Feature: Cross-product model awareness (Chat → Code prompt compatibility)

Problem

When users write prompts in Claude Chat (e.g., Opus) intended for execution in Claude Code (e.g., Sonnet or Haiku), the chat model generates prompts calibrated to its own capability level. Lower-tier models in Code then fail to follow these prompts correctly, wasting tokens and producing poor results.

This is a system-level gap, not something users should manually compensate for.

Proposed Solutions

Option A — Explicit query: Chat asks the user which model they're using in Claude Code, then adjusts prompt complexity/specificity accordingly.

Option B — Internal sharing: Anthropic internally shares the user's Claude Code model setting with Chat, so it can automatically calibrate prompts for the target model's capabilities.

Why This Matters

• Users are already managing token budgets carefully (using lower models in Code to save costs)
• The current workaround requires users to manually "dumb down" prompts — defeating the purpose of using a smarter model for prompt authoring
• This friction increases token waste (failed attempts, re-prompting) rather than reducing it

Additional Context

• Also, the /usage stats comparison (e.g., "21x more tokens than Les Misérables") would be more useful if it showed the reference book's actual size (pages/tokens) for context. Without that, the comparison is meaningless to most users.