In short
Hormesis is a biphasic dose-response phenomenon in which a low dose of a stressor triggers a beneficial adaptive response, while a high dose of the same stressor is harmful.
The classic hormetic dose-response curve is a U-shape (or inverted U, depending on the outcome measured): no stimulus produces no adaptive benefit, a mild stimulus produces an adaptive overcompensation that leaves the system more resilient than baseline, and an excessive stimulus overwhelms the system and causes net damage. This is distinct from a simple linear dose-response relationship where more is always worse (or always better).
Exercise is the most concrete, well-evidenced example of hormesis in human physiology: the mechanical and metabolic stress of a workout causes temporary muscle micro-damage and oxidative stress, and the adaptive response to that stress — increased mitochondrial density, improved antioxidant capacity, muscle protein synthesis — is what produces the fitness benefit. This is also the theoretical basis cited for interest in other mild stressors: brief cold exposure, sauna use, and some phytochemicals in plants (which produce mildly toxic compounds as a defense mechanism that, in small doses, may trigger a protective cellular stress response in the animals that eat them).
Hormesis is a well-established general biological principle, but it does not mean every stressor is hormetic at every dose a person might choose, and the effective dose window varies enormously by stressor and by individual. It is a reason to take dosing seriously in interventions built on this idea, not a justification for assuming that more stress, or an unstudied stressor, is automatically beneficial.