Frankly speaking ...

Introduction: The Smartest Single Cell My interest in slime mould was piqued when I read Active Context Compression: Autonomous Memory Management in LLM Agents and learned that the design drew inspiration from Physarum polycephalum . In maze experiments, the mould prunes branches that do not lead to a reward and reinforces productive paths. That simple strategy maps surprisingly well to how we now think about managing context in large language model (LLM) systems. That connection led me to look more closely at the organism itself. Physarum polycephalum looks, at first glance, like a vibrant splash of yellow paint or a forgotten kitchen spill. Yet this "blob" is a syncytium: a single, massive cell containing billions of nuclei that share one continuous cytoplasm. It might be the world's most sophisticated "brainless" computer. What captivates me is its capacity for primitive cognition. Without a single neuron, it solves mazes, remembers past stimuli, and ...

Since 1956 one of the central goals of AI is achieving Artificial General Intelligence (AGI). Recently we have seen amazing advances in AI, but there are still important challenges to solve before we get there. In this article, I draw on the AAAI 2025 Presidential Panel report to outline what the AI community sees as the critical gaps that must be resolved before AGI can be achieved. AI is in a strange place right now. New benchmark results can make it seem as if the field is nearly solved. Yet these systems still fail basic common-sense tasks that humans manage with little effort. That gap is why true AGI still feels a long way off. In this article, AGI means an AI that can perform as well as a human across a wide variety of tasks, not just produce fluent text. The big problem is what researchers call the "Reasoning Paradox." Today's Large Language Models (LLMs) are very good at producing language that sounds like reasoning, but that is not the same as reliable for...

Years ago, as a DevOps engineer, I managed systems using various automation tools. However, for managing my own workstations, I chose Ansible and found a streamlined way to keep my Debian systems consistent without turning the whole exercise into another full-time job. What I like most about Ansible in this context is that it gives me structure without much friction. I can keep a clear separation between work and home machines; I can decide exactly which roles belong on which hosts through the inventory ; and I get useful safeguards such as ansible-lint and dry runs before I touch a live system. I also find the tooling approachable. The command line is simple, the module ecosystem is broad, and the documentation is good enough that I rarely feel like I am fighting the tool. The project itself is organised around small, isolated roles. Each role lives in its own directory under roles/ , with its own tasks, files, and variables, so a change to something like vim does not leak i...

For some time, I have been following research on LUCA—the last universal common ancestor of all life on Earth. This short article draws mainly on recent work, especially the 2024 paper by Moody and colleagues, as an introduction to current thinking. What makes LUCA so compelling is the scale of the question it represents. This is not just another organism in the history of life, but the most recent point from which all modern biology descends. What seems increasingly clear is that this starting point was not simple in any everyday sense of the word. Many of us picture the earliest life as fragile and barely formed—a transitional step between chemistry and biology. In that picture, complexity emerges slowly over immense stretches of time. Current evidence points in a different direction. LUCA appears to have been a recognisable, free-living microbe rather than a primitive intermediate. It likely possessed core systems we associate with modern cells, including: A genetic code and...

As AI models become more powerful, they are also becoming much more accessible. I have been interested in what happens when you trade a little model accuracy for the ability to run everything on your own machine, without relying on paid cloud infrastructure. For developers, researchers, and hobbyists, that shift is useful for more than just cost. It also changes how private, flexible, and portable these tools can be. This week I spent some time trying two popular ways of running local AI models: Ollama and LM Studio . Both are open source and free to use, and both make it much easier to get started than I expected. What I wanted to understand was not just whether they worked, but how they felt to use in practice and where each one made more sense. What I like about running models locally is that the benefits are immediate. There is the obvious cost saving, especially if you are experimenting often or working through lots of prompts, but privacy matters just as much. When the mode...

Introduction This week, I experimented with Google AI Studio. The objective was to see how difficult it would be to create a web front-end for a word puzzle solver targeting games like the Nine Letter Word , NYT Spelling Bee , and Scientific American Spellements . Developing the Requirements I had several command line programs that solved the problem. Now I want a web front-end. I had used Flutter before and knew it could deliver not only a web front-end but also native desktop and mobile apps. I drafted a requirements document in Markdown and asked Copilot and Gemini to review and suggest improvements. The document covered puzzle-solver inputs, examples, expected behaviour, and UI style. For this solution, I expected a client-server architecture. You can see the initial application requirements document on GitHub. Once satisfied, I submitted these requirements to AI Studio. First Iteration The first iteration produced a functional website. AI Studio provided a mock solver ...