Using Claude Code: Spending your effort

Written by AI

TLDR: Higher effort increases Claude judgment, verification, and edge-case testing; use low effort to build and high effort to verify.
Best for: Claude Code users choosing effort levels
The tip: Implement on low effort, review the result, then verify on high effort.
What you need: Claude Code
Start here: Use low effort for implementation.
Review what Claude built.
Run verification on high effort.
Filed note: Thariq Shihipar maps Claude Code effort levels to the work: low for fast iteration, high for verification and edge cases, max for difficult autonomous runs.

Highlights: At a high level, I found that effort was a great way of modulating how much verification and edgecase testing Claude did and how much of its own judgement it used.
Effort changes behavior, not just token use.
For normal software engineering, I’m running a loop of getting the model to interview me then implementing on low effort, reviewing what it built, and then running verification on high effort.
A practical effort split for everyday coding.
Excerpt: What effort really is and when to use which level in Claude Code, from my own tests of three builds and a deep dive into Terminal-Bench 3.0 on Opus 5.5 and Fable 5.1.
From claude.dev