I went to the Denver Auto Show with a friend this moring. I drove my S2000 and he paid for parking. It's always fun to look at all cars, but the only things I liked better than the S2000 were much more expensive

In The Potential of the Cell Processor for Scientific Computing, Williams, et. al., compare the STI Cell Processor with contemporary commodity microprocessors and the Cray X1E vector processor. They find that it far outsrips the other architectures for single precision floating point and is competive with the X1E for double precision floating point calculations, with both significantly outperforming AMD and Intel chips. The results are even more impressive when power consumption is considered. They used the Cell simulator for this work, as the Cell wasn't available at the time. I'd like to see an update now that actual hardware is here.

Rand Simberg has a wrapup of the recent Space Access conference. In short, while Big Space is struggling, Little Space is doing fine, one baby step at a time.

The Connaught Type D Hybrid seems to offer a lot, decent looks, respectable performance, and excellent fuel economy. I just want to know if they'll export it to the U.S. and how much will it cost?

I've looked at Zope 3 and appreciate it's goal of making Zope development more Pythonic, but it has a steep learning curve and the extensive use of XML for configuration put me off a bit. Grok is an interesting new web application framework that allows you to configure almost everything in Python, much prettier. I may have to try it out.

Michael Chichton's Prey recounts the misadventures resulting from a swarm intelligence project developed for the federal government by a Silicon Valley startup. It started a little slowly for me, and some of the characters were a bit too stereotypical, but it was still enjoyable. While the nanotechnoly gone wrong theme of the book has received most of the attention, the real point is that science funded by venture capital may be too focused on short term revenue goals and dangerous shortcuts might be employed.

PyPy, the Python in Python project, has reached it's version 1.0 milestone release. This is an interesting project, though currently of limited usefulness. Some of the lessons learned have fed back into Stackless and the RPython, the Restricted Python that it is written in, can be used in place of C to write Python extensions.

PyPy has benifited from two years of EU funding, but will now have to continue as a volunteer effort or find other funding sources. I think they'll find a way to continue. Congratulations to the PyPy team.

Technology Revies reports on Brittish Petroleums effort to make butanol from plants like sugar beats. Butanol is easier to ship and has higher engergy density than ethanol. BP hopes the process they'll developing will be price conpetitive with gasoline.

I bought Python Scripting for Computational Science because it covers ground that isn't covered by any other text that I've found. While it has a place on my refernce bookshelf, I have some problems with it. It directs the user to install packages in non-standard locations and depends upon the author's personal extension library which has limited availability. One can hope for somethng better in the near future, but I haven't heard of anything.

Little did I know when I saw Pixel and Texel at the X Prize Cup, that I was looking at a proof of concept for a modular orbital launch vehicle. Some of the competition criticized the Armadillo Aerospace design because the four engine modules resulted in non-centerline thrust, complicating the control problem. It makes more sense now that we know more of the plan.

The Sony PS3 is now providing the bulk of the processing power for the Folding@Home project despite only being introduced last year. Protein folding simulation is an important research area which could lead to better treatments for cancer, Parkinson's, and other diseases, but it is very computationally intensive. This ins an impressive testimonial both to the PS3's performance and the advantages of distributed processing.

... via Richard Jones

ffnet is a feed forward neural net library for Python. The heavy lifting is done with Fortran and NumPy, so it should execute quickly while allowing the user to model the NN within Python. Version 0.6 has just been released and it looks useful for pattern recognition and machine learning work.

Blue Origin apparently had another test flight yesterday. It'd be nice if they weren't so secretive.

The first of ten validation prototypes for the Tesla Roadster has arrived in the U.S. Though they look almost identical to the engineering protoypes, there are a lot of detail differneces and these cars should be very close to the final production configuration.

I attended the Front Range Pythoneers meeting last night. We had two talks that I missed at PyCon. Matt Boersma gave his talk about how to Write Less Code with XRC for wxPython and Sean Reifschneider gave his talk on Python and vim: Two great tastes that go great together. Good talks, but not among my primary interests.

Of more interest to me were Fernando Perez's cmments on IPython and the coming merge of IPython 0.7.4 and IPython1, the version enhanced with Twisted Python to distribute interactive computations over multiple processors. Duncan McGreggor, a Tristed core developer was present and indicated that he plans to attend the IPython1 sprint.

Despite failing to reach the intended orbit, SpaceX had a mostly successful test launch of its Falcon I rocket. CEO Elon Mush said that he expects two more launches this year.

In Simulating with SimPy, Klaus Mueller shows how his Open Source system for Simulaition in Python can be used to build models of a pizza shop. These models may be used to reach a better understanding of the system and make useful infereneces about it.

John Backus, the father of Fortran, has died. Fortran was my first computer language, and while I prefer dyanamic languages like Python, for some computationally intensive tasks Fortran still cannot be beaten.

Open Bayes for Python is an Open Bayes implementaion written in Python rather than C++. Bayesian Networks are useful for reasoning under uncertainty and machine learning. Open Bayes for Python uses Numarray for the heavy lifting, but the project is moving to Numpy for future development.

Since baesian networks are computationally intensive and trivially parallelizable, it will benefit from the IPython1 work.

I recently responded to a call for participation for a Web Survey on the Use of Robots Capable of Lethal Force in Warfare issued by the robotics lab at Georgia Tech. "Robots" are currently used in warfare mostly in combat support roles like reconnaisance and explosive and explosive ordenance disposal, and are controlled by a soldier, sailor, or airman, and they have little or no autonomy. Ths U.S. and other militaries are contemplating more extensive use of unmanned machines in direct combat and these machines will be capable of greater autonomy. Someone will almost certainly do this, and it may be a good thing, but it is somewhat disturbing and requires careful consideration.