Devin-kun in Japan: AI agents accelerate legacy code modernization amid a shrinking software workforce

Japan has become an unexpected proving ground for AI-powered software engineering, with Cognition AI's Devin agent earning the local nickname "Devin-kun" as it tackles legacy code modernization in a market facing a severe software engineer shortage. For AI builders, the Devin-kun story offers a real-world case study in deploying AI agents in aging infrastructure markets with talent gaps, and highlights the operational advantages of distributed compute across time zones.

What happened

Cognition AI, the San Francisco startup behind the AI coding tool Devin, opened a Tokyo office in April as the first step in its Asian expansion. The company plans to make Singapore its Asia-Pacific headquarters later this year. Japan was already one of Devin's most popular countries by user engagement before the official launch, according to Cognition president Russell Kaplan.

The timing aligns with a structural need. Japan's Ministry of Economy, Trade, and Industry (METI) estimated in 2023 that the country would face a shortage of 789,000 software engineers by 2030. With nearly 30% of residents over 65 and a working-age population projected to decline by over 30% by 2060, the talent gap is acute. Kaplan described the situation as "aging infrastructure with a declining workforce."

A concrete example: Sapporo's city government needed to modernize over one million lines of legacy code to meet a national IT compliance mandate. Kaplan estimated the work would normally take 200 engineering months. Using Devin, Sapporo's engineers completed it in roughly a quarter of that time.

Cognition AI raised over $1 billion in a funding round in late May, valuing the startup at $26 billion. The company's annualized run rate reached $492 million, up from $37 million a year earlier. The company also opened a Tokyo office in April and plans to establish Singapore as its Asia-Pacific headquarters later this year.

Why AI builders should care

Japan's adoption of Devin-kun is not a niche experiment. It reflects a structural shift: a country with a projected shortage of 789,000 software engineers by 2030 is turning to AI agents as a primary tool for maintaining and modernizing critical infrastructure. For builders shipping AI coding tools, this validates the thesis that agent-based software engineering can scale in markets where human talent is scarce and legacy systems are deep.

Devin operates as a full AI software engineering teammate. It codes, debugs, and deploys autonomously inside existing tools like Slack. Japanese engineers can work with Devin entirely in Japanese, which Kaplan noted helps overcome the language barrier that has isolated some Japanese businesses. This multilingual capability is a practical differentiator for any AI agent targeting non-English markets.

Cognition AI's growth metrics are striking. The company raised over $1 billion in a funding round in late May, valuing it at $26 billion. Its annualized run rate hit $492 million, up from $37 million a year earlier. Compute demand is doubling roughly every seven weeks, and the company is using geographically distributed teams to run workloads during off-peak hours across time zones.

Why AI builders should care

For teams building AI coding agents or planning to deploy them in enterprise settings, Japan's adoption pattern offers several takeaways.

First, the market need is structural, not cyclical. Japan's METI-estimated shortage of 789,000 software engineers by 2030 means AI agents are not replacing workers but filling a gap that cannot be filled through traditional hiring. This is the same dynamic that will play out in other aging economies across Europe and parts of Asia.

Second, Devin's multilingual support matters. Japanese engineers can work with Devin entirely in Japanese while collaborating through the agent with teams globally. For any AI agent targeting non-English markets, native language support is a practical requirement, not a nice-to-have.

Third, compute efficiency through time-zone distribution is a real operational lever. Kaplan noted that demand for compute at Cognition is doubling roughly every seven weeks, but geographically diverse teams allow workloads to run during off-peak hours in different regions. This is a pattern any AI company with global teams can replicate.

Practical implications

The Sapporo project is the most concrete example. Modernizing over one million lines of legacy code in roughly a quarter of the traditional effort (from an estimated 200 engineering months to about 50) shows the potential for AI agents in government and enterprise IT modernization. For builders, this validates the use case of AI agents for legacy code refactoring, a pain point in many large organizations.

Cognition's expansion strategy also offers lessons. The company opened a Tokyo office, plans Singapore as its Asia-Pacific headquarters, and has launched an Applied AI Engineering program in Malaysia to train engineers in directing AI agents. This suggests that successful AI agent deployment requires not just the tool but also local teams skilled in managing agents.

Compute efficiency is another takeaway. With demand doubling roughly every seven weeks, Cognition uses distributed teams to run workloads during off-peak hours in different regions. This is a practical pattern for any AI company with global operations.

Caveats

The details in this article come primarily from a Fortune report based on company claims and interviews with Cognition AI president Russell Kaplan. The Sapporo legacy code modernization numbers are Kaplan's estimates, not independently verified. The Mythos model access via Project Glasswing was halted in mid-June after the U.S. barred foreigners from using it, highlighting geopolitical risks in AI deployment.

Cognition AI's rapid growth and high valuation ($26 billion) reflect investor enthusiasm, but the company's ability to sustain its run rate and expand into new markets like Malaysia and South Korea remains to be seen. The impact on traditional software engineering roles, particularly in India, is a subject of investor concern, though Kaplan expressed optimism about role evolution.

FAQs

What is Devin-kun and how does it work?

Devin-kun is the nickname Japanese developers gave to Devin, an AI coding tool from Cognition AI. Devin operates as a full AI software engineering teammate: give it a task, and it codes, debugs, and deploys code autonomously inside the tools an engineering team already uses, such as Slack. Japanese engineers can work with Devin entirely in Japanese.

How can Devin help modernize legacy code in government or city systems?

Devin can accelerate modernization of large legacy codebases. In Sapporo, the city government needed to modernize over one million lines of legacy code to meet a national IT compliance mandate. Kaplan estimated the work would normally take 200 engineering months, but using Devin, Sapporo's engineers completed it in roughly a quarter of that time.

Why is Japan a strategic market for AI software engineers and Devin-kun?

Japan has the world's oldest population and a projected shortage of 789,000 software engineers by 2030. The country's critical infrastructure and government systems run on aging code, and the declining workforce makes AI-assisted software engineering a practical necessity rather than an experiment. Japan has also become a preferred beachhead for U.S. AI companies, with OpenAI and Anthropic opening Tokyo offices and SoftBank investing heavily in OpenAI.

What is Cognition AI's funding and valuation?

Cognition AI raised over $1 billion in a funding round in late May 2026, valuing the company at $26 billion. The company's annualized run rate reached $492 million at the time of the raise, up from $37 million a year earlier.