diff --git a/alembic/versions/0053_drop_chat_journal_push_curator_tables.py b/alembic/versions/0053_drop_chat_journal_push_curator_tables.py
new file mode 100644
index 0000000..e21823a
--- /dev/null
+++ b/alembic/versions/0053_drop_chat_journal_push_curator_tables.py
@@ -0,0 +1,82 @@
+"""drop chat / journal / push / curator / weather tables
+
+Revision ID: 0053
+Revises: 0052
+Create Date: 2026-05-27
+
+Phase 9 of the MCP-first pivot. The Python models for these tables were
+deleted in Phase 8 (commit 91bafb6); this migration drops the orphan
+SQL tables and cleans up dead per-user setting rows.
+
+Hard-cutover: existing rows in these tables are discarded. The
+accompanying spec at docs/superpowers/specs/2026-05-26-mcp-first-pivot-design.md
+explicitly accepted this loss.
+"""
+from alembic import op
+
+
+revision = "0053"
+down_revision = "0052"
+branch_labels = None
+depends_on = None
+
+
+# Order: junction tables first so the parent tables don't trip FK drops
+# even with CASCADE — keeps the migration readable.
+_DEAD_TABLES = [
+    # Moments + junction tables (journal entity links)
+    "moment_notes",
+    "moment_tasks",
+    "moment_places",
+    "moment_people",
+    "moment_embeddings",
+    "moments",
+    # Chat / generation telemetry
+    "generation_tool_log",
+    "messages",
+    "conversations",
+    # Curator approval queue
+    "pending_curator_actions",
+    # Push notifications
+    "push_subscriptions",
+    # Weather cache (journal-prep background fetcher)
+    "weather_cache",
+    # Legacy RSS-item embeddings (pre-MCP pivot, briefly experimented with)
+    "rss_item_embeddings",
+]
+
+
+# Specific setting keys to drop. Anything matching the LIKE patterns in
+# upgrade() is also nuked; the explicit list catches one-off keys that
+# don't fit the namespaced patterns.
+_DEAD_SETTING_KEYS = [
+    "default_model",
+    "background_model",
+    "assistant_name",
+    "auto_consolidate_tasks",
+    "chat_retention_days",
+    "think_enabled",
+    "rag_default_scope",
+]
+
+
+def upgrade() -> None:
+    for table in _DEAD_TABLES:
+        op.execute(f"DROP TABLE IF EXISTS {table} CASCADE")
+
+    keys_csv = ",".join(f"'{k}'" for k in _DEAD_SETTING_KEYS)
+    op.execute(
+        "DELETE FROM settings WHERE "
+        "key LIKE 'voice_%' OR "
+        "key LIKE 'journal_%' OR "
+        "key LIKE 'briefing_%' OR "
+        "key LIKE 'curator_%' OR "
+        f"key IN ({keys_csv})"
+    )
+
+
+def downgrade() -> None:
+    raise NotImplementedError(
+        "No downgrade — hard cutover per the MCP-first pivot spec. "
+        "Rolling back is not supported at the database layer."
+    )