Anthropic has released a new AI model, Claude 3.5 Sonnet, which boasts significant improvements in coding and tool use tasks. Most notably, it is the first publicly available frontier model to offer computer use, enabling it to interact with computers like a human would. This groundbreaking ability was achieved by training Claude to accurately count pixels on a screen, allowing it to give precise mouse commands and carry out instructions. While still in beta and prone to errors, this capability presents a significant leap in AI development, allowing developers to automate processes, build and test software, and conduct open-ended tasks. However, Anthropic acknowledges the potential safety and security concerns, particularly prompt injection attacks, and has implemented measures to mitigate risks and monitor for misuse, especially regarding the upcoming U.S. elections.