CONCEPT
What CrossFit is, methodologically and physiologically.
CrossFit is a training methodology built on a specific definition of fitness: work capacity across broad time and modal domains, in Greg Glassman's phrasing. It trains that with constantly varied functional movements performed at high intensity, and it scores the session. The scoring is not decoration. It is the mechanism by which intensity is actually delivered, because a clock and a whiteboard extract effort that a rep target does not.
Three modalities are mixed: weightlifting, gymnastics, and monostructural conditioning such as running or rowing. Sessions rotate across time domains, from efforts under a minute to efforts over half an hour. The claim is that a program covering the whole space produces a generalist who is competent anywhere in it.
Physiologically that is coherent. Repeated high-intensity intervals drive both peripheral adaptations, meaning mitochondrial density and capillarization, and central ones, meaning stroke volume. Loaded and gymnastic movements provide enough tension for strength and some hypertrophy in trainees who are not already strong.
The cost is structural, not incidental. Concurrent training carries an interference effect, documented since Hickson's 1980 work: endurance volume blunts strength and hypertrophy adaptations, roughly in proportion to how much of it there is and how close it sits to the lifting. A generalist program is submaximal in every specific by design. That is the trade, and it is a reasonable one if general capacity is what you want.
Fatigue degrades technique, and a scored workout rewards continuing anyway. Rhabdomyolysis is real but rare, and it dominates the discourse because it is dramatic. The common injury is boring: shoulders and lower backs accumulating overuse from high-rep loaded and overhead work performed under fatigue by people who cannot perform it cleanly when fresh.
The fix is not exotic. Scale the movement, cap the reps, and do not put a novice on the clock doing something they have not demonstrated rested. Whether a given gym does this is a coaching question, which is why the variance between affiliates is larger than the variance between methodologies.
Both "CrossFit is dangerous" and "CrossFit produces the fittest people alive" argue from the sport to the methodology. The competitive sport selects genetic outliers who then train like specialists, with dedicated strength blocks and periodization that look nothing like the class hour.
Varied is not random. Variety without progression is just novelty, and novelty is not a training stimulus.
The class model, the fixed time slot, and the people who notice when you skip are the strongest thing about it. Adherence outperforms programming, and this is a program that engineered adherence on purpose.