Anthropic's Claude Code becomes a primary driver for company growth
The AI coding agent created by Boris Cherny is expanding beyond software engineering into general office automation and business productivity.

The short version
- Claude Code is an AI agent that can write, run, and execute code while accessing a user's local files and operating system.
- The product reached an annualized run-rate revenue of $2.5 billion in February.
- Anthropic used Claude Code to build its own 'Claude Cowork' agent in just 10 days.
- Non-technical users are using the tool to automate tasks like paying parking tickets and creating educational games.
Why it matters
This shift shows that AI agents are moving from simple chat interfaces to active tools that can perform actions on a computer. It signals a transition where AI becomes a functional worker rather than just a writing assistant.
From a side project to a core product
Boris Cherny, the head of applications at Anthropic, developed Claude Code as a personal experiment. After joining the company in 2024 from Meta, Cherny spent a month prototyping ways for the AI to interact with a computer's operating system.
The tool was born from a specific moment when Cherny asked the model to identify the music he was listening to. Even though he did not explicitly instruct it to do so, the model wrote the necessary code to access the system and provide an answer. This ability to figure out how to use tools independently became the foundation for the product.
The success of this tool has been central to Anthropic's commercial strategy. As the company prepares for an initial public offering, investors are looking at how indispensable Claude Code can become to everyday office work. The product has already helped the company expand into sectors like finance, law, and marketing.
How the agent interacts with systems
Unlike standard chatbots that only generate text, Claude Code functions as an agent. This means it can execute commands, access files, and interact with the user's computer operating system to complete multi-step tasks. It essentially acts as a bridge between the AI's reasoning and the computer's actions.
The tool's versatility was demonstrated when Anthropic used it to build its own product, Claude Cowork, in only 10 days. This highlights how the coding agent can accelerate internal development by automating the heavy lifting of software creation. It allows for rapid prototyping of tools that manage files and automate routine office tasks.
Cherny has focused on refining the tool by maintaining a direct line of communication with its users. He frequently responds to questions on social media and has stated that talking to people about their specific use cases is central to his work. This feedback loop helps ensure the tool meets the practical needs of both technical and non-technical workers.
Expanding beyond software engineering
While it began as a tool for programmers, Claude Code is being adopted by people who do not write code. Non-technical staff at organizations like the Mid-Ohio Food Collective have used it to create a game for families to choose healthier foods. This project had been discussed for years but lacked the resources to develop until the AI agent provided the necessary 'unlock.'
Cherny himself uses the tool for mundane practical tasks, such as paying parking tickets and managing projects for his team. This shift toward general office automation is what led Anthropic to build Claude Cowork. The company recognized that users were already pushing the coding agent to perform tasks far outside the realm of software development.
The rise of these coding agents has also influenced broader business spending on AI. According to Ara Kharazian of Ramp, enterprise spending accelerated in late 2025 after Claude Code became generally available in May. These tools were among the first to deliver direct productivity gains for specific groups of workers by automating complex workflows.
The future of the engineering role
The growth of Claude Code has sparked a significant debate about the future of employment in the tech industry. Cherny has acknowledged that the technology will likely disrupt engineering jobs, but he suggests the work will evolve rather than disappear. He predicts that software engineering jobs could begin to vanish by the end of 2026 as the role shifts toward higher-level responsibilities.
In Cherny's view, the job title of software engineer might eventually become 'builder.' He argues that engineers do more than just write code; they work with designers and other teams to decide what products to build. As AI takes over the manual coding tasks, engineers may spend more time on these strategic and collaborative responsibilities.
Despite the optimism, there is significant pushback from the programming community. A survey by Stack Overflow found that six in 10 technology workers actively try to avoid using AI. Some workers worry that these tools are increasing burnout by creating pressure to keep up with new technology, while others remain skeptical about the tools' actual capabilities.
What to watch
- How many more non-technical industries adopt coding agents for internal tool building.
- Whether the 'builder' role becomes the standard job description for software engineers by 2026.
- How Anthropic balances its warnings about AI risks with the rapid release of automation tools.
Try it
Look for opportunities to use AI agents to automate a single repetitive task on your computer, such as organizing a folder of files or drafting a project timeline.
Sources and links
- ReportingSSBCrack: the article we sourced
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.
Listen to the AI Wire headlines
Listen to the AI WireListen to our daily AI shows every morning: on your commute, in the shower, wherever you are
Listen to the daily showsFirst reported by SSBCrack · Read the original. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
