Learning Over Time

With each new example:

1. the model adjusts its internal math
2. the loss becomes smaller
3. the result becomes better

If training is done correctly, the model gets closer and closer to the correct solution.