diff --git a/my/twitter/all.py b/my/twitter/all.py
index 29196e4..f1e39a7 100644
--- a/my/twitter/all.py
+++ b/my/twitter/all.py
@@ -1,17 +1,29 @@
 """
 Unified Twitter data (merged from the archive and periodic updates)
 """
+from itertools import chain
 
 from . import twint
 from . import archive
 
 
+from more_itertools import unique_everseen
+
+
+def merge_tweets(*sources):
+    yield from unique_everseen(
+        chain(*sources),
+        key=lambda t: t.id_str,
+    )
+
+
 def tweets():
-    yield from archive.tweets()
-    yield from twint.tweets()
+    # NOTE order matters.. twint seems to contain better data
+    # todo probably, worthy an investigation..
+    yield from merge_tweets(twint.tweets(), archive.tweets())
 
 
 # TODO not sure, likes vs favoites??
 def likes():
-    yield from archive.likes()
+    yield from merge_tweets(archive.likes())
     # yield from twint