Google has moved its recently formed AI coding strike team into a new group focused on a stage of model development the company calls midtraining, according to a report from The Information. The change suggests Google is continuing to adjust how it organizes work on coding tools as it tries to strengthen its position in a fast-moving race with rivals, including Anthropic.
The original team was assembled to accelerate progress on AI systems designed to help with software development. By shifting that effort into a dedicated midtraining group, Google appears to be narrowing the team’s mission around one part of the model-building pipeline rather than keeping it as a broad response unit.
Midtraining refers to work done between the earliest phases of training a model and the later fine-tuning stages that shape how it behaves for specific uses. In practice, that means the new group is likely focused on improving how Google's models learn and perform during a critical phase of development, especially for tasks tied to coding.
The reorganization comes as Google looks to keep pace in a competitive corner of artificial intelligence where coding assistants and developer tools have become a major point of comparison among large AI labs. Anthropic has drawn attention for its coding capabilities, and the reported move indicates Google is still trying to catch up in that area.
Google has made AI coding tools a more visible part of its product strategy, but the competitive pressure has intensified as rivals have released systems that many developers see as strong performers. In that context, internal changes to the structure of AI teams can reflect both technical priorities and a desire to move faster.
The report does not say how many people are part of the new group or whether the change is tied to any specific product launch. It also does not suggest that Google is abandoning its coding ambitions. Instead, the move appears to be an internal reshuffling meant to concentrate expertise on a phase of model development that may have outsized importance for code generation and related tasks.
Google has repeatedly reorganized parts of its AI efforts in recent years as it responds to a shifting competitive landscape. Those changes have often aimed to reduce duplication, sharpen focus, or accelerate delivery in areas where the company wants better results.
For now, the reported transition of the coding strike team into a midtraining unit signals that Google is still refining how it builds and improves its AI models. Whether that leads to a stronger coding product line will depend on how effectively the company can turn the reorganization into technical gains.