I’ve spent the past few days reading 𝘛𝘩𝘦 𝘖𝘯𝘦-𝘏𝘶𝘯𝘥𝘳𝘦𝘥-𝘗𝘢𝘨𝘦 𝘔𝘢𝘤𝘩𝘪𝘯𝘦 𝘓𝘦𝘢𝘳𝘯𝘪𝘯𝘨 Book by Andriy Burkov, and I’m genuinely glad I did. This is one of those rare books that manages to strike what feels like a 𝗽𝗲𝗿𝗳𝗲𝗰𝘁 𝗯𝗮𝗹𝗮𝗻𝗰𝗲: it covers a remarkably wide range of machine learning algorithms and techniques in a very short space, explains them in a clear and engaging way, and yet 𝗻𝗲𝘃𝗲𝗿 𝘀𝗮𝗰𝗿𝗶𝗳𝗶𝗰𝗲𝘀 𝗿𝗶𝗴𝗼𝗿. The mathematical foundations are always there — equations included — but introduced only when they are truly needed.

I’ve recently been looking for a dataset to evaluate some TSP algorithms. Naturally, the TSPLIB benchmark came to mind. However, one quickly realizes that the TSP instances in this library are stored in a proprietary format. There’s an XML port of these files, but that format has long been superseded by JSON.

The term “filter bubble,” coined by Eli Pariser, captures how social platforms can trap us in narrow content silos. YouTube is no exception: its algorithm often serves up videos from the same handful of channels we already watch—boosting our dwell-time and, ultimately, the platform’s ad revenue.

In this article, we’ll walk through the implementation of a one-dimensional, binary-state cellular automaton built on top of the parametric CA framework I introduced in a previous post. The single-page app is split into two pieces: the cellular-automaton engine itself, and a graphical user interface (GUI). The GUI includes two PixiJS renderers—one for the rule and one for the grid.

Cellular automata are mathematical models built on a finite population of cells. Each cell holds a state selected from a predefined set—finite for discrete automata and potentially continuous for analog variants. In the classic binary case, every cell is either ALIVE or DEAD. At each time step the system advances in lockstep: the new state of every cell is computed by applying a simple local rule that considers the current state of the cell and those of its neighbors. Repeated iterations of this rule generate successive “generations,” allowing intricate global patterns to emerge from straightforward local interactions.

Promptheon is an LLM-powered system for exploring and structuring knowledge about ancient deities from Wikipedia. The project begins by crawling the List of Deities category page, using a Gemini-based language model to classify portal links by cultural origin or divine role.

A Thompson Sampling algorithm using a Beta probability distribution was introduced in a previous post. The Beta distribution is well-suited for binary multi-armed bandits (MABs), where arm rewards are restricted to values of 0 or 1. In this article, we introduce an alternative MAB sampling algorithm designed for the more general case where arm rewards are continuous: Thompson Sampling with a Gaussian Distribution (TSG).