diff --git a/CHANGELOG.md b/CHANGELOG.md
index 41154305..0bdb7fff 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -1,11 +1,16 @@
 # Changelog
 
-## Unreleased - v4.0.0-DEV
-Version 4 is a thorough rewrite of major parts of Vamb.
+## v4.0.1
+* Fix Random.choice for Tensor on Python 3.11. See issue #148
+
+## v4.0.0
+Version 4 is a thorough rewrite of major parts of Vamb that has taken more than a year.
+Vamb now ships with with an upgraded dual variational autoencoder (VAE) and
+adversatial autoencoder (AAE) model, usable in a CheckM based workflow.
 The code quality and test suite has gotten significant upgrades, making Vamb
 more stable and robust to bugs.
-Vamb version is slightly faster and produces slightly better bins than v3.
-The user interface has only gotten slight changes.
+Vamb version is slightly faster and produces better bins than v3.
+The user interface has gotten limited changes.
 
 ### Breaking changes
 * The official API of Vamb is now defined only in terms of its command-line
@@ -14,11 +19,8 @@ The user interface has only gotten slight changes.
   If you are using Vamb as a Python package, it means you should precisely
   specify the full version of Vamb used in order to ensure reproducibility.
 * Benchmark procedure has been changed, so benchmark results are incompatible
-  with results from v3.
-  In v3, a complete bin was defined as the total set of covered basepairs in any
-  contig from the input assembly. In v4, it's defined as the genome of origin,
-  from where contigs are sampled.
-  This new procedure is more fair, more intuitive and easier to compute.
+  with results from v3. Benchmarking is now considered an implementation detail,
+  and is not stable across releases.
 * Vamb no longer outputs TNF, sequence names and sequence lengths as .npz files.
   Instead, it produces a `composition.npz` that contains all this information
   and more.
@@ -33,15 +35,18 @@ The user interface has only gotten slight changes.
   (though read the Notable changes section below).
   
 ### New features
+* Vamb now included an optional AAE model along the VAE model.
+  Users may run the VAE model, where it behaves similarly to v3, or run the mixed
+  VAE/AAE model, in which both models will be run on the same dataset.
+* The Snakemake workflow has been rehauled, and how defaults to using
+  the VAE/AAE combined model, using CheckM to dereplicate.
 * Vamb is now more easily installed via pip: `pip install vamb`. We have fixed
   a bunch of issues that caused installation problems.
-* Added new flag: `--noencode`. With this flag, Vamb stops after producing the
-  composition and depth outputs, and does not encode nor cluster.
-  This can be used to produce the input data of Vamb to other clustering models.
 * By default, Vamb gzip compresses FASTA files written using the `--minfasta`
   flag.
 
-### Notable changes
+### Notable other changes
+* Using the combined VAE-AAE workflow, the user can get significantly better bins.
 * Vamb now uses `CoverM` internally to calculate abundances. This means it is
   significantly faster and more accurate than before.
   Thus, we no longer recommend users computing depths with MetaBAT2's JGI tool.
diff --git a/test/ci.py b/test/ci.py
index e850f3d7..efc3bdb5 100644
--- a/test/ci.py
+++ b/test/ci.py
@@ -26,13 +26,13 @@ def changelog_version(path):
     with open(path) as file:
         next(file)  # header
         textline = next(filter(None, map(str.strip, file)))
-    regex = re.compile(r"v([0-9]+)\.([0-9]+)\.([0-9]+)*(?:-([A-Za-z]+))")
+    regex = re.compile(r"## v([0-9]+)\.([0-9]+)\.([0-9]+)(-[0-9A-Za-z]+)?")
     m = regex.search(textline)
     if m is None:
         raise ValueError("Could not find version in first non-header line of CHANGELOG")
     g = m.groups()
     v_nums = (int(g[0]), int(g[1]), int(g[2]))
-    return v_nums if g[3] is None else (*v_nums, g[3])
+    return v_nums
 
 
 def readme_vamb_version(path):
diff --git a/vamb/__init__.py b/vamb/__init__.py
index 61dbf78a..f32f66e5 100644
--- a/vamb/__init__.py
+++ b/vamb/__init__.py
@@ -19,7 +19,7 @@
 7) Split bins using vamb.vambtools
 """
 
-__version__ = (4, 0, 0)
+__version__ = (4, 0, 1)
 
 from . import vambtools
 from . import parsebam
diff --git a/vamb/cluster.py b/vamb/cluster.py
index 52cc963f..f20291ed 100644
--- a/vamb/cluster.py
+++ b/vamb/cluster.py
@@ -529,11 +529,11 @@ def _wander_medoid(
     )
 
     while len(cluster) - len(tried) > 0 and futile_attempts < max_attempts:
-        sampled_medoid = int(rng.choice(cluster).item())
+        sampled_medoid = int(cluster[rng.randrange(len(cluster))].item())
 
         # Prevent sampling same medoid multiple times.
         while sampled_medoid in tried:
-            sampled_medoid = int(rng.choice(cluster).item())
+            sampled_medoid = int(cluster[rng.randrange(len(cluster))].item())
 
         tried.add(sampled_medoid)