diff --git a/src/content/post/2024-11-13-kuzu-v-0.7.0.md b/src/content/post/2024-11-13-kuzu-v-0.7.0.md
index 61e7ae6..690ef54 100644
--- a/src/content/post/2024-11-13-kuzu-v-0.7.0.md
+++ b/src/content/post/2024-11-13-kuzu-v-0.7.0.md
@@ -138,8 +138,12 @@ creating its persistent disk-based structures. This meant that if you had machin
 to manually chunk the source data into smaller parts, otherwise the `COPY FROM` pipeline could run out of memory and fail.
 
 In this release, Kùzu automatically spills the records it scans from the source table into
-a temporary file on disk during `COPY FROM`.This is in fact done automatically by the buffer manager
-as `COPY FROM` demands more and more memory. As a result of this improvement, Kùzu can now ingest very
+a temporary file `.tmp` located inside the database directory on disk during `COPY FROM`. 
+The spilling is in fact done automatically by the buffer manager
+as `COPY FROM` demands more and more memory.
+It is by default turned on when you open a local Kùzu database under read-write mode, and you can choose to turn if off by `CALL spill_to_disk=false;`.
+Note that we don't support spilling under in-memory mode for now.
+As a result of this improvement, Kùzu can now ingest very
 large sets of relationship records (such as the edges of the 17B Graph500-30 graph in the above experiment) 
 when it is running on a machine with limited memory.