diff --git a/2013-09-07.md b/2013-09-07.md
index e69de29..59b5c5b 100644
--- a/2013-09-07.md
+++ b/2013-09-07.md
@@ -0,0 +1,18 @@
+<h3>Weekly Reflections for the week 9/29-10/5</h3>
+
+<h4>Review of IPython Distributions on Windows Platform</h4>
+
+From time to time, I lost access to my virtual box,  so I decided to try IPython distribution running on Windows platform. I tried two of them, the Anaconda and Enthopy Canopy.  Both of them are recommended on the IPython's offical website.
+
+The first one is Anaconda. Anaconda is larger in terms of size, and the latest release 1.7.0 come with Python 2.7.5. According to the offical website, the IPython package is part of the release. However it turn out to be not true for 1.7.0. But the 1.4.0 does come with IPython package. 
+On the contary, the latest Enthought Canopy comes with Python 2.7.3 and a few popular packages include IPython, matplotlib, pylab etc.
+
+Canopy appeared to be more attractive at the very beginning. It looks like commercial software with a user-friendly GUI to configure the packages. And you're requested to have some kind of subscription to install any additional package, even for gspread. But it seems that Canopy does not have a mature package version management system. The package manager suggested there are a couple of packages to upgrade. I followed the suggestion and the IPython broke after upgrade. The issue cannot be fixed by downgrading to original version. Finally I have to reinstall the whole thing to get it back to work, which was a hassle.
+On the contrary, Anaconda is not that user-friendly. For example, you cannot uninstall it from control panel as other applications. You need to go to the Anaconda directory and run uninstall program. But I didn't notice any restrictions on installing new packages I have no issue with upgrading packages.
+
+Overall, the IPython environment on Ubuntu server is still the best choice for me. But if I have to work on Windows, Anaconda will be my choice.
+
+<h4>Reflection on Professor Stark's presentation</h4>
+As for Professor Stark’s presentation, it is very intriguing. The main point I got is predictions of earthquake is very complicated and less successful. Seismologists have been trying to predict its occurrence but have been unsuccessful due to its complicated reasons. Statisticians have tried to predict earthquake using statistic models but got nothing close to accurate prediction. For statisticians, the whole idea of thinking of earthquakes as having probabilities is based on metaphor. For example, earthquake occurs as if in casino game, so when you think about the probabilities of earthquake, you have to think of the rule of game. There is lots of argument about the rule of game. Another example is to model earthquakes using Poisson process, however, it is possible that rate change during a certain period of time, and also because large earthquakes have been scarce during human history, it is difficult to evaluate its long-term rate. People tried to analogize earthquake to weather prediction, since they have things in common. For instant, “if rain today, predict rain tomorrow” works well for weather prediction, similarly, “if earthquake today, predict earthquake tomorrow” also works well. But the question is still there, how to predict the first rupture (today’s earthquake)? 
+
+For earthquake prediction, many problems left to be solved, this is why our project will be very intriguing and worth deep thought.