Scott Wu

At Cognition, one of our values is to go for it all: when faced with a tradeoff, pick the ambition-maximizing direction. When we launched Devin in 2024, we envisioned a future where every team had an infinite army of junior engineers. We were early and Devin wasn't good enough. Now it is! The only constraint left is how ambitious you're willing to be.

linkedin.com/posts/scott-wu-8b94ab96_at-cognition-one-of-our-values-is-to-go-activity-7495907619568517120-I9FyAug 19, 2026

The best part of being an engineer is only about 10% of the job. Historically, software engineering has been roughly 10% fun stuff, like solving tough architecture problems, making trade-offs, and thinking about what to actually build. The remaining 90% of it is routine execution. The way we’ve always thought about Devin, the AI software engineer, is we want it to be the tool allowing human engineers to do 10x more of the 10%. Today, teams around the world are using Devin for tasks ranging from bug fixes to building out new features to handling complex migrations and modernizations. Instead of being the person who hand-writes every line, you’re now able to deploy an army of junior software engineers with Devin. You launch multiple Devins, delegate the execution, and run them in parallel on each piece of what you're building. The routine 90% gets handled so you spend more of your time on the parts that matter most. The ceiling on what one engineer can build goes way up, and the most human part of the work (taste and deciding what's worth building) becomes more of the job instead of a small fraction of it.

linkedin.com/posts/scott-wu-8b94ab96_the-best-part-of-being-an-engineer-is-only-activity-7483197449512058881-4kdLJul 15, 2026

A few months ago, a malicious actor initiated a sophisticated attack on a widely used open-source library with over 100 million weekly downloads. Within minutes, Devin (our AI software engineer) identified the attack. For context, most modern software is built out of shared, free building blocks - bits of code thousands of companies reuse instead of writing everything themselves. It's a massive time-saver, but someone can also slip harmful code into a popular building block and spread it to thousands of apps before anyone notices. One just hit a widely used building block called Axios, which a huge number of projects rely on. Minutes after the attack went live, Devin flagged to developers that it had reviewed the package and identified a security vulnerability. This happened before the public warnings went out. Because of that, specific customers were protected from the attack and able to make sure their data was safe. As AI capabilities improve, what we'll see is more autonomous use cases where AI is proactively taking action rather than being kicked and driven off by a human. We wanted to take a step further to put proactive defensive capabilities in the hands of all our customers. So we're launching Devin for Security. Devin Review will now come with security reviews out of the box, for all your PRs. Security Swarm will deploy a team of Devins to analyze your code base the way a team of security researchers would. Learn more here: devin.ai/security

linkedin.com/posts/scott-wu-8b94ab96_a-few-months-ago-a-malicious-actor-initiated-activity-7478145428337643520-zw0FJul 1, 2026

I'll be giving a mainstage keynote at RAISE Summit in Paris next week to discuss the Golden Age of Software Engineering. Even within the last 6 months, the speed and quality of what engineers leveraging AI can deliver has changed dramatically. That's worth getting excited about, and worth being honest about what this changes. If you're at RAISE, come find me and the Cognition team at our AI Lounge.

linkedin.com/posts/scott-wu-8b94ab96_ill-be-giving-a-mainstage-keynote-at-raise-activity-7477356976402993152-VGPxJun 29, 2026

5 days before Devin became generally available, our team looked at the product and decided it was close enough to ship but not yet what we wanted it to be. So that morning we told everyone, “get on a flight to Park City this afternoon, and let your family and friends know you'll be unreachable for the next five days.” Our whole team (all 18 or 19 of us at that point) flew out to Park City, Utah and moved into a single Airbnb. I remember us barely leaving the house and running on adrenaline and Diet Coke for the next five days straight. Halfway through the week we realized we should probably have a launch video to go along with the product, so we hired a local videographer to come out and film us in the house while we were still working on the product. The video and the launch barely got finished in time for the Tuesday morning go-live. Now we also learned the hard way that Park City sits at altitude. Some of our people started getting headaches, so we ordered a stack of oxygen tanks. There’s this wild photo from that week where around three or four in the morning the night before launch, everyone is plugged into oxygen and grinding. It's still one of my favorite memories. We were just a group of people who genuinely liked one another, working on a product we actually thought was cool, and committed to giving it everything we had.

linkedin.com/posts/scott-wu-8b94ab96_5-days-before-devin-became-generally-available-activity-7475582300815233025-FwlxJun 24, 2026

People often ask how we hire engineering talent to work on Devin, the AI software engineer. Of our first 35 people we hired at Cognition, 21 had founded a company previously. We didn't set out to hire only former founders, but the qualities that make someone good at building a company from 0 are increasingly the same qualities that make someone good at engineering work when agents handle implementation. Self-ownership, comfort with ambiguity, and the ability to look at a half-formed problem and decide how to move forward, are becoming the engineering job description for the next decade. This is different from what the industry's been selecting for over the last two decades when hiring junior engineers. Things like memorizing facts about a particular language, knowing every framework's syntax cold, or being optimally efficient on a leetcode-style algorithm question, are far less important today. We're always looking to hire talented, high-agency people who want to work on Devin with us. A line we use a lot internally is that it's fun being on The Avengers. The reason is that we genuinely want everyone here to look around the office and see people that are excellent at what they do, while still feeling like they're a bunch of friends in a room trying to build something cool.

linkedin.com/posts/scott-wu-8b94ab96_people-often-ask-how-we-hire-engineering-activity-7473045566394953728-6exIJun 17, 2026

Measuring someone's productivity by their token usage is a horrible idea. Giving everyone the same fixed token budget isn't much better. So what's the right way to roll out AI across your org? We built a system to measure how many productive engineering hours every Devin task is worth, validated against a dataset of real engineers’ times estimates. The goal is to answer the fundamental question that companies are grappling with: how much real value are you getting from each of your agent sessions? On top of that, we're giving an AI productivity guarantee! Now if Devin delivers less engineering value than you're paying for, we fund your usage until it does, up to $10M. The whole industry needs to move from measuring activity to measuring output. We hope to see more AI companies taking this approach.

linkedin.com/posts/scott-wu-8b94ab96_measuring-someones-productivity-by-their-activity-7468370692036526081-oB2kJun 4, 2026

Today we’re launching Devin Desktop, the next generation of Windsurf. We built this for a world where engineers are spending more time managing coding agents than looking at code. Devin Desktop makes an agent command center the default surface in the IDE. You can manage local agents, cloud agents, PRs, files, and context from one place. We’re excited to be working with teams like Ramp, Harvey, NVIDIA, Modal, and Intact Financial as they bring multiple agents into real engineering workflows. We’re also introducing Devin Local, our new primary local agent and the successor to Cascade. It was rewritten from scratch in Rust, is up to 30% more token efficient, and supports modern features like subagents. The IDE still matters. Developers still need their editor, extensions, keybindings, LSPs, terminal, and workflows. So Devin Desktop is a full IDE with an agent command center built in. Really proud of the team for shipping this one. Try it out and let us know what you think:

linkedin.com/posts/scott-wu-8b94ab96_devin-desktop-devin-activity-7467693301127618560-gAHSJun 2, 2026

We've raised over $1B at a $26B valuation, led by Lux Capital, General Catalyst and 8VC. Two years ago, we launched Devin as the first AI software engineer. It's hard to believe how far we've come since. Enterprise usage has grown >10x since the start of just this year, and our run-rate revenue is now $492M. Our goal is to build a future where software engineers operate more like architects, creatively structuring problems for armies of Devins to reliably execute on. I'm proud to see this already unfolding on our team where Devin already ships 89% of the code committed internally. We remain hard at work every day to make Devin even better. If it's been a while since you tried it out, take Devin AI for a spin and let us know what you think! Grateful that existing investors Founders Fund, Elad Gil, Definition, Omri Casspi, Conversion Capital, Positive Sum, Avenir, Vitruvian Partners, Bain Capital Ventures (BCV), 137 Ventures, Soma Capital, Alpha Wave Global have doubled down on our mission, and excited to welcome new investors Ribbit Capital, Atreides, and Layer Global on board.

linkedin.com/posts/scott-wu-8b94ab96_weve-raised-over-1b-at-a-26b-valuation-activity-7465501717044961280-I2SeMay 27, 2026

Devin just got a Windows PC. Windows runs on over 1.4 billion devices and is deployed everywhere from Fortune 500 companies to the living rooms of indie developers. But these codebases couldn't use cloud AI agents because every agent ran on Linux. So we gave Devin AI its own Windows VM. It builds, tests, and iterates in a real Windows environment. Users can now run Windows modernization projects and feature development workloads with the benefits of an AI cloud agent. We're iterating quickly based on early customer feedback. Really proud of the team for shipping this one. Learn more here http://devin.ai/windows

linkedin.com/posts/scott-wu-8b94ab96_devin-just-got-a-windows-pc-windows-runs-activity-7463261136390594560-Ow03May 21, 2026

Growing up, I was obsessed with math and programming competitions. My friends and I often wondered if some day, AI would just do all these things. It still feels a bit crazy how far we've come. These days, I'm obsessed with making software engineering 10x faster with Devin AI. Working with Anthropic has allowed us to deliver an experience to customers where building software is as easy as telling your computer what to do. Full story:

linkedin.com/posts/scott-wu-8b94ab96_the-problem-solvers-scott-wu-at-cognition-activity-7462913810908123136-mCwqMay 20, 2026

Really excited to work with this team!

x.com/ScottWu46/status/2115982062058373268Oct 30, 2026