The conventional wisdom is that AI’s defining companies are being built by seasoned engineers and researchers with decades of experience. The evidence out of San Francisco suggests otherwise. A cohort of founders in their mid-to-late twenties, some of them college dropouts, some of them competitive programming prodigies, some of them Berkeley PhD students who went viral on Hacker News before finishing their dissertations, are building companies that the largest technology corporations in the world are now scrambling to acquire, partner with, or outcompete. This is who they are and what they built.
Cursor: Four MIT Dropouts and a $60 Billion Deal
Michael Truell, Aman Sanger, Sualeh Asif, and Arvid Lunnemark left MIT in 2022 without finishing their degrees. Their first two ideas failed. Their third became Cursor, an AI-native code editor that reached $100 million in annualized revenue faster than Slack did in its early days, crossed $2 billion in ARR by February 2026, and is now used by more than half of the Fortune 500. Lunnemark departed in October 2025 to found Integrous Research, an AI safety lab, with the three remaining founders continuing to lead the company. In April 2026, SpaceX announced it had secured the right to acquire Anysphere, the company behind Cursor, for $60 billion before the end of the year, or pay $10 billion for the joint AI development work the two companies are undertaking together. Truell, the CEO, is 25 years old. Forbes estimates his stake in the company puts his net worth at approximately $1.3 billion. Nvidia CEO Jensen Huang has called Cursor his favorite enterprise AI service. The founders initially avoided the AI coding space because, as Truell told Y Combinator’s AI Startup School, they thought it was too competitive.
Cognition AI: The Competitive Programming Mafia
Scott Wu, Steven Hao, and Walden Yan share an unusual credential: all three won gold medals at the International Olympiad in Informatics, one of the most demanding competitive programming competitions in the world. Wu, born in 1997 in Louisiana, won three IOI gold medals and placed first overall in 2014 before dropping out of Harvard. Hao earned his IOI gold in 2014 and was one of the earliest engineers at Scale AI before co-founding Cognition. Yan, the youngest of the three, won IOI gold in 2020 and became a Thiel Fellow, dropping out of Harvard to focus on Cognition full time. Together they built Devin, billed as the world’s first fully autonomous AI software engineer, capable of writing, testing, and deploying software without constant human input. Devin’s annualized recurring revenue grew from $1 million in September 2024 to $73 million by June 2025. In July 2025, Cognition acquired Windsurf after Google poached Windsurf’s CEO and senior leadership. The company reached a $10.2 billion valuation in September 2025 and is currently in funding talks at a reported $25 billion valuation.
Scale AI: The Youngest Self-Made Billionaire
Alexandr Wang was born in January 1997 in Los Alamos, New Mexico, to two physicist parents who worked at the Los Alamos National Laboratory. He enrolled at MIT at 17, dropped out after his first year, and co-founded Scale AI in San Francisco in 2016 at age 19. Scale became the dominant provider of labeled training data for AI models, working with clients including OpenAI, Google, Microsoft, and the US Department of Defense. Wang became the world’s youngest self-made billionaire at 24, after a 2021 funding round valued Scale above $7 billion. In June 2025, Meta invested $14.3 billion in Scale for a 49% stake, pushing Scale’s valuation to $29 billion. As part of the deal, Wang stepped down as CEO and joined Meta to lead its new superintelligence team, reporting directly to Mark Zuckerberg. His estimated net worth at the time of the deal was $3.6 billion. He was 28.
Letta: Memory for the Agentic Web
Charles Packer and Sarah Wooders met as PhD students in UC Berkeley’s Sky Computing Lab, the research environment that also produced Apache Spark, Databricks, and vLLM. They built MemGPT, an open-source system designed to give large language models persistent long-term memory, and went viral on Hacker News before they had even finished writing the paper. They commercialized the project into Letta, a platform that lets developers build AI agents with memory that compounds over time, emerging from stealth in September 2024 with $10 million in seed funding at a $70 million valuation. Backers include Google’s Jeff Dean, Hugging Face’s Clem Delangue, and Runway’s Cristóbal Valenzuela. Clients include Accenture, Nokia, Bain, and Tufts University. Both founders were named to the 2026 Forbes 30 Under 30 list. Wooders has described Letta’s vision as one where developers eventually interface with agents as services rather than models, making memory and state management the foundational layer of the AI stack.
DayDream Ventures: Betting on the Next Generation
Jonathan Chang is not building an AI model. He is building the infrastructure around the people who will. Born and raised in San Francisco and a UCLA graduate, Chang founded DayDream Ventures as a pre-seed fund backing early-stage founders who are Gen Z or building for Gen Z audiences. He also founded SF IRL, a curated weekly newsletter covering the Bay Area’s startup and venture events, and DayDream Fellows, formerly GenZScouts, a program now in its fourth year that has helped students from non-target universities break into venture capital and startups. To date the program has produced more than 15 venture-backed founders, more than 30 full-time investors, and more than 70 operators in tech. Chang has been named one of Business Insider’s top Gen Z VCs to watch. His bet is not on any single company. It is that the founders who will define the next decade of AI are mostly not yet in the room, and that getting them there before anyone else does is itself a competitive advantage.
