In July 2026, Emergent Intelligence learned that Cognition, an AI enterprise co-founded by three Chinese gold medalists in the International Olympiad in Informatics (IOI), has achieved a valuation exceeding $26 billion. Its self-developed product, Devin—the world’s first autonomous AI software engineer—has been disrupting the global developer ecosystem since its launch in March 2024, marking the beginning of autonomous agents in software engineering.

Founded in 2023, this startup has built a technological moat with top-tier competition talent, secured early investment from Peter Thiel’s Founders Fund, and proven the commercial viability of autonomous programming AI with real-world data. This has also prompted the market to reevaluate the core value of algorithm competition talent in the AI industry.

Unlike traditional AI tools that only assist in code completion, Devin can independently complete development tasks end-to-end. Users simply provide requirements in natural language, and Devin autonomously breaks down tasks, sets up environments, writes code, debugs issues, and delivers complete projects. This fully autonomous operational model represents an unprecedented innovation in the industry.

The SWE-bench dataset, an authoritative benchmark for measuring AI’s true programming capabilities, is sourced entirely from real-world bugs in open-source projects. Previously, all AI models solved an average of less than 2% of issues independently. Devin achieved a 13.86% closed-loop resolution rate in its first public test—nearly seven times the industry standard—creating a tangible technological gap. This data became the core basis for the capital market’s willingness to assign a multi-billion-dollar valuation.

01

How Does a Genius Team’s Technical Moat Transform Competition Strength into AI Advantage?

The foundational strength behind Devin’s technological breakthrough lies in Cognition’s nearly elite talent structure. All three founders have competed in the U.S. IOI national team, each holding highly prestigious competition medals. They respectively oversee core business lines: corporate management, underlying technology, and product implementation, forming a complete capability closed loop (closed loop).

Steven Hao (left), Scott Wu (center), Walden Yan (right)

CEO Scott Wu holds three IOI gold medals, achieving a perfect global score in 2014. His early experience of dropping out to start a business has given him both top-tier algorithmic skills and a commercial vision. CTO Steven Hao combines a background in computer science and mathematics from MIT with early engineering experience at Scale AI, a leading data AI company, where he led the development of the model’s underlying reasoning and engineering architecture. Product lead Walden Yan, the youngest member and a post-2000 IOI gold medalist, leveraged the Thiel Fellowship to take a leave of absence and start a business. Specializing in the intersection of cryptography and machine learning, he oversees product implementation.

Beyond the trio, the initial startup team of over ten members collectively holds ten IOI gold medals. Scott Wu’s brother, Neal Wu—who also holds three IOI gold medals—joined the team. From the highest global rankings in Codeforces to placements in the ICPC World Finals, this team consistently ranks among the top echelons in global programming competitions. Their long-term, high-intensity algorithmic training has honed exceptional logical decomposition and complex problem-solving abilities—precisely the core technical requirements for autonomous AI agents.

Emergent Intelligence believes that while general large models rely on massive text data for basic generation, autonomous engineering agents require sustained, multi-step, long-chain reasoning. The complex problem-decomposition thinking cultivated by IOI competitors naturally aligns with Devin’s underlying architecture for multi-round autonomous decision-making—a unique barrier difficult for ordinary AI teams to replicate.

In today’s homogenized AI competition, most startups choose to fine-tune general large models for lightweight applications, making it hard to establish long-term barriers. Cognition has taken the opposite path, treating top algorithmic talent as its core asset and using the foundational logical capabilities honed through competitions to build a proprietary agent system. By avoiding the crowded field of general models, it has established a differentiated advantage in the vertical software engineering track (sector).

02

What Profound Changes Will Autonomous Programming AI Bring to the Software Industry?

Devin’s implementation doesn’t simply replace programmers but restructures the division of labor across the entire software development process. Previously, developers spent significant time on mechanical tasks like environment configuration, basic bug fixes, and repetitive feature development. Now, such standardized tasks can be autonomously handled by AI, allowing human engineers to focus on high-value work like architectural design, business logic innovation, and complex system decision-making. Both software development cycles and labor costs for enterprises will be significantly optimized.

The capital market’s $26 billion valuation essentially bets on software engineering becoming the first industry scenario for large-scale implementation of AI autonomous agents. Traditionally, AI tools served only as auxiliary plugins embedded in development tools, incapable of independently completing full projects. Devin has proven that AI labor with independent thinking and autonomous iteration capabilities holds commercial value. This will drive more enterprises to pay for vertical-domain autonomous agents, accelerating AI’s transition from conversational tools to an era of intelligent agents capable of independently executing complex tasks.

Emergent Intelligence argues that the event’s more noteworthy implicit (implicit) value lies in bridging information science competition talent with cutting-edge AI industries. For too long, the public has viewed IOI competitions merely as academic bonus point (bonuses) for student admissions, overlooking how the logical, reasoning, and system-building skills honed through competition training are precisely the core needs for developing next-generation autonomous agents.

Globally, a growing number of IOI gold medalists are entering AI startups and large model R&D sectors. From Silicon Valley startups to leading AI labs in China, competition talent is becoming the driving force behind technological breakthroughs. Cognition’s valuation myth will prompt the industry to reevaluate the long-term supply value of algorithmic foundational education for the AI industry.

At the same time, we must objectively recognize Devin’s current limitations. Nearly 86% of real-world engineering problems remain unsolvable independently. Scenarios involving complex business architectures, cross-system collaboration, and production-level risk management still heavily rely on human engineers. Autonomous AI programming will not cause large-scale industry unemployment in the short term but will drive developers to upgrade their skills. Overall industry talent demand will shift toward higher-level engineers capable of AI collaboration and complex system design.

03

Conclusion

The rise of Cognition and Devin represents not just a commercial success for a genius startup but a significant signal for technological and industrial development. Autonomous agents are moving from concept to implementation, with vertical industry-specific AI labor becoming a new focus for both capital and technical teams. Meanwhile, technologists with deep foundational algorithm expertise and hardcore competition training backgrounds will continue to shape the next wave of AI technological iteration. A new era of human-machine collaboration in the software industry has begun, and the bidirectional empowerment between algorithm competition talent and autonomous AI agents will continue to influence the global AI industry’s trajectory for years to come.