token-efficiency

Token-efficiency minimizes token consumption by strategically selecting models, reading only necessary file sections, and preferring bash commands over file operations. Choose Opus for complex learning tasks and Sonnet for routine work to balance cost and quality.

Use Sonnet for routine development tasks like code reviews, debugging, and refactoring where full reasoning depth isn't required. Reserve Opus for complex learning, architecture design, and problem-solving where deeper analysis justifies higher token cost.

Use targeted bash commands like grep to locate relevant code sections, then read only those snippets instead of entire files. Minimize output by summarizing findings concisely rather than dumping full file contents.

Read-efficiency strategies include filtering files by pattern before loading, reading partial sections with line ranges, and using bash to extract relevant code rather than passing whole files. This reduces overhead while preserving answer quality.

Yes. Token-efficiency preserves quality by targeting high-value reads and focused analysis. Filter files systematically, ask specific questions, and let the model work with minimal context rather than loading entire repositories.

Yes. Awareness of Opus versus Sonnet capabilities, strengths, and typical token costs per task type is essential. Understanding model trade-offs lets you assign work appropriately and predict cost impact before executing tasks.