Leon Furze

Synopsis

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. Narayanan and Kapoor’s framework identifies three speeds of progress: invention (research breakthroughs, currently very fast), innovation (developing products like ChatGPT, moderately slower), and adoption/diffusion (broad societal integration, measured in decades). Furze illustrates this with internet history: the World Wide Web launched in 1993, but only 14% of US adults had internet access in 1994, reaching 50% by 2000, and taking over thirty years for many schools to achieve consistent, meaningful use. He traces edtech history from Sidney Pressey’s 1920s Automatic Teacher through Skinner’s teaching machines to modern analytics platforms, noting how applications like Edmodo come and go. Rather than treating GenAI as unprecedented, Furze suggests viewing it as “normal edtech” operating on a 10-20 year adoption timeline, which may reduce panic and allow more deliberate, thoughtful integration.

Originally published at: https://leonfurze.com/2025/10/21/genai-is-normal-edtech/

Links

Synopsis

In Processes are More Important than Prompts, Leon Furze argues that “prompt engineering” has become less important as GenAI technologies have matured. His early 2023 article focused heavily on crafting perfect prompts, but technological improvements—multimodal inputs (PDFs, images, spreadsheets), internet access, thinking models, and research applications—have reduced the need for lengthy contextual prompts. Instead, he introduces “The Practical AI Process,” a five-step framework applicable across platforms. Step one: Apply your expertise by leaning into domain, technological, and situated knowledge before using AI. Step two: Select the right model for the job—understanding when to use fast models versus thinking models, Deep Research, Canvas/Artifacts, custom chatbots, AI agents, or local models. Step three: Add context through various file formats including documents, images, audio, and video. Step four: Use the internet by enabling web search (enabled by default in some platforms, toggleable in others). Step five: Iterate and refine through dialogue with the chatbot and potentially bouncing between platforms for different parts of a task. Furze emphasizes there’s no magic formula or acronym for prompting—the chatbot interface itself is inadequate—but this systematic process offers more control over outputs across text, image, audio, and video generation.

Originally published at: https://leonfurze.com/2025/11/03/processes-are-more-important-than-prompts/

Links

Synopsis

In What Happens When the AI Bubble Bursts?, Leon Furze argues the current GenAI bubble will inevitably burst, much like the dot-com bubble of the late 1990s. He highlights the absurdity of the situation—Microsoft “investing” in OpenAI with server access that both companies then book as revenue, while in reality people are refusing AI, experiencing reduced productivity from “workslop,” and companies seeing little return on investment. The paradox is that the technology actually works: Furze uses GenAI for administrative tasks, and image, video, and audio generation have reached human-comparable quality. But “it works” isn’t enough—investors need promises of revolutionary change and huge returns, so the technology gets hyped beyond reasonable expectations. When the bubble bursts, Furze predicts both negative residue (fraud revelations, court cases, economic damage) and positive residue (cheap GPUs, skilled statisticians in a buyer’s market, optimized open source models). Like the dot-com bust that led to Web 2.0, useful technologies will remain: LLMs will continue to be useful though perhaps peaked, and generative tools will recede into the woodwork as mundane, ordinary technology. Furze concludes with his flash fiction piece “Afterglow,” imagining a future where GenAI components are “harvested” for practical applications like medical imaging and real-time audio processing, leaving something greater than chatbots behind.

Originally published at: https://leonfurze.com/2025/10/17/what-happens-when-the-ai-bubble-bursts/

Links

Synopsis

In Technology is Actually Pretty Great, Leon Furze argues that the problem isn’t technology itself but corporate capture of digital life in education. Returning from a hiatus spent finishing his PhD and developing a new course, he reflects on how OpenAI’s problems—privacy, sustainability, copyright, labour exploitation—are symptoms of the broader tech industry. Furze positions himself as a technology enthusiast who grew up with dial-up internet and IRC, arguing that loving technology is a prerequisite for being a good tech critic. He points out that digital platforms come and go (MySpace, GeoCities, MSN Messenger) but education treats corporate tools like permanent infrastructure. The real issue is that tech companies have convinced education that “digital literacy” means proficiency in their specific platforms—Google doesn’t want students who understand algorithms, they want students who know how to Google. This brand loyalty masquerading as education serves corporate interests while sidelining genuine critical engagement with technology. Furze rejects blaming education for being “too slow” and instead argues we should reject the myth of inevitability, explore open-source alternatives, question whether we need cloud services for everything, and teach students that the online world doesn’t have to be owned by five companies. Technology is great, he concludes, but only when we shape it rather than letting corporations shape us.

Originally published at: https://leonfurze.com/2025/10/06/technology-is-actually-pretty-great/

Links

Synopsis

In What Do Educators Want to Learn About GenAI?, Leon Furze shares findings from a survey of approximately 3,000 educators across K-12, early childhood, universities, TAFEs, and related industries spanning multiple countries including Australia, the US, UK, Europe, Argentina, South Africa, and Denmark. Despite survey fatigue, he received hundreds of detailed, passionate responses revealing five main themes. First, educators want to know how students can use AI without losing critical and creative thinking skills—Furze argues these skills are innate and only threatened if we buy into tech company narratives of inevitability. Second, there’s strong interest in learning and teaching AI ethics, with over half the responses requesting topics on bias, environment, and the digital divide between premium and free AI systems. Third, educators want to create high-quality resources using AI without being replaced by low-quality auto-generated lesson plans from companies like Google. Fourth, there’s widespread concern about rethinking assessment, with many attempting frameworks like the AI Assessment Scale but hitting institutional walls. Fifth, educators consistently asked where they’ll find time to learn all this while developing technological, domain, and situated expertise. Respondents indicated they want online, asynchronous, bite-sized PD that’s self-paced but scheduled, actionable, and ideally completed during office hours.

Join the mailing list and stay up to date with new episodes: https://subscribe.leonfurze.com

Originally published at: https://leonfurze.com/2025/09/08/what-do-educators-want-to-learn-about-genai/

Links

Synopsis

In Professional Development for AI in Schools: A Three-Dimensional Approach, Leon Furze presents a framework for supporting teachers in developing expertise with GenAI. He opens with three vignettes: an experienced literature teacher with strong domain and situated expertise but no technological knowledge; an industry-trained maths teacher with technological and domain expertise but lacking classroom experience; and an out-of-field early career teacher using AI to generate lessons without the expertise to judge quality. Furze argues schools need professional development that addresses all three dimensions—technological expertise (understanding AI’s ethical and practical implications), domain expertise (disciplinary knowledge and curriculum design), and situated expertise (classroom dynamics and professional identity developed over time). He proposes a three-stage PD pathway using the Dreyfus model of skill acquisition: Stage One (Explore) involves auditing staff skills and hidden expertise; Stage Two (Design) maps staff capabilities to identify clusters and professional learning communities; Stage Three (Lead) identifies emerging leaders across dimensions and creates structures for ongoing development. The model respects teachers’ right to resist AI when appropriate while ensuring all staff have sufficient competence to discuss it transparently with students.

Originally published at: https://leonfurze.com/2025/09/01/professional-development-for-ai-in-schools-a-three-dimensional-approach/

As internet search gets consumed by AI, it’s more important than ever for audiences to directly subscribe to authors. Mailing list subscribers get a weekly digest of the articles and resources on this blog, plus early access and discounts to online courses and materials.

No data sharing, no algorithms, no spam. Unsubscribe any time.

Links

Synopsis

In Artificial Intelligence in Vocational Education, Leon Furze reflects on his experience at the Future Skills Organisation National Forum in Canberra, arguing that the Vocational Education and Training (VET) sector is ideally positioned to meaningfully integrate GenAI. The VET sector serves 5 million Australian students annually across diverse industries—from trades to allied health to creative arts—and roughly two-thirds of Australians work in small to medium enterprises where GenAI can be genuinely useful for administrative tasks. While updating VET training packages is notoriously slow, Furze highlights calls from sector leaders to move quickly on skills training while being more measured on regulation. He argues that VET has more assessment flexibility than K-12 or higher education, relying heavily on observations, simulations, role plays, and practical assessments that are naturally less vulnerable to AI interference—you can’t use ChatGPT to fix a circuit or support a special needs child. VET teachers could play a crucial role in helping time-poor small business owners and tradespeople understand AI’s practical applications, particularly for ESL students and literacy support. Furze is optimistic that with quality professional development for teachers and accessible resources for SMEs, the VET sector can make the best use of AI technologies among all education sectors, focusing on real-world applications rather than hype.

Originally published at: https://leonfurze.com/2025/08/25/artificial-intelligence-in-vocational-education/

Links

Synopsis

In Five Principles for Rethinking Assessment with Gen AI, Leon Furze argues that assessment reform should focus on good pedagogical practice rather than policing AI use. While the AI Assessment Scale has become widely adopted, Furze emphasizes that the principles behind it—validity, authenticity, transparency—were important long before ChatGPT.

His five principles are: (1) Validity First—design assessments that generate trustworthy evidence of learning through content, construct, and consequential validity while balancing security with accessibility; (2) Design for Reality—create authentic assessments that reflect real-world workflows where AI may legitimately be used, and acknowledge the “brutal reality” that GenAI is more capable and harder to detect than most educators think; (3) Transparency and Trust—communicate clear expectations about AI use, provide teachers with substantive professional development, and trust that students want guidance and will generally choose to do the right thing; (4) Assessment is a Process, Not a Point in Time—move away from one-shot high-stakes assessments and challenge the “stuck thinking” that end-of-school exams should dictate all upstream assessment; and (5) Respect Professional Judgement—resist rigid rules and surveillance technologies, trust teacher expertise over time, and ensure assessment doesn’t create unsustainable workload.

Furze concludes these principles are pro-learning, not anti-AI, and reflect practices that matter regardless of whether students use generative AI.

Originally published at: https://leonfurze.com/2025/08/18/five-principles-for-rethinking-assessment-with-gen-ai/

Links

Synopsis

In About, With, Through, Without, Against: Five Ways to Learn AI, Leon Furze pushes back against simplistic binaries about whether AI helps or harms learning. He argues that young people are adaptable and capable, and proposes five overlapping approaches to thinking about AI in education. Learning about AI covers AI literacy—how to use and understand the technology—but Furze warns against making every teacher responsible for teaching AI skills. Learning with AI includes tutor chatbots and using AI for feedback, though he remains skeptical about personalized learning claims. Learning through AI treats the technology as a conduit or medium, like semantic search engines. Learning without AI acknowledges that fundamental skills and disciplines remain vital regardless of technology. Learning against AI encompasses critical awareness, disruption, and understanding why students should actively resist problematic aspects like bias, surveillance, and environmental harm. Furze argues that learning against AI is where the most interesting classroom work happens—analyzing deepfakes in English, exploring algorithmic discrimination in math and science, or examining why AI companies are being sued by artists. He concludes that the key is balance and nuance, not outright rejection or uncritical embrace.

Originally published at: https://leonfurze.com/2025/08/12/about-with-through-without-against-five-ways-to-learn-ai/

Links

Synopsis

In Open Source AI is Going Mainstream, Leon Furze explores the growing significance of open source artificial intelligence beyond the GPT-5 hype cycle. He begins by clarifying what “open source AI” actually means, noting that the term has become contentious—many so-called “open source” models from companies like Meta are more accurately described as “open weights,” since their underlying architecture, training processes, and datasets remain proprietary. True open source projects expose their entire codebase to public examination and adaptation.

Furze highlights the rising geopolitical importance of open source AI. The European Union is recommending member states prioritize open source to reduce dependence on US big tech, with German government institutions and Switzerland leading the charge. China has released large-scale open source models like DeepSeek, and even the US AI Action Plan now recommends focusing on open source to compete globally.

Originally published at: https://leonfurze.com/2025/08/11/open-source-ai-is-going-mainstream/

Links