1.1 billion tokens. That's how many units of text - roughly 750 tokens equals 1,000 words, so we're talking approximately 825 million words of back-and-forth - one developer burned through Claude in a single month of intensive vibe coding.
Vibe coding is the practice of describing what you want software to do in plain English and letting an AI write the actual code. It sounds lightweight until you see usage at this scale. Every file Claude reads to understand your project's structure, every code block it generates, every explanation it writes back all count against your token budget. On a large codebase, a single feature request can consume 50,000 to 100,000 tokens before the model writes one line of output. Stack dozens of those sessions per day and the numbers compound fast.
At 1.1 billion tokens, you're well past any consumer subscription. Claude's Pro plan ($20/month) and Max plan (starting at $100/month) both have rate limits that cap usage far below this - reaching 1.1 billion requires API access at usage-based pricing. Claude Sonnet 3.5 runs $3 per million input tokens and $15 per million output tokens at standard API rates. At a typical input-heavy coding ratio, 1.1 billion tokens could run anywhere from $5,000 to $15,000 depending on how much is input versus output.
The practical lesson: context management is the biggest cost lever when building with Claude Code. Feeding the model only the files directly relevant to each task - rather than the entire repository - can cut per-session token use by 60-80%. The model doesn't need to read your whole codebase to fix a single function. Getting disciplined about what you put in the context window is less exciting than the coding itself, but it's the difference between a $100 monthly bill and a $5,000 one.