A visual guide to self-driving codebases

Recently an article titled 'something big is happening' went viral.

It was a wake-up call to those not in the tech industry about how AI has hit this inflection point, since December 2025. It does a great job of putting into words what those of us keeping up with the frontier of coding AI feel: an inflection point, and like things are 'going exponential'.

I feel it, and see it in my own GitHub contributions graph.

The bottleneck of software development has shifted violently. The bottleneck isn't writing code anymore. And if it still is for you, it won't be for long.

Now, everyone is wrestling with the idea of parallel agents, sandboxing and worktrees or even buying multiple Macs to run more agents in parallel. The longer you invest in coding agents without addressing the system around it, the deeper you entrench. This is the false summit.

As agent autonomy has increased, the interest in 'background agents' has intensified, yet many are still confused about what they are, how they work, and where the industry is heading.

That's why today we are announcing: background-agents.com.

Into the background, and the cloud

It's become increasingly obvious that the software development life cycle must change. With this, we felt the need to unify the conversations we were having with customers, seeing in the ecosystem, and exploring internally in our organization.

To get the most out of coding agents, we have to shift agents not only to the background, but also to the cloud. On schedules, and on event triggers, so we can start to automate the new bottlenecks that are springing up in our lifecycle.

The website is not a feature announcement, but an attempt to guide and explore background agents in an interactive, visual way. As we enter a world where agents can remediate vulnerabilities on-demand, review our code automatically, and where we can wake up in the morning to fresh pull requests without prompting.

As an industry, we must name some of our patterns, and our terms. We think it makes sense to start with 'background agents'. From here, we can bring clarity to other emerging terms like:

Parallel agents, fleets, swarms, proactive and sub-agents.

Over the coming weeks, we will publish a lot about the challenges with parallel agents, the primitives to successfully deploy background agents, the use cases to focus on and eventually how to get to a self-driving codebase.

We hope this website becomes a vehicle of clarity in a time of uncertainty and rapid change.

We'd love to know what you think of background-agents.com