The iteration of Google's large model has been slow due to limited computing power and internal disagreements. Google's co-founder, Sergey Brin, urged core AI employees to fully commit to the development of the Gemini large model, emphasizing the necessity of catching up to the industry's cutting edge. According to sources familiar with the matter, Brin has pushed for resources to be allocated towards Recursive Self-Improvement (RSI) technology, which aims to enable AI to iterate itself without human intervention. Internal tests have shown that the performance of the new version of Gemini in areas such as code writing still lags behind competitors, prompting Google to postpone the launch of the new version of its flagship large model by two months to August.