The Loading Verb Leaderboard
A developer recently built claude-spinner-verbs.vercel.app, a community leaderboard tracking Claude Code's "loading verbs"—those brief status messages that appear while Claude processes your requests. While the site itself is a fun community project, understanding these verbs reveals crucial information about what Claude Code is actually doing behind the scenes.
What Loading Verbs Actually Tell You
When you run claude code "refactor this function", you'll see verbs like:
thinking- Claude is analyzing the problemplanning- Breaking down the task into stepsexecuting- Running commands or writing codereviewing- Checking its own workbrowsing- Searching documentation or files
These aren't just decorative animations. They're real-time indicators of which subagent or process is active. The recent Claude Code update revealed subagent features in the interpreter for isolated task execution—these verbs correspond to those different execution modes.
Why This Matters for Your Workflow
Know when to wait vs. intervene - If you see
thinkingorplanningfor more than 30 seconds on a simple task, Claude might be stuck in analysis paralysis. Consider breaking your request into smaller chunks.Identify resource-intensive operations -
browsingindicates file system operations. If this appears frequently, consider using the/compactflag or providing more specific file paths in your initial prompt.Track multi-step processes - Complex tasks will cycle through multiple verbs. Watching the pattern helps you understand Claude's approach to your problem.
Practical Applications
Debugging stuck agents:
# If you see "thinking" for too long, interrupt and try:
claude code --compact "Just fix the syntax error on line 47"
Optimizing token usage:
When you see browsing frequently, it means Claude is scanning files. Pre-load context:
# Instead of letting Claude browse
claude code "What's in utils.js?"
# Provide the file directly
cat utils.js | claude code "Analyze this file"
Timing complex operations:
Note which verbs take longest for your codebase. If reviewing is consistently slow, your code might need better documentation or simpler patterns for Claude to parse quickly.
The Community Aspect
The leaderboard project shows developers are paying attention to these details. While the site currently has no submissions, it represents a growing understanding that Claude Code's interface provides meaningful feedback—not just progress bars.
As more developers contribute verbs they've observed, patterns will emerge about common workflows and potential bottlenecks in Claude Code's processing pipeline.
What's Next
Watch for new verbs in future updates. Anthropic has been expanding Claude Code's capabilities with new MCP servers and subagent features. New verbs likely indicate:
- Integration with additional tools (design systems, databases, etc.)
- More sophisticated error recovery patterns
- Parallel processing capabilities
For now, use the verbs you see as diagnostic tools. They're free information about how Claude Code is tackling your requests.



