Risk Assessment Before A Rework

Run this twice on your own project, in two fresh conversations so neither run can read the other one's answer. Once with effort set to low, once with it set to max. Check /usage between them.

Before we change anything: assess what could break if we rework how the projects grid filters. List every risk you find.

Swap the projects grid for whatever piece of your own project other parts depend on. What matters is that it is a question with edges, so some risks sit on the surface and some only show up when you follow what depends on what.

What to compare

On an easy question, the two answers may land in the same place. On a question like this one, look for whether the deeper run found a specific dependency or edge case that the quick run missed.

Then look at the cost difference between them. That gap, right there on the screen, is the whole lesson. You are not buying more intelligence in the abstract, you are buying more thinking, and thinking costs, so you spend it where it pays off.

The mental model

Low effort is a quick sketch, fast and cheap, fine for something you will glance at and throw away. Max effort is an architectural blueprint, slower and more expensive, and worth every bit of it when the thing you are deciding is costly to reverse.

You would not pay for a blueprint of a grocery list, and you would not build a house foundation off a sketch. Most of your work lives in between, and the skill is matching the effort to the stakes.

Two honest cautions

The setting does not always behave the way you expect. People have found cases where a run at high did worse than a run at max, so if an answer feels oddly shallow, it is worth moving one level higher.

Watch the names, because they do not line up the way you would guess. A level called high sounds close to the top. It sits in the middle. The whole ladder is low, medium, high, xhigh, max, and ultracode at the top. On the models you are most likely to be using, high is where you already are without choosing anything, so you are starting on a sensible rung rather than at the bottom. The one reason to go looking at the rest of that list is a problem that is genuinely harder than usual.

Where to set it

/effort works anywhere. In the desktop app, Cmd Shift E opens the menu and you pick a level with a number key, the same way you pick a mode.

This resource is part of

Claude Code: From Beginner to Pro Agentic AI Engineer

by Arun Nagarathanam

Take your skills to the next level with

AI GEO 2026: Get Recommended by ChatGPT, Google, Perplexity

by Arun Nagarathanam