Founder Stack

Cognition Labs built its early reputation on Devin, marketed as the first AI software engineer capable of autonomously completing multi-step coding tasks, from reading a bug report to writing, testing, and submitting a fix, rather than just autocompleting code inside an existing editor. Founded by Scott Wu, a competitive programmer who previously ran the recruiting platform Bloomberg-adjacent Adventive, alongside Steven Hao and Walden Yan, Cognition's pitch was always more ambitious than the code-completion tools like GitHub Copilot that preceded it: full task ownership rather than line-by-line suggestion. The company's most consequential move came in 2025, when it acquired Windsurf, the AI coding editor and IDE, in a deal struck in the aftermath of a collapsed acquisition attempt by OpenAI, which had reportedly discussed buying Windsurf at a valuation near $3 billion before the deal fell apart over terms. Cognition's acquisition, completed rapidly and reportedly on more favorable terms given Windsurf's weakened negotiating position after the OpenAI deal collapse, gave Cognition both an existing IDE user base and engineering talent that significantly expanded its product surface beyond Devin's standalone agent product into a full development environment. That combination pushed Cognition's valuation to a reported $10.2 billion in its 2025 funding round, positioning it alongside Cursor-maker Anysphere as one of the two dominant AI-native coding companies, though the two have pursued different strategies: Anysphere's Cursor emphasizes human-in-the-loop editing assistance, while Cognition has pushed harder toward autonomous task completion with less continuous human supervision. Enterprise rollout has focused on engineering organizations willing to let Devin operate on well-scoped, lower-risk tasks like dependency upgrades, test writing, and documentation, while keeping humans in the loop for architecture-level decisions. The unresolved question across the entire AI coding agent category is trust calibration: engineering leaders want productivity gains from autonomous agents but remain wary of agents making consequential changes without adequate review, and several high-profile incidents across the industry involving agents deleting data or introducing subtle bugs have made enterprise buyers more cautious about scope than the marketing of 'autonomous software engineer' products originally suggested. Cognition has responded by building more granular permissioning and audit trails into Devin's enterprise deployment rather than pushing purely on autonomy as a selling point. What to watch: whether the Windsurf integration meaningfully grows Cognition's paid seat count relative to Cursor's, whether enterprise customers expand the scope of tasks they trust Devin to complete autonomously through 2026, and whether OpenAI or another major lab makes a renewed push into the coding-agent IDE category now that its Windsurf deal fell through.