We have a wealth of cool science announcement to round off the year with.

• We’ll start with the announcement about how two separate teams have successfully entangled pairs of calcium monofluoride molecules. This is exciting in the context of quantum computing in which the basic unit of computing is a qubit. The question is what do we make qubits out of, physically? This is the first time that a qubit has been constructed out of pairs of single molecules and will for obvious reasons make it much more feasible to scale a quantum computer to a useful level. Of course so far this news is only about creating the entangled pairs which is a long way from an actual quantum computer.
• Next is a product that I’d actually want to use myself if it really works. A new company claims that they have a modified strain of bacteria that when introduced into a person’s mouth, will eventually outcompete the other native bacteria, making it the dominant strain in your oral microbiome. What makes their strain different is that it lacks the gene to create lactic acid from breaking down sugars and it is this acid that breaks down tooth enamel and causes cavities. Their version breaks down sugars into a small amount of alcohol instead. Theoretically a single application would last a lifetime and significantly reduce or perhaps even negate the need to brush teeth or visit a dentist. Incredibly the science behind it has been known since 2000 but it has never been made commercially available to the general public. The usual caution applies in that all this is claimed by the company itself and will need to be independently verified.
• Next are a couple of papers in the field of AI. The first of these describe an autonomous laboratory that successfully synthesized novel materials, specifically inorganic powders. The platform integrates machine learning models to propose synthesis recipes to attempt, a robotic laboratory that carries out the recipe, and characterization of the result through X-ray diffraction. When the recipe fails, the system is able to propose improved follow-up recipes to try. The system was able to create 41 novel compounds over 17 days of continuous operation. I’m sure how useful these novel compounds are or how this compares to the productive output of traditional laboratories staffed by humans. But it’s proof that AI-led scientific development is a real thing.
• The other one is a deep dive into a research project by Anthropic to understand what actually goes on inside a neural network. As we all know, these networks are essentially black boxes. We feed it an input and we get an output but it’s impossible to understand how it arrived at that output. So Anthropic built a very small and simple network and with the help of another network built an interface to understand and break down the first network into its smallest components. In particular, the smallest component isn’t a single neuron because the activation of a given neuron might have a specific meaning but that same neuron activated in combination with some others might mean something else entirely. The paper really only for the experts but it’s a fascinating attempt at looking inside the black box.
• Finally we end with a weird bit of science news to prove that reality is crazier than the imagination of any science-fiction writer. This article talks about the Japanese green syllid worm, a species that reproduces through a process called stolonization. This means that its rear end, or butt, can detach from the main body to swim away and seek out other detached rear ends from other worms to mate with. To make this possible, the rear end of the worm has its own set of eyes, antennae and a simple brain. The article goes on about the genes that regulate the development of this strange organism and speculate that at some point in its evolution, it started to develop a second head further down its body leading to this form of reproduction.