Google is facing internal discord over the increasing emphasis on artificial intelligence (AI) within the company. While Google’s leadership promotes the use of AI tools for coding and productivity, not all employees are convinced that the current approach is effective or fair. The core issue revolves around access to AI resources. Engineers at Google DeepMind reportedly have access to Claude, a prominent AI coding assistant, while most other teams are limited to using Google’s internal models, primarily Gemini.
The unequal access to AI tools has caused frustration among engineers in other divisions who believe that Google’s internal tools are still lagging in terms of coding performance. This disparity has created a situation where employees are urged to integrate AI more into their daily tasks, yet many feel they are not equipped with the best available tools.
In some teams, the pressure to utilize AI has become a component of performance evaluations. Engineers are given specific targets to not only use AI tools in coding but also to develop systems that enhance efficiency through AI. This shift in measuring productivity marks a departure from Google’s historical emphasis on engineering excellence.
Google justifies its preference for internal tools by highlighting the complexity of integrating external tools with its highly customized infrastructure. The company also upholds the principle of “dogfooding,” advocating for employees to use the same products developed for users to expedite tool improvements. Despite this rationale, some employees find this practice restrictive to their productivity.
The contrast with competitors like Meta is stark, as Meta reportedly allows more flexibility for employees to utilize external AI tools, including Claude. This openness at Meta has led some Google employees to question whether Google’s stringent approach is impeding internal progress.
A public debate ensued when programmer Steve Yegge criticized Google’s AI adoption, suggesting it was lagging behind expectations. Demis Hassabis refuted these claims, labeling them as “completely false” and “pure clickbait.” Yegge later revealed support from Google employees regarding the unequal access to tools like Claude.
Reports surfaced that an internal proposal to either expand Claude access or remove it entirely faced resistance, even from DeepMind. Some engineers strongly opposed the idea, with a few threatening to leave if access to Claude was revoked.