Midas, Led by World’s Leading Mathematicians, Unveils Platform to Transform AI from Plausible to Provable
Midas, a new company focused on bringing mathematical certainty to artificial intelligence, has publicly launched following a $10 million funding round led by Nova Global. Backed by investors behind companies such as OpenAI, Tesla, and SpaceX, the startup is positioning itself at the center of a new approach to AI—one that replaces probabilistic outputs with mathematically verified results.
The company’s launch reflects growing concern across industries about the reliability of modern AI systems. While today’s models are capable of producing fluent, confident, and highly persuasive outputs, they still lack a fundamental element that underpins most human systems of knowledge: proof. Midas aims to address that gap by building infrastructure that mathematically verifies AI outputs, data, and reasoning before they are used in real-world applications.
A Team Built on Mathematical Excellence
At the core of Midas is an unusually accomplished founding team. The company was formed by 11 medalists from the International Mathematical Olympiad (IMO) and the International Olympiad in Informatics (IOI), two of the world’s most selective academic competitions. Each country sends only a handful of participants to these events, making medalists part of an extremely small global talent pool.
Beyond their competition credentials, team members bring experience from major technology and finance companies, including Jane Street, Google, AWS, NVIDIA, and Mercor. Their academic backgrounds span institutions such as Stanford, MIT, Cambridge, Princeton, and Duke.
Together, the group is applying formal mathematical verification—a technique used in fields like aerospace, cryptography, and hardware design—to one of the most pressing challenges in AI: trust.
From Plausibility to Proof
Modern AI systems are built on probabilistic models. They generate outputs based on patterns learned from vast datasets, but they typically cannot prove that those outputs are correct. As a result, the same systems that can produce brilliant insights can also generate subtle or catastrophic errors.
Modern AI produces fluent, convincing answers, but it cannot prove they are correct,” said Shalim Monteagudo-Contreras, President and Co-Founder of Midas. “Midas is building the barrier between probabilistic outputs and real-world systems. We enforce correctness mathematically, so results are not inferred, argued, or hoped for, but proven before they are allowed through.”
The company’s philosophy is grounded in a simple distinction: fluency is not auditable, but proof is. While language models may appear confident, that confidence is not evidence of correctness. Midas aims to make proof the foundation of AI systems, rather than an afterthought.
Renzo Balcazar, CEO and Co-Founder, described the problem in broader terms. “Every human institution, from law to science to finance, runs on evidence. Artificial intelligence is the first form of intelligence that operates without it,” he said.
According to the company, AI is the first intelligence deployed at scale without a “proof loop”—a system that ensures outputs are supported by verifiable reasoning. Instead, current systems often deliver answers without explanation, confidence without causality, and results without evidence.
The Limits of Plausible Machines
As AI systems become faster and more capable, the gap between output and verification is widening. Machines can generate information far more quickly than humans can evaluate it, creating a world in which plausible answers may be mistaken for correct ones.
In this environment, coherence is often confused with correctness. Fluent language replaces evidence. Confidence replaces truth.
Midas argues that this trend is unsustainable. Plausibility scales easily with modern machine learning techniques, but proof does not—unless it is embedded directly into the system.
Rodrigo Porto, Tech Lead at Midas, explained that the company’s approach focuses on verifying reasoning from the start, rather than checking for errors after the fact. “When systems become too complex for manual review, trust can only come from built-in verification,” he said. “We introduce mathematical evidence at the core of AI, verifying outputs, data, and reasoning where mistakes are not an option.”
Building Infrastructure for High-Stakes Domains
The $10 million funding round will allow Midas to move formal verification techniques out of academic research and into production-grade infrastructure. The company is already targeting sectors where correctness is critical, including:
- Biotech and life sciences
- Defense systems
- Hardware and chip design
- Financial infrastructure
- Core AI and cloud platforms
In these environments, a small error can have massive consequences. According to the company, correctness is not a competitive advantage in such sectors—it is the minimum requirement.
Midas describes its platform not as a typical software product, but as a structural layer designed to sit beneath AI systems. Rather than iterating through product cycles, the company says it is building foundational infrastructure intended to reshape how AI is deployed in critical applications.
Investor Confidence and Long-Term Vision
The company’s backers include investors associated with OpenAI, Tesla, and SpaceX, organizations known for pushing technological boundaries. Nova Global, which led the funding round, says its investment reflects both the technical ambition of the project and the strength of the founding team.
At Nova Global, we focus on backing founders with the potential to become historical figures,” said Carlo Agostinelli, founder of Nova Global. “Shalim Monteagudo-Contreras and Renzo Balcazar are already operating at that level. They’ve built a world-class team from scratch and are taking on one of the most fundamental challenges in AI: trust. Their proof-native approach to ensuring AI reliability demonstrates the technical ambition and founder-market fit that can turn Midas into a generational company.”
Toward a Proof-Native AI Future
As artificial intelligence continues to move into safety-critical and high-value domains, the question of trust is becoming central. For decades, software verification has been used to guarantee correctness in systems where failure is unacceptable. Midas is attempting to bring that same rigor to AI.
The company’s vision is an AI ecosystem where every output can be backed by mathematical evidence. In such a system, results are not merely plausible—they are provably correct.
If successful, Midas could help shift AI from a technology defined by probability and persuasion to one grounded in certainty and verification, marking a significant step toward more reliable and accountable intelligent systems.
About Midas
Midas is building the verification layer for AI — mathematical trust infrastructure that uses formal verification to ensure provable correctness of AI outputs and training data. Founded by a team of 10 IMO/IOI medalists from Cambridge, MIT, Princeton, Duke, and Stanford, alongside senior engineers from leading technology companies, Midas applies formal mathematics to enable enterprise AI deployment across mission-critical sectors including biotech, defense, hardware design, and finance. The company is backed by Nova Global, and additional tier-one investors.
