Evo has described an update to its autoresearch system that moves the workflow into Anthropic Claude Code and adds dynamic behavior to the process.

In a post shared by Alok Bishoyi on X, the company said it ported Evo's autoresearch loop into workflows and then made it dynamic. The update comes after Anthropic introduced dynamic workflows in Claude Code on June 2, a feature that allows Claude to write a small amount of code as part of a workflow.

The post points to an article titled "Self-Evolving Autoresearch Workflow Loops," which outlines the approach in more detail. According to the description, the project focuses on making the autoresearch loop self-evolving, with workflow steps that can adapt rather than follow a fixed path.

The source material does not include technical benchmarks, performance results or deployment details. It also does not say how widely the system is being used or whether the workflow changes are available to the public. Still, the post suggests the work is intended to build on Anthropic's newly announced workflow capabilities rather than replace them.

Claude Code has become one of Anthropic's more closely watched developer tools, and dynamic workflows appear designed to let the model do more inside automated processes. Evo's update shows one example of how developers may use that capability to redesign research-oriented agents and loop structures.

The company's description frames the effort as a port of an existing autoresearch process, not a from-scratch system. That distinction matters because it suggests the work is about translating an established workflow into a new environment and then letting the workflow change dynamically as it runs.

The post did not provide a timeline for future releases or mention whether the workflow is tied to a specific product. It also did not identify the size of the team behind the work. What is clear from the announcement is that Evo sees Anthropic's dynamic workflows as a useful foundation for more adaptive agent systems.

As AI labs and toolmakers race to add more autonomy to coding and research assistants, workflow design has become an important part of the conversation. Rather than focusing only on model output, teams are increasingly experimenting with how models can structure, revise and extend their own tasks.

Evo's update fits that trend. By combining an autoresearch loop with dynamic workflows in Claude Code, the company is signaling interest in systems that can iterate on themselves as they work, instead of following a static script from start to finish.