Cognition
Cognition (also known as Cognition AI or Cognition Labs) is an applied AI lab based in San Francisco, founded in November 2023 by Scott Wu, Steven Hao, and Walden Yan, three competitive programmers who between them had won multiple International Olympiad in Informatics gold medals. The company is best known for Devin, marketed as the first AI software engineer capable of autonomously planning and executing coding tasks. Cognition has since raised hundreds of millions of dollars at a multi-billion-dollar valuation.
X / cognitionNotable people
People map
Current peopleNotable public-facing people we cover. Grouped by public roles, not reporting lines; this is not a complete staff directory.
Leadership
- Scott Wu
Co-founder and CEO
Product & engineering
- Jared Palmer
VP of Engineering
Other public roles
- Walden Yan
Team member
- Shawn Wang (swyx)
Advisor
Writing, research & talks
Cognition adds Gemini 4 Argon to Devin Desktop and CLI at 75% off through October 27. On its proprietary FrontierCode 1.1 it scores 60.5%, the top Gemini result, with strengths in backend bug fixes and refactors.
Cognition's FrontierCode 1.1 benchmark, which scores coding models on whether open-source maintainers would merge their PRs, now has a leaderboard on Artificial Analysis. The evaluation is private and was updated after its June 2026 debut.
Cognition added Claude Haiku 5.5 to Devin, citing 58.4% on FrontierCode 1.1 at roughly one eighth the cost per task of Sonnet 5. It also positions the model as a sidekick in Fusion, with Opus 5.5 as Lead in Devin Desktop and CLI.
- Agent Memory Repo | Cognitioncognition.com
Cognition's open spec stores agent memory in a git repo with a short MEMORY.md index, sourced single-line entries and wiki-style links. A scheduled Dreaming agent adds patterns and prunes duplicates, and repos compose across users.
Cognition added GPT-6.1 Sol to Devin, reporting 60.4% on FrontierCode 1.1 versus 60.7% for GPT-6 Sol, at $0.31 per task on medium effort. At low effort it scores 58.1% for $0.21, the top result under $0.30 per task.
Cognition adds Claude Sonnet 5.5 to Devin Desktop and CLI, reporting 64.4% on FrontierCode 1.1 Main versus 56.2% for Sonnet 5 and ahead of Fable 5.1 at extra high effort. Anthropic's token prices are unchanged.
Cognition added GPT-6 Sol and Luna to Devin. On FrontierCode 1.1, Sol matches GPT-5.6 Sol's score at 61% lower cost per task, and Luna beats GPT-5.6 Luna at about a quarter of the cost, under $0.10 per task.
Cognition added Claude Opus 5.5 to Devin, where it now leads Cognition's FrontierCode 1.1 benchmark for mergeable, real-world engineering tasks at 65.3% on the Extended split, at a lower cost than the prior leader.
Cognition added Grok 4.7 to Devin, reporting 59.4% on FrontierCode 1.1 Extended. It's strong on hard backend tasks but tends to over-scope, trailing Grok 4.6 in aggregate.
Cognition added a voice interface to Devin, letting you dictate coding tasks by phone, built on GPT-Live. Alongside it, the company released SWE-2, a model it claims matches recent frontier coding performance at up to 70% lower inference cost.
Dioxus Labs, maker of the open-source Rust UI framework, has joined Cognition to work on Devin's VM, computer use, and testing infrastructure. Cognition already relies on Dioxus internally, folding a key dependency's creators directly into the agent's engineering team.
Cognition's SWE-2, post-trained from Kimi K3, applies a single RL run with per-effort cost penalties to lift its whole cost-performance frontier, matching Fable 5.1 coding scores while taking 58% fewer turns and costing far less.
Cognition's SWE-2 coding model, post-trained on Kimi-K3, adds adjustable reasoning effort and ships in Devin Desktop and CLI, with vendor benchmarks claiming frontier-level FrontierCode scores at a fraction of SWE-1.7's cost.