In a few years, buying a robot might feel like buying a car, you go to Tesla (or another brand), pick your model, sign the contract… and a shiny humanoid rolls into your house ready to work, dance, cook, or clean, powered by serious AI. But that’s only half the fun. The real chaos starts when people begin “modding” these robots the way gamers mod PCs. Picture this: you order a legit humanoid robot from Tesla, then jump on Alibaba at 3 a.m. and fill your cart with random “upgrades”, extra sensors, weird robotic arms, laser pointers, microphones, 3D-printed grippers, maybe a sketchy camera module that ships with no manual and broken English in the listing then you go to YouTube, search “How to add a third arm to Tesla Bot,” and follow a 27‑minute video from a guy in his garage who swears “this is totally safe, I’ve only blown the fuse twice.” You open access panels, plug in mystery cables, zip-tie brackets where they definitely weren’t meant to go, then flash some community-made firmware that unlocks hidden features the official company doesn’t want to support. Now your “normal” home robot has a DIY arm that hands you snacks, controls your smart lights, or sprays water at your friends when they walk into your room.This future would create a new culture. People flexing their modded robots in videos the way they flex custom keyboards, communities running robotics build nights where everyone brings their bots and weird Alibaba parts. A black market of “do not try this at home” mods coming soon! -Jordi Lopez

My work involves lots of digital signal processing, frequency-specific filters and whatnot, and occasionally machine learning. I was looking at cepstrograms, and wondering if there's another way to look for oscillations layered on oscillations. I finally found some documentation on wavelet transforms, generalizing the Fourier transform by leaving some uncertainty in both frequency and time, with band resolution varying by frequency. Wavelet techniques are popular for image compression, and now recognition, as machine learning algorithms decompose images in order to identify important components. A popular neural network for such tasks is the convolutional neural network (CNN), which adjusts scalar weights to extract information from images in order to assign labels to the input. There was a beautiful insight in recent years to use the cascading structure of a CNN to iteratively apply filters to different regions of the signal, namely decomposition by wavelet convolution. Different layers of scalar 'neurons' correspond to different categories of resonances in the input signal, and more weights per layer yields sharper resolution. Remarkably, neural networks were originally developed to model human thought. This effectively means that the most efficient algorithm for time-frequency analysis is based on an abstraction of how people think. So, I've been thinking about the duality of time-frequency analysis and its applications, and considering the possibility that a mind, regardless of its material substance, could have varying levels of awareness of different components of reality. It is conceivable that such a cascading process could occur emergently in some natural systems, and perhaps can't not emerge. That is, the mechanical equivalent of thought may come to exist in systems we wouldn't consider sentient, much less alive. Neither the material substance nor the dimensional presence need be constant throughout such an experience; I think of the CKM or PMNS matrix. It sounds trite, by way of spiritual awareness and relationship with the Divine, but to understand the way broadly is to see it in all things.