AI & Cybersecurity: Your AI Agent Is the Next Phishing Victim

In this article, we’ll explore a specific type of AI-based attack (and remediations) that can be executed using any LLMs capable of coding and used as agents, without the need for Anthropic Mythos or OpenAI Astra.

AI: Debugging a Tiny LLM to Understand What Happens Between a Prompt and a Response

In this article we’ll explore how LLMs (Large Language Models) actually work internally, with practical examples so everyone can understand how they work and behave.

RISC OS: Running ShareFSServer alongside VirtualRPC, RPCEmu and Arculator with a virtual network adapter

In this tutorial, we’ll see how to configure Windows and Linux hosts to run my ShareFSServer together with your favorite RISC OS emulator.

AI: AI Systems Are No Longer Just Models, and Runtime Verification Is Becoming the New Security Boundary

AI systems are increasingly being discussed as if the model were the system. That was never entirely true, but it is becoming dangerously misleading. Moreover, modern AI applications are not simply prompts sent to a model and responses returned to a user. They are pipelines, wrappers, agents, tools, APIs, third-party services, memory layers, orchestration logic, and increasingly autonomous or pseudo-autonomous behaviours. So let’s analyse what is going on and where we should focus our efforts next.

DME Project: Porting a lot of Themes to the new ExtTheme!

While the DME project slowed down due the need for a better development environment for which I have designed UltimaVM for, the parts I can continue to work on are still moving forward, here is an updated with pictures of Themes ports to the new DME ExtTheme!

UltimaVM: Update 2025-11-08 – More features added to Ultima – Function pointers, Code sections and more

Given the interest Ultima has been receiving from enthusiasts and engineers, both online and at the few computer shows where I’ve demonstrated it, I think some will appreciate a more detailed look at its ongoing progress. So here’s the first of these updates.

ZFP Systems and the RISC OS Community on GitHub at the RISC OS London Show 2025

Next Saturday (25th October 2025) I’ll be at the RISC OS London Show, together with many other developers and enthusiasts who keep our favourite operating system alive and evolving. Let’s see what to expect!

AI: Why Generative AI Hallucinates. All the Causes You Should Know

Large language models can produce text that looks fluent and convincing but turns out to be wrong; a problem known as hallucination. These errors are not random glitches: they stem from how AI systems are trained, the data they learn from, the way we evaluate them, and even how people interact with them. In this article, we unpack all the main causes of hallucinations to help you better understand and use GenAI tools with clearer expectations.

Merlin OS: Building Trust in RISC OS Merlin with Formal Verification Methods

In this article, we’ll take a look at an important new effort I am adding to my operating system, Merlin OS: formal verification, using a set of tools available to all Rust developers, as well as some custom ones.

Software Engineering: Rethinking the Two Most Misused Mantras in Software Engineering

In this short article we’ll see my personal “corrections” over some common advices given in the industry which have been proved to cause more problems than helping people.

AI: The Security Challenges of Liquid Neural Networks

In this second article exploring Liquid Neural Networks (LNN), we’ll take a look at the security challenges behind developing both a foundational library and foundational models based on LNNs.

AI: Liquid Neural Networks in few words

Given all the questions I have received on the matter of Liquid Neural Networks from people interested in my zLNN foundational LNN library, I have decided to write an article to explain LNNs in the hope of helping people to find answers to initial questions.

Software Development: Software Engineering liability

In this article, we explore the complexities of assigning liability to software developers compared to traditional engineers. We discuss the roles of various stakeholders, the impact of management decisions, and the need for a balanced approach to improve software quality and reliability. Discover how categorizing software into critical and non-critical types can help advance the field responsibly.

Let’s write a bytecode interpreter in… many programming languages!

In my career, quite a few of my co-workers have had wild dreams of creating their own bytecode interpreter virtual machine (VM). Some have actually taken the plunge and attempted it. The allure of this magical endeavor still captivates many. And let’s not forget about “that colleague” who’s convinced that everything needs to be rewritten because, apparently, everything is just “wrong”! I guess all these enthusiasts might as well start from scratch with a VM, unless, of course, they also want to design a brand new CPU!

New alpha release v0.9.4 for the CROWler GoLang port

📣 I am excited to announce the new alpha release 0.9.4 of theCROWler (my Content Discovery and Security development platform) GoLang Port available for Testing!