Coding Is Only 8.7% of What People Do With Claude's Agent — 48 Hours Later, OpenAI Cloned It

Editor's take

Anthropic revealed that users primarily employ Claude's agent for tasks beyond code generation, with coding constituting a mere 8.7% of observed interactions in the initial 48 hours. This data challenges the prevailing narrative that advanced LLM agents will predominantly serve developers. It suggests a broader, more general-purpose utility for these tools, impacting how enterprises might integrate them for workflow automation across non-technical departments.

The rapid cloning of this agent functionality by OpenAI, as reported by Towards AI, underscores the intense competitive pressure in the LLM space. This move signals an accelerated pace of feature parity and innovation, forcing companies to continuously differentiate beyond core capabilities. The focus now shifts to Anthropic's ability to maintain a lead in nuanced, user-centric applications rather than solely on raw performance metrics.

Future developments will likely hinge on whether Anthropic can cultivate unique use cases that competitors struggle to replicate, or if OpenAI's broader ecosystem and developer adoption will quickly eclipse Anthropic's initial niche successes. Observing the evolution of agent prompts and the emergence of specialized agent workflows across different industries will be crucial indicators.