Cognition
Devin, the AI software engineer. The first real agentic coding product.
cognition.ai ↗Cognition is the best-positioned agent-first company in the market. Devin is the only autonomous dev product whose demos hold up under scrutiny, and the Windsurf acquisition was the most strategically interesting M&A move in AI last year — it bought Cognition enterprise distribution at a price that will look cheap in retrospect. The execution question now is whether the team can operationalize a high-touch enterprise motion on top of the research intensity that built Devin.
What's going for them.
- 01Devin is the first agentic coding product that a non-technical observer can watch for five minutes and understand — it plans, writes, runs, debugs, and reports. The demo does real work.
- 02The Windsurf acquisition (July 2025) — completed days after Google raided Windsurf's CEO — gave Cognition an IDE surface, an enterprise pipeline, and 30%+ instant ARR uplift. Unusual opportunism, well executed.
- 03The founding team (three former International Olympiad in Informatics medalists) is unusually deep on algorithms — which shows in how Devin handles multi-step planning versus competitors' flatter agents.
- 04$400M round in September 2025 at $10.2B gave the company the balance sheet to out-execute on agent R&D while Cursor and Replit focus on the editor surface.
- 05Positioned one layer above the code-editor wars: if agents start doing a material fraction of end-to-end software work, Devin is the reference product.
What they built
Devin is an autonomous software engineer — a product that takes a task description, plans a sequence of steps, executes them in a sandboxed environment, debugs its own failures, and reports back with a completed pull request. Cognition has built the planning, execution, and evaluation infrastructure around Devin to handle end-to-end tasks that previously required a human engineer: code migrations, bug triage, feature builds, integration work. The Windsurf acquisition in July 2025 added a full AI-native IDE, which Cognition now ships as the front-end to Devin for users who prefer an editor-centric workflow.
How they got here
Scott Wu, Steven Hao, and Walden Yan — all former IOI medalists, all with backgrounds in competitive programming and research engineering — started Cognition in late 2023 with an explicit agent-first thesis. The company stayed quiet for six months, then launched Devin in March 2024 with a demo that went viral and set the benchmark for what agentic coding was supposed to look like.
The first year was uneven. Devin’s early releases impressed on benchmarks but struggled on messy real-world codebases. Through 2024 and early 2025 the team rebuilt the planning and tool-use infrastructure, and by mid-2025 Devin’s ARR grew from $1M (Sep 2024) to $73M (June 2025). Then in July 2025 — two days after Google hired away Windsurf’s CEO — Cognition acquired Windsurf outright, absorbing the team, the IDE product, and an existing enterprise customer base. The September 2025 $400M round at $10.2B followed.
What’s ahead
Three things will determine whether Cognition becomes a generational company or a category leader that gets compressed by the frontier labs. First, reliability: Devin’s planning is good enough to be useful but not yet good enough to be trusted on high-stakes work without supervision. Second, enterprise: Windsurf gave Cognition a pipeline, but the company has not yet proven it can run the multi-quarter enterprise sales cycles that the category requires. Third, pricing: autonomous agents are the first AI product category where per-task pricing makes more sense than per-seat, and Cognition has a chance to define the pricing architecture for the whole space.
Why it matters
Cognition is the cleanest investable bet on AI agents actually working — not as marketing copy but as revenue. If agentic software engineering becomes a real category over the next two years, Devin is the product that will be cited as the one that made it possible. For founders building in the agent space, the Cognition playbook — deep research, slow launch, aggressive M&A at the right moment — is the reference case for how to compete when the frontier labs could move into your space any quarter.
Founder interview coming soon.
We'll be sitting down with the founders and operators of the companies we profile — on fundraising, product decisions, and what they're building next. If you're part of the Cognition team and want to share a perspective, get in touch.
Thinking about fundraising or M&A?
Amafi Advisory works with AI companies on strategic, fundraising, M&A, and technical advisory. Even if you're just exploring.