Google DeepMind Has Reportedly Assembled an A-Team to Take on Anthropic’s AI Coding Prowess

Coding seems to be the flavour of the month for artificial intelligence (AI) companies. Recently, OpenAI increased its focus on its coding platform Codex, and now Google appears to be shifting its focus as well. As per a report, the Mountain View-based tech giant's DeepMind division has created a specialised team that has been tasked with building AI models focused on coding. The aim is, reportedly, to close the gap between the company and Anthropic and is likely to secure a larger share of the lucrative high-ticket enterprise revenue.

Google's A-Team Targets Anthropic's AI Models

According to The Information, Google DeepMind has created a new team of researchers and engineers to develop the coding capabilities of future Gemini models and to build coding-focused large language models (LLMs) from scratch. This team is reportedly being led by research engineer Sebastian Borgeaud, who previously looked after pre-training of AI models.

Some of the areas the team is focused on are said to be complex coding work, long-horizon programming, writing an entire software from scratch, and enabling the models to read files to contextually understand the requirements of the user. It appears that the company is aiming to develop an AI model capable of handling end-to-end coding tasks.

Google's Co-Founder, Sergey Brin, and DeepMind's CTO, Koray Kavukcuoglu, are reportedly directly involved with the team. In an internal memo seen by the publication, Brin reportedly highlighted the need to “urgently bridge the gap in agentic execution and turn our models into primary developers.” The Co-Founder is also said to have asked engineers working on Gemini to use internal agents for complex tasks.

Google is reportedly pushing to improve the coding capabilities of its models because the executives believe that Anthropic's models are better at such tasks, and the Gemini maker does not want to be left behind. With the enterprise adoption of AI coding tools increasing, major AI players have switched their focus to developing models that can handle complex tasks autonomously.

Just last week, OpenAI upgraded Codex with Computer Use and image generation capabilities, allowing the desktop app to access files, apps, and interfaces to write code, run programs, test software, and iterate on existing builds. The image generation feature also bolsters the coding tool's ability to handle frontend tasks.