What actually breaks
Tectonic plates move continuously, but faults can lock because of friction. Stress builds until part of the fault slips, releasing stored elastic energy as seismic waves. Earthquakes can also be induced by human activities such as fluid injection, mining or reservoir loading.
Why forecasting works better than prediction
Geologists can identify active faults, measure past ruptures and calculate the probability of damaging shaking over years or decades. Those hazard forecasts are useful for building codes and preparedness even though they cannot name the day of the next event.
Why short-term prediction keeps failing
Proposed precursors such as unusual animal behaviour, clouds, groundwater changes or small foreshocks have not produced a reliable, reproducible method that predicts major earthquakes with low false-alarm rates.
What early warning actually does
Earthquake early-warning systems do not predict an earthquake before it starts. They detect a rupture after it begins and use the fact that electronic signals travel faster than damaging seismic waves to provide seconds of warning farther from the epicentre.
Research record
USGS — Can you predict earthquakes?Official · open sourceUSGS — Earthquake facts and earthquake fantasyOfficial · open sourceUSGS — Can human activity cause earthquakes?Official · open sourceWhat Causes Earthquakes — and Can We Predict Them?
Direct answer: Earthquakes result mainly from sudden slip on stressed faults. We can forecast hazard and sometimes warn after rupture begins, but we still cannot reliably predict a major earthquake’s exact time, place and magnitude.
Hazard forecasting is established; deterministic short-term prediction is not. Claims of a reliable earthquake predictor should be judged against prospective, pre-registered performance rather than anecdotes after the event.