Many individuals find it challenging to maintain a single conversation, yet Boris Cherny, the head of Claude Code at Anthropic, manages multiple Claude instances simultaneously. His method resembles that of an air traffic controller, skillfully managing numerous tasks in a highly dynamic environment.
#How Did Claude Code Evolve from Prototype to Essential Tool?
Claude Code commenced its journey as a straightforward terminal-based prototype in late 2024. It was not intended to be a complete product, laden with features typically expected from conventional software. Instead, it was designed for developers to engage with Claude directly via the command line, allowing a fluid exchange of tasks between local and web sessions.
In early 2026, Cherny shared a detailed post about his workflow that quickly gained traction, accumulating millions of views on social media. This post highlighted his method of orchestrating multiple Claude sessions, delegating tasks, and shifting workloads as needed, depending on each job's context.
Anthropic has experienced noteworthy results since the adoption of Claude Code, reporting a nearly 70 percent boost in productivity per engineer. This increase is particularly remarkable given that the organization's workforce expanded significantly during this same timeframe.
#What Are the Key Phases of Working with AI?
Cherny has outlined a crucial progression regarding how developers engage with AI technology. The first phase involves straightforward interactions where users input commands and receive responses, followed by iterations of refinement.
The subsequent phase introduces what he refers to as autonomous loops, which allow developers to set up tasks that Claude can perform independently. When these loops reach completion or present points requiring human input, developers can intervene as necessary.
The final phase is characterized by the execution of thousands of overnight sub-agents handling intricate tasks, showcasing the advanced capabilities of Claude Code. Presently, 80 to 90 percent of the code generated within Claude Code is produced through these innovative methods, reflecting a significant trend in Anthropic's overall engineering output.
#Why Should Investors Pay Attention to These Productivity Gains?
The productivity accomplishments emerging from the deployment of Claude Code at Anthropic are compelling for observers of the AI tools landscape. A consistent 70 percent increase in productivity, especially during a period of aggressive team growth, is a transformative indicator that can influence corporate strategies regarding research and development investments.
The transition toward managing autonomous agents presents a fresh perspective for investors. Companies that focus on the infrastructure needed to oversee thousands of concurrent AI agents, while also integrating their outputs into existing development frameworks, are addressing a crucial bottleneck. This operational complexity highlights the necessity for review, testing, and deployment of generated outputs, cementing the future potential of AI technologies within various industries.