A lighthouse keeper pastes three seasons of logs into a chat box and the tail of the prompt is silently dropped. What limit was hit?
Select an answer to reveal the explanation.
Short Explanation
Paste three seasons of logs and the tail vanishes because the model can only hold so many tokens at once. That cap is the context window. It is not a learning-rate shrink, a thermal alarm, or a tree-depth knob.
Full Explanation
A context window is measured in tokens. Overflow is silently dropped (or rejected), not converted into a learning-rate, cooling, or tree-depth limit.