Devin, the AI software engineer, just got its 2025 performance review! Over the past eighteen months, thousands of companies have hired Devin, including Goldman Sachs, Citi, Santander, and Nubank. Using real-world examples and metrics from customers, we looked at where Devin shines and where it struggles. You might find this interesting if you’re thinking about deploying coding agents in the real world. Link below:
About us
Makers of Devin, the first AI software engineer. We are an applied AI lab building end-to-end software agents. We’re building collaborative AI teammates that enable engineers to focus on more interesting problems and empower engineering teams to strive for more ambitious goals.
- Website
-
https://www.cognition.ai
External link for Cognition
- Industry
- Software Development
- Company size
- 51-200 employees
- Type
- Privately Held
Employees at Cognition
Updates
-
Cognition reposted this
Another example: A squad across Ops / HW / Product used DevinAI/Cognition to build a ticket analyzer that reads customer tickets, spots patterns, and suggests fixes automatically. Before: engineers spent 10-20 hours a week doing this manually Now: 20 minutes a day reviewing and training the workflow
We automated 35 hrs/week of busywork with AI. Now teaching other teams our exact model. No engineers needed. Bootcamp format. Example below 👇 A senior finance leader used to spend 4h analyzing burn vs our forecast. A rote task that requires juggling multiple spreadsheets to analyze anomalies. Used Claude to build a Claude sub-agent to pull and organize all the data, create docs highlighting anomalies, and be able to ask deep dive questions on the dataset.
-
AHEAD just published a detailed look at how enterprises are using AHEAD + Cognition to apply AI directly to the SDLC. The focus is on real workflows with measurable baselines, not abstract “AI strategy.” Across joint customers, the pattern is consistent: start with work that already has cycle-time data (dependency and framework upgrades, regression testing, onboarding), run a narrow pilot where correctness and throughput can be measured, use agentic tools like Windsurf and Devin to automate the repetitive parts, then scale once governance and data boundaries are in place. The outcomes are equally consistent. Teams are cutting multi-week regression testing down to days, accelerating onboarding by auto-generating configs and scaffolding, clearing backlog items like SonarQube fixes and dependency bumps, and producing release notes and architecture docs tied directly to the codebase—all without pulling senior engineers off roadmap work. Read the full article here: https://lnkd.in/gRx-cSer
-
Every engineering team has a migration story: late nights, brittle builds, endless dependency errors. Moving from .NET Framework to .NET Core is one of the most notorious of them all. We wrote up a guide to automating .NET migrations using Devin: https://lnkd.in/gs_qckHD
-
Proud to be joining GitHub's new Agent HQ as launch partners. More choices and more ways to try Devin, coming later this year!
Welcome to your Agent HQ 📍Orchestrate any agent, any time, anywhere. Coding agents from Claude, OpenAI, Cognition, Google, xAI and more will become available in GitHub as part of your paid Copilot subscription. https://lnkd.in/g3q7sM7M
-
-
SonarQube is available in Devin's MCP Marketplace. Equip Devin with Sonar's security analysis and code review tools. Getting started with the MCP Marketplace: https://lnkd.in/dHwG7ReZ
Today, we're excited to announce the Sonar Integration Program🚀, a major expansion of our partner ecosystem designed to unify code governance across the entire software development lifecycle. With new integrations like Google Gemini, Atlassian Jira, JFrog, Windsurf Cognition, Jellyfish, and Port.io we're making it easier than ever for developers to write better, more secure code. This program provides a holistic ecosystem for orchestrating code quality and security, enabling development teams to embed automated code review and governance directly into their existing processes. Read the full press release to learn more 👉: https://bit.ly/4oBPttz #CodeQuality #CodeOrchestration #CodeSecurity #SonarQube #AI #TechPartners
-
-
Introducing SWE-grep and SWE-grep-mini: Cognition’s model family for fast agentic search at >2,800 TPS. Surface the right files to your coding agent 20x faster. Now rolling out gradually to Windsurf users via the Fast Context subagent – or try it in our new playground! Read more: https://lnkd.in/gxFEvJq5
-
Check out Russell's conversation with Goldman Sachs CIO Marco Argenti on AI in the workforce:
Founding President at Cognition, Russell Kaplan, sat down with Marco Argenti, CIO of Goldman Sachs, to discuss the compounding capability of language models, and what happens when AI transitions from a tool to an autonomous workforce. As JPMorganChase, Santander, Goldman Sachs, Wells Fargo, BlackRock, Citi, Bank of America, Morgan Stanley, PNC, and others report this week, a clear theme runs through their 10-Qs: AI spend is accelerating - deployed to drive real productivity gains and expand profits. Across marketing, treasury, operations, and customer service, banks are deploying AI to improve efficiency, reduce costs, and expand margins. It’s already showing up in earnings - AI is now a direct driver of both top-line growth and bottom-line leverage. For those leading technology or strategy in financial services, this conversation is more than a preview - it's a blueprint for whats to come. The winners in this next wave won’t just build faster, they’ll operate differently. They’ll learn to orchestrate humans, systems, and AI agents to move as one. The impact will be profound: shorter development cycles, lower technical debt, and entirely new operating models for financial institutions and industry alike. As Russell Kaplan put it, “Every individual engineer is becoming a tech-lead manager of a fleet of AI agents. The real skill now is breaking down complex projects and delegating them into units of work for AI.” Talks at GS – Russell Kaplan https://lnkd.in/eT-pAePq
-
Cognition reposted this
Two months ago, Paul Graham tweeted that one of the best engineers he knows was using AI tools to write 10,000 lines of code a day. Yesterday, I wrote 180,000. At Kaizen, we’ve spent the past few months optimizing our development process to work with dozens of coding agents (we love Cognition's Devin) in parellel. It’s changed how we build. Our engineers move faster, Kaizen automates more, and ship features at a pace customers are consistently astonished by. Customers often ask how Kaizen's browser automation platform is so feature-complete: we support batch execution, document processing, uploads and downloads, scheduling workflows, triggering workflows via email and API, TOTP and SMS authentication, and human-in-the-loop operations... all built out in a matter of months. This isn’t vibe coding. It’s architecture design, process optimization, and ruthless product management. We talk to customers constantly. We break down their requests into the smallest actionable tasks, then identify minimum cuts so work can be executed with the highest possible parallelism. From there, we execute these tasks one by one, designing workflows that automatically trigger new work as soon as their blocking tasks are completed. I’m writing a post detailing how process working with coding agents at Kaizen. To get this right we had to rewrite our entire product development process. Would folks be interested in reading? Comment "ship" and I'll send over an early copy.
-