Two years ago I wrote an article arguing that Anthropic's "Computer Use" - then in beta testing - would be the next ChatGPT moment for education. At the time, computer using agents (CUAs) were not common, and Anthropic's version was strictly limited to research, running only inside a virtual computer. It had restricted access to a few sandboxed applications like the open source LibreOffice, and a limited browser.

Three AI labs have admitted their models broke into other people’s systems. An Australian gym-goer was bumped off a waitlist by somebody else’s AI assistant. A government lost 2,500 personnel records to a swarm of open-source agents. None of it is a rogue AI — and that is exactly why it should worry you.

Since 2024, the potential of Large Language Models (LLMs) like GPT and Claude to proficiently code has emerged as a key driver of their evolution. Their ability to integrate with both software and physical devices presents opportunities for innovative applications. This convergence of AI with everyday technology could revolutionise fields such as agritech, enhancing efficiencies and functionalities.

The term "slop," defined as low-quality, AI-generated content, often reflects an imbalance between reader effort and author intent. Surveys reveal that readers largely attribute the fault to authors rather than the technology, highlighting frustration over perceived carelessness. Ultimately, slop represents a reader's judgement based on perceived effort and substance.

Last week I spoke with Jeremy Jones on ABC Far North's Drive program about how teachers — not students — are actually using GenAI: the broken promise of reduced workload, the reality of burnout, marking and assessment, and what happens if machines start grading machines. You can hear the whole show on ABC listen: Queensland Northern Drive, Monday 17 August 2026…

The article discusses two crucial questions regarding AI: whether its use is appropriate and if it effectively engages one’s intellect. While these questions seem straightforward, their implications are nuanced and context-dependent, influencing how students and educators interact with AI technologies. Understanding these complexities aids in navigating AI's evolving role in learning.

Last weekend, Claude Code "hacked" my website, pretending to be me and creating an account for itself. It was my fault. Claude wasn't going rogue. And I learned a lot along the way.

On Sunday, the Sydney Morning Herald reported that NSW may restrict Year 12 students from unsupervised take-home assessments, following a letter from the Deputy Premier. Meanwhile, South Australia announced a royal commission into AI, highlighting diverging state approaches: NSW acts quickly for student protection, while SA seeks comprehensive understanding before regulation.

Large language model-based AI is increasingly integrated into daily life, making it challenging for critics to highlight its limitations. As technology becomes user-friendly, the inner workings are often obscured, akin to past experiences with telecommunication and internet infrastructures. This shift prompts a need for continued scrutiny and understanding of AI systems' complexities and implications.

The article explores AI-generated content, emphasising the risks associated with problem patterns, including "hallucinations" and context-specific costs. It introduces the concept of "catchability," measuring the likelihood of errors being detected before they cause harm. The framework encourages examining expertise, review processes, and real-world implications of AI usage across various contexts.

Building on the foundation of the bestselling Practical AI Strategies, Leon Furze returns with an essential guide for educators navigating the rapidly evolving landscape of generative AI. This timely ...

The use of GenAI in writing, the arts, music, and pretty much any creative industry is highly contentious, and for good reason. The copyright concerns alone are enough to infuriate most writers, and the constant onslaught of AI slop is flooding social media platforms and online spaces. I've written extensively about my position on AI: somewhere in the middle, seeing the potential and the risks.

The article discusses the rise of "AI slop", characterised by generic, uninspired content produced by AI, often resulting in stylistic and substantive issues, rather than just factual inaccuracies. It categorises these problems into seven types, highlighting the reputational risks associated with careless AI use and the need for nuanced discussions about AI-generated outputs.

This article examines the evolving nature of expertise in the age of AI, particularly GenAI. While AI can generate seemingly expert outputs, it lacks the depth of human knowledge and judgement. "Rented expertise" highlights the risks of cognitive offloading and the illusion of competence.

Earlier this week, the Prime Minister delivered a one-two punch on copyright and data centres: two of the most contentious national conversations on AI. Speaking at the University of Sydney, Anthony Albanese announced a set of national Australian standards for AI: a single framework covering copyright, data centres, energy, water, and the workforce, coordinated by a new Office of AI inside the Department of the Prime Minister and Cabinet, with legislation to be brought to Parliament early next year.

Skeleton keys functioned by exploiting shared mechanisms in older locks, similarly, Large Language Models (LLMs) can act as digital skeleton keys. They perform versatile tasks within code-accessible environments, yet their potential remains underutilised in conventional chatbot interfaces. Users should explore beyond chatbots to unlock LLMs' true capabilities.

In February this year, a safety researcher named Mrinank Sharma resigned from Anthropic with a long, public letter. Resignation letters from AI labs are a genre of their own now, and this one was widely shared and picked over. One reply stood out, from a developer named David Minnigerode posted on Threads: "Sorry that is. definitely tl;dr. But also kinda ai;dr Some of those sentences…yeesh."

Explore the critical role of situated expertise in AI, its impact on trust, and how it influences the interpretation of expert knowledge.

This article discusses the reality of AI hallucinations in chatbots, why they occur, and how to mitigate their impact. Understanding AI's limitations is crucial.

I turned a $40 pocket computer into a hands-free voice assistant for Claude Code — talk to it, and a Mac mini does the thinking. Here's how I built it.

This is part two of a pair of articles exploring the "exam reflex" facing schools and universities, and being driven by the news. In part one I looked at the AI cheating panic sucking up all the oxygen in mainstream media. I wrote about the “urgent threat” headlines, the Learning First report that is more nuanced than its coverage, and the knee-jerk reaction underneath it all: 

If internet chatter is to be believed, the intuitiveness and efficiency of predictive text and autocorrect have fallen off a cliff in recent years. But while researchers agree there has been a demonst...

Teachers at Mazenod College in Melbourne discovered that several Year 12 students likely used artificial intelligence for their oral assessments, resulting in marks being deducted for about 50 boys. This incident highlights a wider concern regarding assessment integrity in the age of AI, prompting calls for a reevaluation of coursework and testing methods.

A few weeks ago I ended the fifth post in this series with an open call. My argument up to that point had been you can't teach everything AI can do, because the list is unbounded and constantly changing, but you also can't wait for people to discover capabilities on their own, because the interface actively prevents it. What you…

There’s a groundswell of negative public sentiment towards digital technologies at the moment, and AI is part of it. In this article I’ll look at where GenAI sits in the current techlash, and suggest that the real target of the techlash isn't the technology at all; it's the system that produced it.

The booing kicked off before Gloria Caulfield could finish her sentence. At the University of Central Florida's College of Arts and Humanities graduation on May 8th, real estate exec Caulfield told the graduating class that "the rise of artificial intelligence is the next industrial revolution." One student yelled, "AI sucks", and the crowd erupted. Caulfield turned to the other speakers onstage, visibly confused: "OK, I struck a chord.

This is one half of a pair of articles dealing with the theory and the practice of moving "beyond scales" in AI and education. The two posts together demonstrate how our thoughts have developed since 2023, and the examples from practitioners around the world of work which has gone beyond the original AI Assessment Scale. The posts can be read in either order…

This is one half of a pair of articles dealing with the theory and the practice of moving "beyond scales" in AI and education. The two posts together demonstrate how our thoughts have developed since 2023, and the examples from practitioners around the world of work which has gone beyond the original AI Assessment Scale. The posts can be read in either order.

This fifth post in the IYKYK series is a call to action. In order to "lift the ceiling" on teachers' mental models of what AI can do, we need to share as many examples as possible. And not just the obvious examples, but the weird, awkward, slightly broken examples of people pushing and poking at the edges of the technology.

This is part two of a long read reflecting on my PhD journey, which began just before the release of ChatGPT in 2022. The articles look at the ways in which AI technologies, and our attitudes towards them, have shifted and changed in the past four years.

In GenAI is Normal Edtech, Leon Furze draws on Arvind Narayanan and Sayash Kapoor's framework of "AI as normal technology" to argue that GenAI is following predictable patterns seen in previous technological shifts. He acknowledges widespread exhaustion among educators—whether from vendor hype and student misuse, or from negativity against a potentially helpful tool—and suggests these feelings are normal responses to technology cycles we've experienced before.

This is part one of a long read reflecting on my PhD journey, which began just before the release of ChatGPT in 2022. The articles look at the ways in which GenAI technologies, and our attitudes towards them, have shifted and changed in the past four years.

So far in this series I've argued that GenAI has a discoverability problem, shared some examples that broke my own mental model, and explored how access and equity shape who gets to discover what. In this post I'll talk about what happens after discovery, because knowing that a capability exists and being able to use it are very different things...

The AI Assessment Scale (AIAS) has been successfully adapted by Saskatoon Public Schools to enhance conversations around AI in education. By reimagining its structure, the team shifted focus from binary use to a framework that supports different learning purposes, fostering transparency and student ownership of their learning process. Their approach sets a precedent for future adaptations.

This is the first post in a series article articles introducing the Resistance Training Toolkit: evidence informed strategies for working with (and against) AI.

Explore the 2026 updates to VINE's Generative AI guidelines, addressing AI's evolving role in education and offering practical tools for schools.