exponential growth of AI’s capabilities means that the generation of raw output is no longer the primary bottleneck. Instead, the challenge lies in the capacity for expert review. As Sater articulates, "AI is producing more technical work than experts can possibly review, so every field is going to need a way to test that work at scale."
This is where the skills and principles inherent in the CPA profession become remarkably relevant. CPAs are trained to provide assurance over work they have not personally created. Their role involves meticulously examining the processes and evidence underlying financial statements, tax returns, and other critical business information, and then communicating a professional opinion on their trustworthiness. This ability to assess, verify, and vouch for the accuracy of complex information, even when generated by others, positions CPAs as uniquely qualified to navigate the challenges presented by AI-generated content.
Sater anticipates that CPAs will discover a natural aptitude for managing AI agents, akin to how they currently manage teams of human analysts. More significantly, he posits that "CPAs are going to be the ones who figure out the assurance layer for a world where output is no longer the constraint." The critical constraint will shift from production to verification. The sheer volume of AI-generated output—whether in the form of scientific research, financial models, legal documents, or engineering designs—will necessitate robust and scalable methods for ensuring its accuracy and reliability.
The core of the CPA’s value proposition lies in their established methodologies for establishing trust in complex systems and outputs. They have a long history of developing frameworks for risk assessment, internal controls, and evidence gathering. These foundational principles can be adapted and applied to the validation of AI-generated work, ensuring that the burgeoning capabilities of artificial intelligence are harnessed responsibly and reliably.
The publicly available preprint detailing Sater’s solution to Kourovka Notebook Problem 10.34 is accessible on arXiv under the identifier arXiv:2608.02970. The comprehensive verification materials, including the GAP files and output transcripts, have been made public on GitHub at github.com/RichieSater/kourovka-10-34, allowing for complete transparency and independent verification by the mathematical community and beyond. Furthermore, a more personal account of his experience is currently undergoing review for publication in the Mathematical Association of America’s esteemed Math Values section, with an anticipated publication date of September 10th. Sater is also scheduled to present the full narrative and the architectural details of his AI system at NC State University’s Algebra & Combinatorics Seminar on September 30th, further disseminating his findings and methodology.
The implications of this development are profound. In fields where accuracy and reliability are paramount, such as finance, law, medicine, and engineering, the ability to efficiently and effectively audit AI-generated outputs will become a critical differentiator. For CPAs, this presents an opportunity to expand their traditional role and become leaders in establishing the necessary assurance frameworks for the AI-driven future. The skills honed through decades of auditing and attestation—critical thinking, analytical rigor, evidence evaluation, and clear communication—are directly transferable and essential for ensuring the integrity of AI’s expanding contributions to knowledge and innovation.
The successful resolution of Problem 10.34 by Sater, a CPA wielding AI, serves as a powerful case study. It not only demonstrates the potential of AI as a problem-solving tool but also highlights the critical need for robust assurance mechanisms. As AI continues its rapid advancement, the demand for professionals who can provide confidence in AI-generated outputs will only grow. The accounting profession, with its inherent focus on trust and verification, appears exceptionally well-positioned to meet this burgeoning demand, redefining its role in an increasingly automated world. The challenge for the future will be to integrate these AI-driven solutions with human oversight and validation, ensuring that innovation progresses both rapidly and responsibly.
