Devin Developer Cognition's 18-month ARR reaches $445 million, with founder Scott Wu stating that their mission is to teach AI to write code

AIMPACT News, May 11 (UTC+8), according to Beating Monitoring, Colossus Magazine published a three-hour exclusive interview with Scott Wu, founder of Cognition. Cognition is the developer of the AI programming agent Devin, founded in November 2023, which reached an annualized revenue of $445 million within 18 months, with usage doubling every 8 weeks, currently raising funds at an estimated valuation of about $25 billion. Clients include the U.S. Army, Goldman Sachs, and Mercedes-Benz. Wu’s entrepreneurial timeline is extremely tight: on October 6, 2023, his mother passed away from lung cancer, after he had already been in Louisiana caring for her for over a year. A month later, on November 17, Sam Altman was dismissed by the OpenAI board. Wu was in New York chatting with friends about startup ideas that afternoon, and after seeing the news, he immediately decided to take action, canceling his original Thanksgiving plans with his father, and flying to the Bay Area to organize a hackathon. Two weeks later, Cognition took shape. Wu is a three-time gold medalist at the International Olympiad in Informatics (IOI), recognized as the strongest competitive programmer in American history. His IOI teammates include Scale AI founder Alexandr Wang, Perplexity co-founder Johnny Ho, and Decagon founder Jesse Zhang, whose combined company valuations total about $80 billion. Cognition co-founder Walden Yan once said, “The gap between me and Scott is like the gap between me and those who aren’t qualified to compete.” The interview also revealed several details: Gennady Korotkevich, widely regarded as the strongest competitive programmer in history, is currently working at Cognition; Wu challenged investor Peter Thiel to a game of chess and cards to negotiate better funding terms; Devin now has a significant proportion of conversations initiated not by humans but triggered automatically, with the agent discovering tasks and then fixing them. In terms of technological progress, Devin’s score on SWE-Bench increased from 13% at its debut in March 2024 to about 90% on the original version and about 80% on SWE-Bench Pro. Wu candidly admitted in the interview that over the past two years, there have been no major breakthroughs in AI research; the real driver of explosive usage has been large-scale training combined with a large amount of practical engineering tailored to real-world scenarios. Regarding the future of AI programming, Wu cited research from METR: the duration that AI agents can work independently has grown from seconds to hours, and this doubling occurs every few months. He believes this trend will continue, ultimately enabling agents to independently complete tasks spanning days, weeks, or even years. He summarized his mission in one sentence: “If I was born to do one thing, it’s to teach AI to write code.” (Source: BlockBeats)