diff --git a/harness2/cli/apply.py b/harness2/cli/apply.py
index 227938af..659463e4 100644
--- a/harness2/cli/apply.py
+++ b/harness2/cli/apply.py
@@ -91,11 +91,11 @@ def _apply(args, ctx) -> tuple[int, dict]:
     todo = [(rel, kind) for rel, kind in delta if args.force or rel not in moved]
 
     label = (
-        "consolidated from the run's tasks"
+        f"Final harness of run {run_dir.name} ({args.mode})"
         if args.task is None and task == "consolidated"
-        else f"adopted for {task}"
+        else f"Harness adopted for {task} (run {run_dir.name}, {args.mode})"
     )
-    print(f"Harness {label} ({args.mode}, run {run_dir.name}) -> {where}\n")
+    print(f"{label} -> {where}\n")
     for rel, kind in delta:
         mark = "!" if rel in moved else kind
         print(f"  {mark} {rel:<48} {_delta(rel, current, adopted)}")
diff --git a/harness2/cli/eval.py b/harness2/cli/eval.py
index 2e44c836..eac094d6 100644
--- a/harness2/cli/eval.py
+++ b/harness2/cli/eval.py
@@ -82,22 +82,28 @@ def _text(data):
     def percent(value):
         return "-" if value is None else f"{value * 100:g}%"
 
-    lines = [
-        f"Comparing the consolidated harness with the starting one on {len(data['tasks'])} "
-        f"task(s), {data['repeat']} run(s) each.",
-        "TASK  START  CONSOLIDATED",
-    ]
-    for task, row in data["tasks"].items():
-        lines.append(
-            f"{task}  {percent(row['start']['mean'])}  {percent(row['consolidated']['mean'])}"
-        )
     start, consolidated = data["mean"]["start"], data["mean"]["consolidated"]
     delta = (
         f"{(consolidated - start) * 100:+g}"
         if start is not None and consolidated is not None
         else "-"
     )
-    lines.append(f"mean  {percent(start)}  {percent(consolidated)}  ({delta} points)")
+    rows = [("TASK", "START", "FINAL")]
+    rows += [
+        (task, percent(row["start"]["mean"]), percent(row["consolidated"]["mean"]))
+        for task, row in data["tasks"].items()
+    ]
+    rows.append(("mean", percent(start), percent(consolidated)))
+    width = max(len(r[0]) for r in rows)
+    table = [f"  {a:<{width}}  {b:>5}  {c:>5}" for a, b, c in rows]
+    table[-1] += f"  ({delta} points)"
+    lines = [
+        f"Comparing the final harness with the starting one on {len(data['tasks'])} "
+        f"task(s), {data['repeat']} run(s) each.",
+        "",
+        *table,
+        "",
+    ]
     if data["skipped"]:
         lines.append(
             f"Skipped {len(data['skipped'])} task(s) with no check or rubric: "
diff --git a/harness2/cli/init.py b/harness2/cli/init.py
index 34f9e4dd..bb8b3f80 100644
--- a/harness2/cli/init.py
+++ b/harness2/cli/init.py
@@ -107,7 +107,7 @@ def _run(args) -> tuple[int, dict]:
         print(message, file=sys.stderr)
         return 1, {"error": message, "exit_code": 1}
     ask = _asker(interactive=not args.yes and sys.stdin.isatty())
-    print(f"Setting up Harness² in {root}\n")
+    print(f"Setting up Harness² in {root}")
 
     interactive = not args.yes and sys.stdin.isatty()
     found = {name: _version(name) for name in ("claude", "codex")}
@@ -159,7 +159,6 @@ def _run(args) -> tuple[int, dict]:
         ignore.write_text("runs/\n.env\n", encoding="utf-8")
     (config.parent / "tasks").mkdir(exist_ok=True)
 
-    print()
     print(_harness_summary(root / harness, executor, editable))
     print(_key_summary(root, (model, improver)))
     print(f"\nWrote {config.relative_to(root)}")
diff --git a/harness2/cli/run.py b/harness2/cli/run.py
index 31680703..fac49701 100644
--- a/harness2/cli/run.py
+++ b/harness2/cli/run.py
@@ -117,7 +117,9 @@ def _estimate(tasks: int, k: int, modes: tuple[str, ...], totals: dict) -> dict:
 
 
 def _spend(totals: dict) -> str:
-    return f" · ${totals['spent_usd']:.2f} spent · {totals['calls']} model calls"
+    # Codex and subscriptions report no money; $0.00 would read as "free".
+    money = f"${totals['spent_usd']:.2f} spent" if totals["spent_usd"] else "spend not reported"
+    return f" · {money} · {totals['calls']} model calls"
 
 
 def _stopped(name: str, reason: str, totals: dict) -> str:
diff --git a/harness2/cli/status.py b/harness2/cli/status.py
index 7a55b273..5ccbbb27 100644
--- a/harness2/cli/status.py
+++ b/harness2/cli/status.py
@@ -136,7 +136,8 @@ def render(snap: dict) -> str:
         head,
         f"started {_when(snap['created'])}"
         + (f" · last activity {_ago(last.get('t'))}" if last else "")
-        + f" · ${snap['spent_usd']:.2f} spent · {snap['calls']} model calls",
+        + (f" · ${snap['spent_usd']:.2f} spent" if snap["spent_usd"] else " · spend not reported")
+        + f" · {snap['calls']} model calls",
         "",
     ]
     if rows:
diff --git a/tests/test_consolidate.py b/tests/test_consolidate.py
index fc7589d7..b1ffd7f8 100644
--- a/tests/test_consolidate.py
+++ b/tests/test_consolidate.py
@@ -213,9 +213,9 @@ def test_two_tasks_retry_adoption_eval_and_json(self):
         self.assertEqual(task["label"], "a-one")
         self.assertIn("+Task note.", task["diffs"]["CLAUDE.md"])
         applied, _ = self.invoke("apply", "trial", "--dry-run", "--task", "a-one")
-        self.assertIn("Harness adopted for a-one (parallel, run trial)", applied)
+        self.assertIn("Harness adopted for a-one (run trial, parallel)", applied)
         applied, _ = self.invoke("apply", "trial", "--yes")
-        self.assertIn("Harness consolidated from the run's tasks (parallel, run trial)", applied)
+        self.assertIn("Final harness of run trial (parallel)", applied)
         self.assertEqual((self.project / "CLAUDE.md").read_text(), "Use shared procedure.\n")
         text, _ = self.invoke("eval", "trial", "--repeat", "2")
         self.assertIn(
@@ -277,8 +277,8 @@ def test_eval_filters_skipped_failed_grades_and_limits(self):
         self.assertEqual(json.loads(Path(data["path"]).read_text()), data)
         with patch.object(GenericAdapter, "score", return_value=None):
             text, _ = self.invoke("eval", "trial", "--only-tasks", "a-one", code=3)
-        self.assertIn("a-one  -  -", text)
-        self.assertIn("mean  -  -  (- points)", text)
+        self.assertRegex(text, r"a-one +- +-")
+        self.assertRegex(text, r"mean +- +-  \(- points\)")
         data = json.loads((self.run_dir / "eval/parallel/results.json").read_text())
         self.assertEqual(data["tasks"]["a-one"]["start"], {"runs": [None], "mean": None})
         text, _ = self.invoke("-c", "run.max_calls=1", "eval", "trial", code=4)
@@ -429,7 +429,7 @@ def test_rollout_grades_and_optional_cost(self):
             "path": "results.json",
             "mean": {"start": 1.0, "consolidated": 0.0},
         }
-        self.assertIn("mean  100%  0%  (-100 points)", eval._text(data))
+        self.assertRegex(eval._text(data), r"mean +100% +0%  \(-100 points\)")
 
     def test_command_order_and_help(self):
         i = COMMANDS.index("diff")