Anthropic's coding assistant Claude Code currently features 83 built-in slash commands (shortcut commands that control the tool itself)—this signals that AI coding tools are transitioning from "chatboxes" into developers' "operating systems."

What this is

Claude Code is an AI coding assistant running in the terminal. We note that beyond having the AI write code, it provides a set of shortcut commands starting with a slash (/). While a standard prompt is content that instructs the AI to do work, slash commands are buttons that operate the tool itself—such as clearing the conversation, switching models, or returning to a previous session. There are currently 83 such commands. The most noteworthy is the context management capability: for example, /btw (temporary bypass Q&A) allows developers to ask an unrelated question without polluting the main conversation history, thereby saving token (the AI billing unit) overhead; /branch forks the current conversation like code version control, allowing developers to trial-and-error in a new branch, and then switch back to the mainline via /resume. These commands essentially manage the AI's attention.

Industry view

We believe that shifting from pure natural language dialogue to a hybrid interaction of "commands + dialogue" is inevitable for AI deployment. Development workflows rely heavily on state switching and context control; relying solely on chat is too inefficient, requiring deterministic shortcuts. However, there is opposition: the addition of 83 commands is turning AI tools back into traditional software that requires rote memorization. Developers' cognitive load increases—if managing conversation branches becomes more complex than writing code, it defeats the original intent of using AI to lower the barrier to entry. Furthermore, when enterprises use terminal AI internally, risk audits for access control (like /sandbox) and sensitive data leaks remain blind spots.

Impact on regular people

For enterprise IT: Permission boundaries for terminal AI must be clarified ASAP, especially whether code review and sandbox mechanisms meet corporate security and compliance requirements. For the individual workplace: The core skill of programmers is shifting from "writing code" to "managing AI context"; those who know how to branch AI conversations and clear caches will achieve higher output efficiency. For the consumer market: Regular users won't interact with slash commands directly, but as developers improve their ability to manage AI attention, software iteration speeds will accelerate further.