Anthropic, the AI startup behind the Claude family of models, released a new report as part of their Economic Index Research. The latest report suggests that Anthropic’s products, the web-based conversational AI platform Claude.ai and the coding agent tool Claude Code, are increasingly being used to automate coding processes, rather than simply using them as assistants.
The company analysed 5 lakh coding-related interactions across both products. Anthropic reported that 79% of conversations on Claude Code involved “automation”, where the AI directly completed tasks. The remaining 21% were categorised as “augmentation”, where the AI collaborated with users to enhance their capabilities. On Claude.ai, 49% of tasks were classified as automation.
Source: Anthropic
Moreover, the analysis revealed that web development languages such as JavaScript and HTML were the most commonly used, indicating that front-end development is at the forefront of AI-enabled coding.
Source: Anthropic
Furthermore, startups were found to be the primary, early adopters of Claude Code. “In a preliminary analysis, we estimated that 33% of conversations on Claude Code served startup-related work, compared to only 13% identified as enterprise-relevant applications,” read a section of the report.
A detailed breakdown of the nature of developers, coding tasks, and usage behaviours can be found in the report.
The analysis builds on one previously published by Anthropic earlier this year, which suggested that 57% of use cases on Claude involved leveraging AI to augment human capabilities, while 43% suggested automation.
The findings in the newer report found an increase in the rate of automation, thanks to the advent of Claude Code.
While Claude.ai can write and preview code with the Artifacts feature, the Claude Code tool released a few months back serves as an ‘agentic coding tool’ that operates directly within the terminal. It is capable of fixing bugs across a codebase, resolving merge conflicts, creating commits and pull requests, and answering questions about the architecture and logic.
Several developers have embraced and appreciated the product. Victor Taelin, founder of Higher Order Company, revealed how he used Claude Code to optimise HVM3, the company’s high-performance functional runtime for parallel computing, and achieved a speed boost of 51% on a single core of the Apple M4 processor.
He also revealed that Claude Code created a CUDA version for the same. “This is serious,” Taelin said. “I just asked Claude Code to optimise the repo, and it did.”