"We study instabilities and pattern formation in reaction-diffusion layers that are diffusively coupled. For two-layer systems of identical two-component reactions, we analyze the stability of homogeneous steady states by exploiting the block symmetric structure of the linear problem. There are eight possible primary bifurcation scenarios, including a Turing-Turing bifurcation that involves two disparate length scales whose ratio may be tuned via the inter-layer coupling. For systems of $n$-component layers and non-identical layers, the linear problem's block form allows approximate decomposition into lower-dimensional linear problems if the coupling is sufficiently weak. As an example, we apply these results to a two-layer Brusselator system. The competing length scales engineered within the linear problem are readily apparent in numerical simulations of the full system. Selecting a $sqrt{2}$:1 length scale ratio produces an unusual steady square pattern."

"It is widely accepted that population genetics theory is the cornerstone of evolutionary analyses. Empirical tests of the theory, however, are challenging because of the complex relationships between space, dispersal, and evolution. Critically, we lack quantitative validation of the spatial models of population genetics. Here we combine analytics, on and off-lattice simulations, and experiments with bacteria to perform quantitative tests of the theory. We study two bacterial species, the gut microbe Escherichia coli and the opportunistic pathogen Pseudomonas aeruginosa, and show that spatio-genetic patterns in colony biofilms of both species are accurately described by an extension of the one-dimensional stepping-stone model. We use one empirical measure, genetic diversity at the colony periphery, to parameterize our models and show that we can then accurately predict another key variable: the degree of short-range cell migration along an edge. Moreover, the model allows us to estimate other key parameters including effective population size (density) at the expansion frontier. While our experimental system is a simplification of natural microbial community, we argue it is a proof of principle that the spatial models of population genetics can quantitatively capture organismal evolution."

"Microblogging services such as Twitter are an increasingly important way to communicate, both for individuals and for groups through the use of hashtags that denote topics of conversation. However, groups can be easily blocked from communicating through blocking of posts with the given hashtags. We propose #h00t, a system for censorship resistant microblogging. #h00t presents an interface that is much like Twitter, except that hashtags are replaced with very short hashes (e.g., 24 bits) of the group identifier. Naturally, with such short hashes, hashtags from different groups may collide and #h00t users will actually seek to create collisions. By encrypting all posts with keys derived from the group identifiers, #h00t client software can filter out other groups' posts while making such filtering difficult for the adversary. In essence, by leveraging collisions, groups can tunnel their posts in other groups' posts. A censor could not block a given group without also blocking the other groups with colliding hashtags. We evaluate the feasibility of #h00t through traces collected from Twitter, showing that a single modern computer has enough computational throughput to encrypt every tweet sent through Twitter in real time. We also use these traces to analyze the bandwidth and anonymity tradeoffs that would come with different variations on how group identifiers are encoded and hashtags are selected to purposefully collide with one another."

"Motivated by the amount of code that goes unidentified on the web, we introduce a practical method for algorithmically identifying the programming language of source code. Our work is based on supervised learning and intelligent statistical features. We also explored, but abandoned, a grammatical approach. In testing, our implementation greatly outperforms that of an existing tool that relies on a Bayesian classifier."

Now that's a Robotic Overlord I can happily play minion to.