diff --git a/CHANGELOG.md b/CHANGELOG.md
new file mode 100644
index 0000000..a2689f2
--- /dev/null
+++ b/CHANGELOG.md
@@ -0,0 +1,71 @@
+0.4.4
+=====
+
+Features:
+
+* Use "human units" by default in `hist` and `plot` sub-commands.  Big numbers
+  (those with many digits) can be hard to read.  `lowcharts` will use some
+  heuristics to use units when helpful (for instance, "412.7 M" as opposed to
+  "412723763251.327").  Heuristics will take into account the range to be
+  displayed and will keep units consistent in all of the visualization.  Command
+  line option `--precision` is available to deactivate this feature and request
+  for an arbitrary number of decimals.
+
+Bug fixes:
+
+* In time histograms, do not panic if all input timestamps are the same.
+
+0.4.3
+=====
+
+Bug fixes:
+
+* Do not truncate numbers in Stats section of `hist` and `plot` sub-commands if
+  they have many digits.
+
+0.4.2
+=====
+
+Features:
+
+* Implement the sub-command `common-terms`.  Use it to display an histogram with
+  the number of occurrences of the most common terms of the input.  Use it with
+  `--regex` if the input is not filtered.  Example:
+
+```
+# Figure out the most used syscalls when listing a directory
+#
+$ strace ls -l 2>&1 | lowcharts common-terms --lines 5 --regex '(.*?)\('
+Each ∎ represents a count of 1
+[  openat] [40] ∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎
+[    mmap] [36] ∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎
+[getxattr] [29] ∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎
+[   close] [29] ∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎
+[   fstat] [25] ∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎
+```
+
+0.4.1
+=====
+
+Bug fixes:
+
+* Allow negative values in min and max arguments
+
+0.4.0
+=====
+
+Features:
+
+* Implement the sub-command `split-timehist`.  It mixes up the time histogram
+  and bar chart in a single visualization.  See README for an usage example.
+
+0.3.0
+=====
+
+Features:
+
+* Implement the sub-command `timehist`.  It displays the frequency of logs that
+  match a regex (by default any log that is read by the tool).  The sub-command
+  can autodetect the most common (in my personal and biased experience)
+  datetime/timestamp formats: rfc 3339, rfc 2822, python `%(asctime)s`, golang
+  default log format, nginx, rabbitmq, strace -t (or -tt, or -ttt), ltrace,...
diff --git a/README.md b/README.md
index 121e42e..b1ba7db 100644
--- a/README.md
+++ b/README.md
@@ -99,8 +99,8 @@ The idea is to depict the frequency of logs that match a regex (by default any
 log that is read by the tool).  The sub-command can autodetect the most common
 (in my personal and biased experience) datetime/timestamp formats: rfc 3339, rfc
 2822, python `%(asctime)s`, golang default log format, nginx, rabbitmq, strace
--t (or -tt, or -ttt),ltrace,... as long as the timestamp is present in the first
-line in the log and the format is consistent in all the lines that contain
+-t (or -tt, or -ttt), ltrace,... as long as the timestamp is present in the
+first line in the log and the format is consistent in all the lines that contain
 timestamp.  It is ok to have lines with no timestamp.  The consistency is
 required because of performance reasons: the 1st log line is the only one that
 triggers the heuristics needed to create an specialized datetime parser on the