diff --git a/codestrain_cli.py b/codestrain_cli.py
index 1d49121..d255e02 100755
--- a/codestrain_cli.py
+++ b/codestrain_cli.py
@@ -17,6 +17,9 @@
 import sys
 from pathlib import Path
 
+# Keep in sync with pyproject.toml [project].version at every release.
+__version__ = "0.1.8"
+
 
 # ── ANSI Colors ──────────────────────────────────────────────────────────────
 
@@ -590,7 +593,7 @@ def print_header_adaptive(mode="auto"):
         # else: skip logo entirely on cramped terminals
     if mode != "none":
         print()
-    print(c(Colors.DIM, "  Your AI coding recovery score."))
+    print(c(Colors.DIM, f"  Your AI coding recovery score.  v{__version__}"))
     print()
 
 
@@ -686,10 +689,15 @@ def print_session_summary(stats_list, label=""):
     print(f"  Tokens:    {c(Colors.CYAN, format_tokens(total_input))} in / {c(Colors.CYAN, format_tokens(total_output))} out")
     print(f"  Cost:      {c(Colors.AMBER, format_cost(total_cost))}")
 
-    if all_models:
-        models_str = ", ".join(sorted(all_models)[:3])
-        if len(all_models) > 3:
-            models_str += f" +{len(all_models) - 3} more"
+    # Hide "<synthetic>" from the displayed list: it's not a real model, just
+    # Claude Code's marker for locally-fabricated events (cached API errors,
+    # interrupted turns, no-response slash commands). Token/turn counts still
+    # include them — only the user-facing Models row is filtered.
+    display_models = sorted(m for m in all_models if m != "<synthetic>")
+    if display_models:
+        models_str = ", ".join(display_models[:3])
+        if len(display_models) > 3:
+            models_str += f" +{len(display_models) - 3} more"
         print(f"  Models:    {c(Colors.DIM, models_str)}")
 
     print()
@@ -728,20 +736,39 @@ def print_project_breakdown(project_stats, anonymize=False):
         reverse=True,
     )
 
+    # Pre-compute each row's plain (uncoloured) text so we can size columns
+    # off the visible width. Padding inside an f-string field on already-
+    # coloured strings doesn't work: ANSI escapes count toward the width
+    # spec and silently eat the padding, which leaves the breakdown jagged.
+    rows = []
     for i, (project, stats_list) in enumerate(sorted_projects, start=1):
         total_duration = sum(s["duration_seconds"] for s in stats_list)
         total_cost = sum(s["total_cost"] for s in stats_list)
         total_turns = sum(s["turn_count"] for s in stats_list)
 
         if anonymize:
-            project_display = f"project-{i}"
+            name = f"project-{i}"
         else:
-            project_display = project[:30] + "..." if len(project) > 30 else project
+            name = project[:30] + "..." if len(project) > 30 else project
+
+        rows.append((
+            name,
+            format_duration(total_duration),
+            str(total_turns),
+            format_cost(total_cost),
+        ))
+
+    name_w  = max(len(r[0]) for r in rows)
+    dur_w   = max(len(r[1]) for r in rows)
+    turns_w = max(len(r[2]) for r in rows)
+    cost_w  = max(len(r[3]) for r in rows)
+
+    for name, dur, turns, cost in rows:
         print(
-            f"  {c(Colors.WHITE, project_display):<36}"
-            f"{bold(format_duration(total_duration)):>10}  "
-            f"{c(Colors.CYAN, str(total_turns)):>6} turns  "
-            f"{c(Colors.AMBER, format_cost(total_cost)):>8}"
+            f"  {c(Colors.WHITE, name.ljust(name_w))}  "
+            f"{bold(dur.rjust(dur_w))}  "
+            f"{c(Colors.CYAN, turns.rjust(turns_w))} turns  "
+            f"{c(Colors.AMBER, cost.rjust(cost_w))}"
         )
 
     print()