Operations | Monitoring | ITSM | DevOps | Cloud

Five coding agents sounds like leverage right up until a developer is the one keeping track of all five: one fixing a bug, one building a feature, one refactoring, and two waiting on input at the same time. AI did not create that problem. It exposed a workflow problem that was always going to surface once parallel development became normal instead of occasional.

Before agents were part of the picture, GitKraken Desktop had one job: make Git make sense. Visualize the commit graph, show branches and remotes clearly, and let a developer feel confident hitting merge. That job has not changed. What changed in GitKraken Desktop 12 is what sits on top of it.

Recently, Google released a couple new Gemini models. They were Gemini 3.6 Flash, 3.5 Flash Lite, and 3.5 Flash Cyber. Many people on the internet roasted Gemini Flash and Flash Lite for their less-than-frontier performance. This makes sense when you consider that OpenAI and Anthropic have been sparring back and forth for first place for quite a while but Gemini almost never gets that first place position. After all, this is Google we are talking about.

We didn’t set out to build a new category of product. We set out to stop juggling. That’s the word Gyo, the senior engineer who built the first version of Kepler, keeps coming back to when he talks about where it started. “I have a lot of terminals opened, and with all those tabs, it was very difficult for me to keep focused on what I was doing,” he says. “I’m not a juggler.”

Merge conflicts rarely make it into a sprint retrospective, but they should. They’re one of the most reliable ways to lose an hour of flow without anyone noticing it’s gone. Every developer expects them eventually, but almost nobody questions the workflow around resolving them.

GitBench is our attempt to see how well LLMs handle git-specific tasks. We put together a suite of 17 benchmark categories with a total of 204 specific test cases. We ran these benchmarks against 43 (and counting) different models with all of their supported effort/reasoning levels.

From plan to main. Software is no longer just a tool. It is the infrastructure of modern life. Software keeps airplanes in the sky and power flowing into our homes. It helps doctors save lives, scientists discover cures, farmers feed cities, and astronauts navigate space. It powers economies, protects supply chains, and connects billions of people across the world. Every major system humanity depends on now depends on software. Which means developers are no longer just building applications.

You’re no longer writing code. You’re managing a pipeline of agents writing it for you. If you’ve been running two, three, or four AI coding agents in parallel, you already know the problem. The agents are fast. The orchestration is chaos. You’re bouncing between terminal windows, manually rebasing branches, cleaning up messy commits, and trying to remember which agent is touching which repo.

Your pull request has been open for three days. Your reviewer hasn’t commented. You’re starting to wonder if anyone will ever look at it—and whether the code you wrote on Monday still makes sense on Thursday. This feeling is common. PR lifecycle time—the duration from first commit to merged code—directly impacts how quickly you ship features, how fresh your code stays, and how engaged your reviewers remain.

Tracking down who made a specific change in your codebase can feel like detective work. Whether you’re debugging an issue or trying to understand why a particular piece of logic exists, knowing the history behind each line is invaluable. GitKraken makes this process simple with tools like GitLens for VS Code and GitKraken Desktop, which bring blame annotations directly into your workflow.