Cognition launches Devin Bots to manage AI coding teams
Devin Bots shift AI coding from completing tickets to managing entire projects without human babysitting.
Tools · Source: Cognition Blog
What happened
Cognition just launched Devin Bots on November 17, 2026. The startup, now valued at $48 billion, gave its AI software engineer a boss. A Bot is a persistent agent with its own memory and tools that owns a goal end-to-end. It breaks large projects into pieces and delegates them to multiple Devin instances running in parallel. You hand it an outcome, not a ticket.
The Bots retain context across sessions and run for days or weeks. They investigate problems, evaluate results, and decide what to pursue next. They only bother human engineers for decisions that require actual judgment. Will Laughlin, a VP of Engineering, noted their Bots merged about 1,600 pull requests in a single week. They handle everything from planning to fixing regressions after a deploy.
Cognition tested this internally by building an entire Android app from scratch. An engineer named Samyok created an Android Bot and gave it a single prompt to manage the product. Over 44 days, the Bot filed tickets, managed Devin sessions, and merged 1,134 pull requests. Samyok only sent 183 messages during that time, mostly for security reviews or high-level direction. On 22 of those days, he sent zero messages.
Key facts
- 44 days — Time it took an Android Bot to build Devin's Android app
- 1,134 — Pull requests merged by the Android Bot
- $48 billion — Cognition's valuation after a September 2026 funding round
- 1,900 — Weekly pull requests achieved by one game studio using Bots
- $900 million to $1 billion — Estimated annualized revenue run-rate for Cognition
Why it matters
The bottleneck in software development is shifting from writing code to managing context. Individual AI agents are great at writing functions, but they need constant human supervision. Devin Bots automate the technical lead role. Founders and engineers can now hand off entire project outcomes instead of babysitting individual tasks. This dramatically scales the output of a small human team. One game studio using the Bots saw its weekly pull requests climb from roughly 50 to 1,900.
This launch also forces a showdown in the always-on agent market. OpenAI introduced Dots in September 2026, and Anthropic released Duet in October. Both are general-purpose agents that can handle code. Cognition is betting that a purpose-built manager for a specific coding agent will win. With an estimated revenue run-rate nearing $1 billion, enterprise customers are already paying a premium for that focus. They are using the technology to modernize legacy systems and manage security at scale.
For builders
Scale output without hiring more managers
One game studio saw weekly pull requests jump from roughly 50 to 1,900 using Bots. Small teams can ship massive projects with minimal oversight, but companies paying for bloated middle management will lose out.
Shift focus from coding to architecture
Bots handle the daily grind of filing tickets, running QA, and fixing regressions. Engineers must adapt to reviewing high-level decisions and security risks rather than writing boilerplate. The value is in setting the direction.
Automate tedious maintenance work
Cognition uses a Layout Bot just to catch UI shifts and lag. Builders can offload annoying, repetitive QA tasks entirely. This frees up human hours for core product features and keeps the application stable.
My take
I have always said the real value of AI is not autocomplete, it is autonomy. Giving an AI a ticket is a parlor trick, but giving it a goal and getting a finished product is a real business. Cognition is proving that the future of engineering belongs to those who know how to manage agents, not write syntax.
Original reporting: Cognition Blog. This is my rewrite and opinion.