Coder

Most platform teams frame it the same way, how do we make AI work at scale? That's a reasonable question to ask, but it skips something important. Are the efficiency gains you��re chasing actually real, or do they just feel that way? On March 31st, we're getting into how teams are managing context so models stay focused. Where subagents and swarms actually help distribute work, and where they just add complexity. The security tradeoffs that show up once agents start getting more autonomy. We're getting into the real patterns, tradeoffs, and the places where this stuff starts to break. Link in comments to register. Hope to see you there.

44% of organizations say onboarding new developers takes more than two months. Across six multinationals, new hires without AI tools took a median of 91 days to reach their 10th pull request. The root cause is almost always the same: local environments that accumulate dependencies, break constantly, and live on individual machines. We co-hosted a workshop with AWS where attendees deployed Coder on AWS sandbox with AI-driven development workflows through Amazon Q Developer and Claude Code via Bedrock. The conversation has shifted. Teams aren't just standardizing where developers write code. They're building infrastructure that makes agentic AI development possible without compromising security. Palantir cut time to first commit from 15 days to one hour. Skydio slashed dev VM spend by 90%. That same infrastructure now powers their agentic development. https://lnkd.in/gwtS3F_Z

KubeCon EU 2026 delivered. Great conversations, great energy, and a lot of love for what we're building at Coder. It's always a highlight to spend time with so many platform teams pushing things forward and hear firsthand how they're tackling real challenges. Already counting down to 2027 😎 Carly Stephens Harlee Clark Nick Bedson Stefan B. Simon Gregory Théo D. Thomas Kosiewski Pawel Kawinski

Vibe coding gets dismissed too quickly. Dan Vega explains why that’s a mistake. Dan sees vibe coding as a way to lower the barrier to entry. Someone with a real idea can finally turn it into an MVP without hiring a team or spending months just to get started. But the catch is vibe coding should just be the beginning, not the finish line. The real value is using that momentum to learn fundamentals, understand tradeoffs, and turn an experiment into something solid. Check out the link to the full conversation in the comments.

Coder and AWS together let you run AI-assisted development entirely inside your own infrastructure. The foundation is Coder's control plane, which orchestrates template-defined workspaces that spin up as Kubernetes pods or EC2 VMs, plugged into the AWS services your team already uses. And because workspaces are template-defined, you can: → Deploy the Q Developer IDE plugin across web-based and local IDEs like VS Code → Run self-hosted agents like Claude Code inside secure workspace boundaries → Roll out the Q Developer CLI org-wide through Coder templates → Route everything through Bedrock as your LLM backend, mixing and matching agents and models as needed Add in integrations with your existing Git and Artifactory providers, and AI-assisted development actually fits into your infrastructure instead of floating outside it. When everything runs through AWS and Coder together, you get the flexibility of modern AI tooling without giving up control. #AWS #CloudDevelopment #AgenticDev

Coder's second company-wide hackathon put 27 people on nine teams for one week. Engineers, marketers, designers, and PMs all building on the same platform. The projects ranged from a Kubernetes operator that deploys Coder clusters from Slack to a workspace agent supervisor that auto-recovers crashed agents. One team built a graphics generator with no professional engineering experience on the team. They started with a product requirements doc, gave the agent structured context before each feature, and shipped a working tool in a week. That's the part worth paying attention to. AI agents have reached the point where clear planning and a shared vision matter more than technical background. The barrier to building is lower than it's ever been, and the range of who can build is expanding fast. When you give people the right infrastructure and the space to experiment, the results go places no roadmap would predict. https://lnkd.in/gceaJh5A

Tomorrow: our hands-on workshop on deploying AI agents in Coder CDE. If you've been meaning to register, today's the day. The session walks through how to run AI coding agents inside secure, governed cloud development environments — on your infrastructure, with full visibility into what the agent is doing and why. March 26 at 9:00 AM PT. One hour. Seats are limited. Register: https://lnkd.in/gBQDaRVP

The question we keep hearing at KubeCon EU this week: how do you standardize development environments across thousands of engineers without slowing anyone down? That's been Coder's home turf for years. Self-hosted CDEs on your infrastructure, defined as code with Terraform, provisioned in seconds, governed by your policies. No more "works on my machine." No more laptop sprawl. But the conversation is evolving. As teams start running AI coding agents alongside their developers, the same infrastructure that governs human workflows needs to govern agent workflows too. Same isolation. Same audit trails. Same templates. That's the problem we're solving, and we've got some exciting things in the works. If you're at KubeCon in Amsterdam, come find us at Booth #794. Whether you want to talk CDEs, AI agents, or both, we'd love to dig in.

When do you stop being a coder and become an engineer? Dan Vega explains why writing code is only a small part of building software. The real shift happens when you understand requirements, production, security, and ownership across the full lifecycle. Especially in enterprise work, engineering is about process, not just output. This clip hits a line too many conversations skip. For the full conversation grab the link in the comments.