Skip to main content

Work

Selected work, in depth.

Platform

The Digital Research Hub

A school-wide learning platform at St Andrew's Anglican College. K-12, curriculum-mapped, with embedded research pathways and AI literacy. Available across the College. Excellence Awardee for Best Use of Technology at the 2026 Australian Education Awards.

The Digital Research Hub homepage at St Andrew's Anglican College

30+

Original courses and modules

K–12

Curriculum-mapped

2026

Australian Education Awards

We wanted them to be the kind of thinkers who know how to interrogate any source and hold it to account.
Karen Gorrie, Principal, St Andrew's Anglican College · quoted in Tech Business News, May 2026

Inside the platform

Research hub view Skill modules and frameworks Staff dashboard Interactive map view
Read the case study

Context

Before the Hub, the school was using LibGuides. It was old, text-based, and out of date. The link between the library catalogue and the actual research students were doing had broken. The databases attached to it sat unused. What had been a research tool had become a graveyard of links and word vomit.

Approach

I pulled the usage stats and confirmed what the surface already suggested: nobody was opening it. Not the LibGuides, not the databases. That was the opening. Knowledge and data don't sit on a page anymore. I designed something that worked for teachers and students at the same time, built from research backwards instead of links forwards.

What I built

I researched, designed, and built the platform. The two Research Hubs and the five Skills Modules, the structural backbone of the site, are entirely mine. The subject-specific hubs were collaborative: I worked with teachers who wanted updated content or interactive labs for their units. The IT team supported deployment and integration.

Two structured research pathways guide students through defining questions, locating credible sources, evaluating information critically, and constructing evidence-based arguments. AI tools sit inside that framework, not above it. Five thinking-skill modules run through everything: Inquiry and Investigation, Systems Thinking, Evidence and Data Literacy, Critical and Ethical Reasoning, and Communication.

Outcome

The platform shipped. K-12, curriculum-mapped, embedded research pathways and AI literacy. Available across the College. Recognised at the 2026 Australian Education Awards as an Excellence Awardee for Best Use of Technology. But the more telling outcome is behavioural.

Teachers ask for custom tools or items to be added, and because it's built in-house, they can see those changes ship.

Students who find the platform get fascinated. Both behaviours are downstream of ownership. A vendor product doesn't allow that loop.

Reflection

If I started over, I would uniform the whole thing from day one. This was my first big project. I expected a longer trial before live use, and the trial folded under demand before I had a chance to standardise. The platform sprawled, and retrofitting uniformity after the fact is much harder than designing it in from the start. The next thing I build, I build the system before I build the surface.

Framework

The VERIFY Framework

Stop typing at AI. Start thinking with it.

A six-step thinking cycle for navigating AI in classroom work. Designed for students. Usable by anyone who has to stand behind what they say.

Validate the Task

Before prompting, establish what the task actually requires. The prompt is where the thinking starts.

Examine the Output

Fluency is not accuracy. Interrogate claims, check sources, and ask what the output is assuming.

Reflect on Fit

The output looking finished is not the same as the thinking being done. Does this serve your task?

Investigate Further

Every source leaves something out. Go looking for what that is. Voices, context, missing evidence.

Filter for Quality

You cannot filter without a standard. Apply check questions before you stand behind what you have.

Your Own Judgement

You cannot defend a judgement you did not make. Y is where you find out whether you made one.

AI does not make students worse thinkers. The absence of explicit teaching does.

VERIFY is the framework that puts the teaching back in.

Read the case study

Context

AI predicts. It does not think. It produces fluent, confident, well-structured answers without ever knowing if any of it is true. And students are reading that performance as understanding.

The conversation about AI in schools has been stuck on cheating. The real risk is quieter. Students stop reaching for their own thinking, because something else got there first. The problem is not that AI is in the room. The problem is that no one has been taught how to stand next to it.

Approach

VERIFY is a six-step thinking cycle. Not a checklist. Not a rule. A way of moving through AI work that turns autopilot into ownership.

Each letter is a question students can ask before, during, and after they use AI. The aim is not to slow them down. The aim is to put their thinking back at the centre of the work and keep it there.

What I built

Six letters. Six moves. A poster for the wall, prompts for teachers, and subject-specific adaptations that drop into the task you already teach. English. Science. History. Media Studies. Health. Same framework, different surface.

VERIFY runs across the lifecycle of a task. Validate before prompting. Examine and Reflect during the work. Investigate, Filter and stand behind Your own judgement before anything goes in.

It maps cleanly onto the Australian Framework for Generative AI in Schools. It also gives teachers something the framework alone does not: a shared language they can put into a student's hands on a Monday morning.

Outcome

The work shifts. From extraction to evaluation. From paraphrasing to probing. From collecting to connecting. Students stop asking "what did it say" and start asking "is this any good, and can I defend it".

It's not about who wrote the first draft.

It's about whether you can explain why it's worth reading.

Reflection

The most common objection is that this is more cognitive load. It is not. It is a reset. Students were already carrying the load of guessing what counts as their own work. VERIFY just gives them a way to answer that question out loud.

The deeper objection is about ownership. If a student verifies AI work, did they write it? The answer matters less than the question, because the question is the thinking. That is the part schools have to teach. That is the part VERIFY teaches.

The article below is the first articulation of VERIFY. The framework has continued to iterate since, in step with what classrooms keep teaching me.

Framework

The EPFL Protocol v2.0

Make the invisible visible.

Five stages for keeping your mind in the game when working with AI. Where VERIFY teaches students to interrogate the output, EPFL teaches them to interrogate the interaction itself.

v1.0 was about safety: don't fall for the polite robot. v2.0 is about strength: don't let the AI make you weak.

Intention

Predictive Calibration

Declare your state before you open the tool. Lost, Partial, or Confident. Each state needs a different approach.

Interaction

The Friction Protocol

Forbid the AI from being polite. If it agrees immediately, that is a red flag, not a reward.

Response

The Fluency Audit

Read your emotion, not just the output. Relief means you got played. Curiosity means your brain is working.

Reflection

Social Triangulation

Take the AI's best output to a real human before you trust it. AI makes disagreement feel optional. They don't.

Academic Branch

Prove You Learned It

If you can't explain it on a blank page without the AI, you borrowed the understanding. You didn't earn it.

Three Layers of Awareness

Running underneath every stage.

These three layers don't turn on and off between stages. They run underneath every interaction you have with AI. Not a checklist, a posture.

Cognitive

Am I thinking better, or avoiding thinking?

How AI affects your thinking process and intellectual growth.

Emotional

Curious, bored, frustrated, or done?

Your emotional state shapes what the AI gives back to you.

Systemic

What forces shape this tool?

AI is not neutral. Who built it, and what were they optimising for?

v1.0 was about safety. v2.0 is about strength.

EPFL is the protocol that keeps your mind in the game.

Read the case study

Context

The conversation about AI in schools kept getting stuck on cheating. Cheating is the symptom. The harder problem is what happens inside every AI interaction long before the output appears.

Help, without awareness, can quickly become a dangerous echo chamber. AI mirrors what you bring to it. If the student is curious, it amplifies curiosity. If the student is angry, anxious, isolated, performing, it amplifies that too, fluently and politely. Cheating was never the core issue. The core issue is that nobody is taught to notice what they are feeling when they prompt.

Approach

Build a model that makes the emotional, cognitive and systemic steps of an AI interaction visible. Not a checklist. A continuous cycle students return to and refine over time. Ethics as lived practice, not policy.

The protocol works for the student in the prompt and for the teacher in the room. Both lenses, same loop. The teacher's role is not to police the prompt. It is to surface insight without creating shame, panic, or defensiveness.

What I built

v1.0 tracked six steps: Feeling, Intent, Prompt, Output, Consequence, Reflection. The student asked: what state am I in, what do I actually want, what words am I using, what came back, what am I doing with it, would I be okay if someone I trusted saw this exchange. The teacher tracked the same six from the outside.

v2.0 collapses the loop into five stages with three layers of awareness running underneath. Predictive Calibration. The Friction Protocol. The Fluency Audit. Social Triangulation. An Academic Branch that closes the loop with: if you cannot explain it on a blank page without the AI, you borrowed the understanding.

The Cognitive, Emotional and Systemic layers do not turn on and off between stages. They run underneath every interaction. Not a checklist. A posture.

Outcome

Students start prompting more slowly and listening to what comes back differently. Relief stops feeling like success. Curiosity stops feeling like a detour. Teachers get a shared language for the part of AI use that used to be invisible.

I am not trying to stop students from using AI.

I am trying to give them a way to think inside the system. To check what they are doing and why. To recognise the emotional weight of their choices.

Reflection

EPFL exists in two versions because the work kept teaching me something. v1 made the loop visible. v2 made the posture visible underneath it. There will be a v3, because students will keep meeting tools I have not seen yet, and the protocol has to meet them there.

The loop is not limited to AI. It applies to every system that shapes how students consume and create information. Social media. Recommendation engines. Newsfeeds. Advertising. If students do not learn to question the feedback they receive, they keep reinforcing patterns that limit their thinking and erode their agency.

The article below is the v1 articulation of EPFL. v2.0 (visible above) is the iteration that came out of using v1 in real classrooms.

Classroom tool

RePrompted

Rules are just drafts. Revisions make the magic.

A set of physical prompt cards for teaching prompting and AI ethics in the classroom. Students draw a card, react, revise, and defend. The cards do the provocation; the conversations do the teaching.

A set of RePrompted prompt-card decks in use at a classroom table.

What it teaches

Prompt engineering without ever opening a chatbot. How small changes in language shift tone, clarity, and power. Bias, agency, and critical trust.

Format

Physical decks for small groups inside an existing lesson. Two pieces: RePrompted (the main deck) and Trained On (a sub-game).

In design & testing
Read the case study

Context

We keep handing students AI tools and asking if they'll cheat. We should be asking if the task was ever worth doing in the first place. With younger learners, launching straight into ChatGPT turns thinking into a transaction. It turns prompting into typing. It turns kids into users before they've ever become thinkers. That's not preparing them for the future. That's skipping the part that actually matters.

Approach

Don't start with the chatbot. Start with the thinking that shapes it. I don't use games to make learning fun. I use them to make learning real. Gamification with intent: to spark energy, deepen engagement, and bring students into the learning together.

The classroom is often the first place students test ideas. I don't want to replace that with a chatbot.

What I built

RePrompted is a card-based tool that teaches students how small changes in language shift tone, clarity, and power. They remix sentences, debate intention, and spot how persuasion works long before AI even responds.

Trained On is a sub-game. Each student acts as an AI trained on a single, biased source. TikTok. War propaganda. 1950s etiquette. They answer prompts from inside that worldview.

Both let students practice prompt engineering without ever opening a chatbot. They are not about prompting. They are about bias, agency, and critical trust.

Outcome

Students lean in. They start asking more. They stop waiting for the answer and start shaping the process. The cards teach what no chatbot can demonstrate as cleanly: the way you frame the question is the work.

It's funny. It's revealing. It lands.

Trained On is the one that surprises teachers most. Watching a kid argue in character as "TikTok" reveals how voices get trained into us, long before anyone says the word algorithm.

Reflection

Different year levels need different approaches. Middle school students need curiosity, connection, and confidence. Shared language before access. Senior students face urgency and need tools, fluency, and good judgment. One deck can't carry both.

The cards are still in development. The fear in classrooms isn't cheating. It's raising a generation that thinks their ideas aren't enough. RePrompted is one small intervention against that.

What readers said back

This is exactly the adoption mistake many institutions make: they treat access as literacy. Giving students a chatbot is not the same as teaching them how systems frame, distort, compress, and reward certain kinds of thinking. The deeper issue is not whether students cheat. It is whether they learn to outsource the messy part of forming judgment before they have built enough judgment to know what they are outsourcing.
LinkedIn response to Don't Start With the Chatbot.

Programme · Sessions and measurable change

AI Literacy Workshop Programme

In-class sessions on prompting and verification, with pre/post measurement of student work. The workshops are the teaching side of the same questions VERIFY and EPFL formalise.

40%

improvement in real student prompt quality after the sessions. Measured from live usage, not a test.

Methodology

  1. Signal. An AI Learning Insights Dashboard surfaces what students are actually doing with AI across subjects. What they do, not what they say they do.
  2. Intervention. In-class sessions on what a good prompt looks like, how AI gets things wrong, and how to verify. Tied to the subjects students are actually working on.
  3. Effect. Prompt quality climbs 40% after the sessions. Read directly from real usage, not a parallel task. Behaviour change, captured in real data.
Wide view of the Hackathon room: students at tables with laptops, a large screen displaying the methodology (IDEAS, SKETCH, DESIGN, BUILD, DEMO), framed photographs along one wall.

Kids Vibe Coding Hackathon

First global site of the Worldwide Kids Vibe Coding Hackathon, co-hosted by Peregian Digital Hub and St Andrew’s Anglican College. Nine students aged 9 to 13 built their own websites, apps, and games in four hours using AI-assisted creation tools.

Co-led with Larry Hay, in partnership with the Peregian Digital Hub.

Sandy Robinson working with a student on an iPad alongside hand-drawn UI sketches

Working alongside

A student moves from hand-drawn UI mockups to working prototypes during the Hackathon.

It was awesome, the kids really loved it. I had a bunch of them comparing the same prompt in ChatGPT, Perplexity, and Claude and deciding which was best. They were also really engaged in writing better prompts, so I think we can consider the presentation a huge success.
DL, Head of Science
AI Literacy workshop
I really appreciate the thought and experimentation you've put into the guide. The format works well. It's clean, accessible, and walks users through the capabilities. I can absolutely see this becoming a valuable model for staff.
BM, Design & Technology
Professional development

My company

Big Wella Apps

Four products for Australian classrooms, building since 2022. The product side of the same work, built on what real classroom use has already taught me.

  • Live
    WellaCite · citation tool for Australian students. Free during build.
  • Live
    Well Read · reading discovery for AU/NZ school libraries. Mood quiz, tier-aware interface, teacher dashboard.
  • In build
    BLACKOUT · narrative game on AI integrity for Years 6–10.
  • In build
    Starlit Watercolour · astrophysics through art for ages 6–10.
Visit bigwella.com
Big Wella Apps watermelon logo

The through-line

None of this is ever finished.

The frameworks shift. The tools rebuild. The workshops change each term, because the teachers and students using them are navigating the same changes in real time. Iterating with them, not at them, is the practice.

Currently exploring

Questions I'm holding.

Not finished work. Not raw notes. The places where current thinking is becoming the next thing.

Scaffold effectiveness

Which scaffolds actually move students from copy-pasting AI output to interrogating it? Which ones produce mimicry instead?

Transparency in AI systems

What students can be told about how the tool they're using actually works, and what difference that knowledge makes to how they use it.

Play as gateway to complexity

When playful framing opens up serious technical territory, and when it stays as decoration. Where the line sits in classrooms.

Student trust in tools

How quickly students stop questioning a tool that keeps being right. What that does to their reading, writing, and reasoning over a year.

Information scarcity

When the answer is too easy to get, how does that shape what students are willing to wonder about?

Tool stability

Frameworks last. Tools don't. How do we build literacy that survives the next platform shift?

Get in touch

Want to talk?

Conferences, podcasts, articles, or a conversation about AI in classrooms. Different topics, same email.