30 years ago, providing a forecast of greater than 4-5 days would result in smirks from responsible forecasters.
But times have changed, and today there is substantial skill in the second week.
Let me prove that to you. The plot below shows how skill declines over time for the American Global Model (the GFS in blue) over the Northern Hemisphere. 1.0 indicates perfect skill by the measure considered (something called anomaly correlation). Anything above roughly 0.6 indicates useful skill.
So according ot this measure, there is skill through 8 days! More than a week!
Buit is better than that. Today, we have enough computer power to run many forecast runs each cycle (called ensembles), and the average of these many runs is even more skillful.
The US ensemble (GEFS, red color) has skill out to 10 days!
The European Center forecasts are even better, adding about another day of useful skill--11 days. Machine learning prediction adds another day.
So what has contributed to the improved skill? The most important contribution is three dimensional obserservations over the entire planet! No data voids to ruin forecasting skill.
Plus, more computer power leads to higher resolution simulations, better physics descriptions, and the ability to run ensembles of many forecasts.
Now that I have convinced you of the potential for extended prediction, what is the forecast for Seattle over the next ten days? Boring highs in the 50s every day (see below). If you want real warmth, you will have to head to Southern California.