Claude Code's status line: Flamerix measures 16.4%, a team member says about 2%

TL;DR

• Flamerix AI measured Claude Code's agent_summary calls at 16.4% of its spend in one week, above the main conversation's 15.7%.
• Edwin Arbus, who works on Claude Code, says that doesn't seem typical, and gives about 2% of cost across users who send telemetry and about 6% in the terminal.
• You can check your own share through the OpenTelemetry export, and Arbus says a fix for summaries that miss the prompt cache rolled out in 2.1.311+.