What actually changes

The effect of a higher effort level shows up less in what you directly see printed and more in the actual choices tCode makes along the way: which file it decides to check first, whether it pauses to search for other call sites before editing something.

It also shows up in whether it notices an edge case that a lower-effort pass might have missed entirely, simply by spending more deliberation on the decision before committing to it.

It is not a visible "thinking" transcript printed separately to the screen; it changes the quality of the decisions behind the visible tool calls, not the format of the output itself. See Outputs for exactly what does get printed regardless of effort level.

This is a deliberate choice: a raw stream of intermediate reasoning would add a great deal of noise to the transcript without necessarily being more useful than the tool calls and closing summary that already show what was actually done and why, in plain language.

If you want insight into why a specific decision was made, the fastest way is usually just asking directly, "why did you check that file first," which gets you a genuine explanation in the model's own words rather than a raw reasoning dump you would have to interpret yourself.