◇ Could this help me?
Yes for the practice, with the specifics unverified. Matching reasoning effort to task difficulty is sound and the video is honest about the trade, since the high setting costs materially more tokens and only pays for itself on problems where the quality of the thinking changes the outcome. The specific feature name and the claim of a hundred or more simultaneous agents could not be confirmed and should be checked against the installed version's command list and the official documentation rather than taken from a video, since command names and effort labels change between releases. General practice, and the useful discipline is to decide the effort level deliberately per task instead of leaving a high setting on and paying for it on trivial work.
I want to use reasoning effort deliberately instead of leaving it at one setting. First, tell me exactly which effort or thinking settings the version I am running actually supports, and how to confirm that myself rather than taking it on trust. Then take the kinds of task I describe below and sort them into ones that genuinely benefit from maximum effort, ones where a middle setting is enough, and ones where the lowest setting gives the same answer for a fraction of the cost. For each category, tell me what signal would show me I chose wrong.