Naive forecasting

The way that the naive method is explained does not make sense to me.
It says "where you assume that the next value will be the same as the previous time step. " let’s say to forecast time step 101, we take the vale of step 100, and for step 102, we take step 101. Unless the naive approach is used to only make a forecast for the immediate next step from our known value, then the prediction for all steps will be the same, and I will just get a flat line, and no trends and seasonality will be absorbed.

so I don’t know how is there a forecasting created for the entire training data in the lab 2-week1, the orange line.

The aim behind building a model is to learn the mapping from features (X) to target (y). Model prediction is associated with symbol \hat y

In the case of naive forecast, \hat y[t] = y[t - 1].

If your problem requires prediction for more than 1 timestep into the future (without considering seasonality), consider using other methods. Should the choice be to retain the last forecasted value, there will be a flat line as you observed. Here’s an approach when performing naive forecasting, taking seasonality into account.