Days after it gave Claude 3.5 Sonnet the ability to use computers, Anthropic has brought yet another powerful new feature to its Claude AI assistant. The AI upstart is now introducing a new analysis tool, and will let the assistant write and run JavaScript code, amongst other tasks.
Anthropic describes the tool “as a built-in code sandbox,” wherein users will be able to leverage an embedded coding environment to write, test, and execute JavaScript directly within Claude.ai’s interface. In other words. this means that they can interact with data by running code in the chat window, even as they request text-based responses. The analysis tool is currently available for all Claude users. This new analysis builds on improvements introduced with the Claude 3.5 Sonnet model, which Anthropic claims to have showcased an increase in its performance.
Users do not need to have a programming background to utilize this feature, since the need to switch to external coding platforms is negated. Claude will process the data on a step-by-step basis to provide “answers that are not just well-reasoned, but are mathematically precise and reproducible.”
A key feature of the code sandbox is that there is no delay, and it will provide interactive visualizations on the spot. In its official statement, Anthropic showcased this using a sales funnel analysis, wherein Claude managed to go through raw data, analyze it. and present the final results through charts and graphs. If users import data from spreadsheets, CSV files, and PDFs, Claude can go through them as well.
Anthropic has highlighted several potential applications for the JavaScript sandbox across different organizational departments. Marketing teams, for example, can upload datasets containing customer interactions and use the sandbox to identify opportunities for conversion improvements. Similarly, sales teams can analyze global sales performance and break down metrics by region to optimize strategies. Product managers can also benefit from the tool by processing user engagement data for sprint planning, while engineers have a way to analyze server logs.
