Hey there, fellow dial-turner,
For the past two weeks I have been working on several various projects, one is helping a friend with some AI brand work for his agency, and another is working on my family's net worth tracking app. This is a project I iwll share more about in another email, on how I am moving our tracking from Google sheets to an app.
Anyway...
Claude has an effort setting: Low, Medium, High, XHigh and Max.
Most people I talk to either never touch it or leave it cranked all the way up. Because more has to be better, right?
Not always.
In one simple website redesign, Low effort took 1 minute. Max took 28 minutes. The results were nearly identical.
That's one example, not a benchmark. But it changed how I work, and it boils down to one rule:
Pick the lowest effort that reliably gets you the quality you need.
What effort does
Effort is how much work Claude puts into a task before it comes back to you.
Think of a LEGO builder. Same builder, same instructions. For a simple house, you want a quick first version you can react to. For a complicated castle, you want the builder to test the structure and find the weak spots.
The model is the builder. Effort is how much work that builder is encouraged to do. One catch: lower effort can make Claude less capable on hard tasks. It isn't just a speed switch.
Match the effort to the job
- Low: changing headings, colours or spacing, or working on a design together (fast feedback helps you steer)
- Medium: building a component with clear requirements, or connecting a CMS to a site (bump it up if it gets tricky)
- High: debugging a Cloudflare Worker or API, or checking an order or payment integration (a hidden error can get expensive)
- XHigh: migrating a big WordPress site to Astro (lots of moving parts)
- Max: only for a really difficult problem, and only when the extra work proves it's worth it
Change gears as the project changes
You don't need the same effort for every stage:
- Explore → Low. Get a first version and react to it.
- Implement → Medium. Connect the forms, routes and content.
- Verify → High. Check errors, accessibility and edge cases.
- Polish → Low. Small tweaks.
How to change it
In Claude Code, type /effort to open a slider, or set it directly with /effort high. Which levels you get, and the default, depend on your model.
Three things I learned the hard way
- Clear prompts come first. More effort won't fill in requirements you never gave it.
- Low is useful. If a quick task has several parts, give it a checklist.
- Checking earns its keep. A small task can still deserve High if a mistake would cost you money, a client or a bad decision.
Here's the thing: more effort isn't the same as more care.
Knowing when to care? That's still the human's job.
Want to play with it? I turned this into an interactive guide with a slider and a quick "what does my task need?" picker.
Keep it on Medium (until it matters),
Chris MBS Latham 🍔
P.S. Know someone who runs every task on Max and then complains it's slow? Forward this to them. You're welcome.
