diff --git a/Manual.ipynb b/Manual.ipynb
index dd3890c..faf9314 100644
--- a/Manual.ipynb
+++ b/Manual.ipynb
@@ -146,7 +146,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "(<i>REPL note</i>: if the output of `flat` looks [wonky](https://github.com/JuliaData/DataFrames.jl/issues/1272), try `Matrix(flat(pd))`)\n",
+    "(<i>REPL note</i>: if the output of `flat` gets [split up](https://github.com/JuliaData/DataFrames.jl/issues/1272), try `Matrix(flat(pd))`)\n",
     "\n",
     "The first column shows what fraction of backtraces (in %) go through the `method at file:line_number` in the `stackframe` column. It's the same quantity as in `Base.Profile.print()`, except for recursive calls: if `factorial(2)` calls `factorial(1)`, that's 2 counts in Base's report, but only 1 count in FProfile.\n",
     "\n",
@@ -576,35 +576,6 @@
    "source": [
     "Use the `get_method, get_file, ...` functions on `StackFrame` objects (see above). `tree(pd::ProfileData)` is defined as `tree(backtraces(pd))`, and similarly for `flat`, so you can modify the backtraces and get a tree/flat view of the results."
    ]
-  },
-  {
-   "cell_type": "markdown",
-   "metadata": {},
-   "source": [
-    "# Scratch"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 1,
-   "metadata": {},
-   "outputs": [
-    {
-     "data": {
-      "text/plain": [
-       "ProfileData(49 backtraces)"
-      ]
-     },
-     "execution_count": 1,
-     "metadata": {},
-     "output_type": "execute_result"
-    }
-   ],
-   "source": [
-    "using FProfile, Calculus\n",
-    "\n",
-    "pd = @fprofile second_derivative(sin, 1.0)"
-   ]
   }
  ],
  "metadata": {