learn.diya.au turns live-classroom AI training into interactive courses — idea cards, hands-on labs with copy-ready prompts, mastery quizzes, and projects that end with something your learners shipped. Built by Design Industries on an Australian-sovereign platform: learner data stays onshore, the AI tutor is grounded in approved content, and educators see provenance — not guesswork.
Early access now open — first cohorts run late 2026 to early 2027. No video playlists, no seat-warming.
Decorative terminal animation showing a learner's session: signing in, opening a course session, asking the grounded AI tutor a question answered from course content on Australian-resident infrastructure, completing a lab step for XP, passing the mastery quiz to unlock the next session, and shipping a live app.
Nothing here is a video playlist. Every session is idea cards, hands-on labs, and a mastery quiz — built from live classroom programmes and illustrated with 170+ commissioned artworks. Filter by who you teach.
Understand how AI actually works, use it safely, think critically about it, and build a first real project — no coding assumed, every subject area welcome.
From a blank screen to a live AI app on the internet in one afternoon — with a QR code in hand to prove it. Zero prior coding, design, or AI experience.
An honest, data-grounded map of where the job market is heading — and a personal AI positioning plan for every learner, whatever their field.
Walk in curious, walk out having shipped a live AI app — a six-session arc with ethics and responsible-AI judgement built in from the first session.
Use AI honestly to win the job hunt — résumé tailoring, interview prep, and the demonstrable AI-fluency portfolio recruiters now screen for.
From anxious to fluent: use AI as a productivity partner, teach it credibly in any subject, and redesign assessment for an AI-aware classroom.
A student track that teaches AI as a study coach — without outsourcing the thinking — plus an owner track that makes AI literacy your centre's edge.
Work like a professional AI-era developer: real sprints, code reviews on real pull requests, live cloud labs — ending in a deployed capstone product.
Course structure, sequencing, and examples are localised with each partner for Australian cohorts and institutional context during onboarding.
Five real components from Ship Your First AI App, Session 2 — "Give Your App a Brain" — live on this page. Click around; the XP is real enough.
Two study apps can look identical — same buttons, same purple gradient — and one feels like a patient teacher while the other rambles like a random chatbot. The difference is never the interface; it's the brain.
💡 Drift comes from missing constraints, not a broken model. The misconception here is "the AI knows what my app is for" — it only knows what you wrote, and whatever you leave unspecified it fills with a confident guess.
"You are a friendly Year 10 maths explainer"
"Explain the step the student is stuck on"
"Max 3 sentences, simple English, warm"
"Maths only — politely redirect the rest"
👀 Look for: The refusal is in quotation marks — exact words you can test for, not a vague intention like 'be appropriate'.
Every word above is verbatim course content (typographic quote styling aside), with one localisation: the session file's "Class 10" reads "Year 10" for Australian cohorts — the same localisation pass every course gets during partner onboarding.
The learning model comes from live classrooms, not lecture capture: learners do the work, prove they understood it, and only then move forward.
Learners sign in and go — nothing to install. Progress, XP, and streaks follow them across devices, and educators see the whole cohort at a glance.
Idea cards build the concept, hands-on labs with copy-ready prompts make it real, autosaving worksheets capture the thinking, and a mastery quiz checks it stuck — pass at 80%.
Progression is mastery-gated and enforced server-side — no skimming ahead, no hollow completions. What a dashboard says a learner finished, they actually did.
The tutor answers from your approved course content and cites where answers come from — "from your course, Session 4" — labels general knowledge as exactly that, and declines to produce submittable assessment work. Every interaction is recorded in an append-only log educators can actually act on.
Whether you're a school answering to parents, a university answering to council, or an IT company answering to clients — your learners' data never leaves Australia. Inference runs exclusively on Amazon Bedrock's Australia-geo-fenced Claude endpoints — verified with live inference calls in our own AWS account in July 2026 — and is never routed through Asia-Pacific cross-region paths.
AI tutoring runs on Anthropic's Claude models via Amazon Bedrock's Australian regional endpoints, geo-fenced to Sydney and Melbourne only. We verified this with live inference calls from our own AWS Sydney account — before writing this page.
Your content and your learners' data are never used to train AI models — a contractual Amazon Bedrock commitment we can cite in procurement, not a marketing promise.
Backups replicate from Sydney to Melbourne — never offshore. Residency holds in failure modes, not just on the happy path.
Integrity conversations should move from suspicion to evidence. Every AI interaction on the platform lands in an append-only record: who asked what, when, which model and version answered, and what the learner then submitted.
The AI tutor answers from approved course content and cites where answers come from — labelling general knowledge as exactly that.
The tutor declines to produce submittable assessment work. Refusal behaviour is a release gate, not an assumption: every build must pass a scored evaluation suite — integrity scenarios and grounding checks included — before it reaches learners.
Exact model and version, prompt, response, and timestamps for every AI interaction. Educators get visibility surfaces over AI assistance, giving integrity reviews evidence rather than hunches.
Grades, quiz results, and progress are computed server-side. Answer keys never reach the learner's browser — integrity is an architectural property, not a policy document.
Whoever signs off on your tools — a school board, a university procurement panel, or an enterprise security team — every line below is an architectural commitment you can test, not a brochure adjective.
Every model and version in use is registered, per-tenant pinnable, and subject to mandatory change notices before anything changes under your learners.
AI usage is metered and passed through at cost — including the ~10% premium Australian-resident endpoints carry. That premium is visible because it is the residency guarantee, never hidden inside a flat licence.
Safety guardrails, per-tenant quotas, and human-in-the-loop review queues are core platform architecture — operated by a team already running production guardrails in AWS Sydney today.
Where offshore support access exists it is declared, just-in-time, role-scoped, and fully session-logged. You will never discover an access path we didn't disclose.
WCAG 2.2 AA from the first component — semantic markup, keyboard paths, and contrast-checked design tokens enforced as a build-pipeline requirement, not retrofitted.
We are pursuing ISO 27001 certification (targeted 2028) and planning IRAP assessment (2029). We publish targets, not premature claims — and invite you to hold us to them.
The platform is pre-release, and we say so. Here is the milestone path — early-access partners get evidence at every step.
AU-geo-fenced Claude inference and production AI guardrails verified live in our own AWS Sydney account; sovereign AI runtime build under way; curriculum library in structured, compiler-ready form; early-access conversations open.
Real learners on the corpus-grounded AI tutor — across schools, colleges, training providers, and enterprise teams — with educator dashboards and the full interaction log in service.
Engagement, educator sign-off, unit AI cost per learner-hour, and a sovereignty audit of the complete AI interaction log — published to early-access partners.
Microsoft Entra ID SSO and SCIM provisioning, LTI 1.3 with an end-to-end Canvas demonstration, staff authoring workflow, and an external WCAG 2.2 AA audit.
Multi-tenant platform with published scale and disaster-recovery evidence.
ISO 27001 certification targeted for 2028; IRAP assessment planned for 2029.
We are onboarding a small number of early-access partners — schools, colleges and universities, training and coaching providers, and enterprise learning teams — for cohorts starting late 2026 to early 2027. Partners shape the educator experience, receive the full evidence pack at the March 2027 review, and get first access as the platform matures. Not sure where to start? Tell us who you teach, and a human will recommend a course.
hello@di.net.au · 1300 73 63 63
One reply from a human. No newsletter, no spam.
The six questions every partner asks first.
Cohorts, not lone browsers: senior schools (Years 9–12), colleges and universities, training and coaching providers, and IT companies or enterprise teams building AI capability. The catalogue filter above shows which of the eight courses fits each audience — most need no coding at all.
No — for most of the catalogue. Ship Your First AI App assumes zero coding, design, or AI experience, and AI & Your Career and AI Foundations assume no coding background at all. The AI Developer Residency is the deliberate exception: a 90-day deep dive for developers and technical teams.
Progression is mastery-gated and enforced on the server: a session unlocks only when the previous one is fully complete — every idea card, every lab step, and a quiz score of at least 80%. No skimming ahead, and no way to fake a completion from the browser: grading is server-computed and answer keys never reach the client.
In Australia, full stop. Learner data, logs, and backups live in Sydney with disaster recovery in Melbourne, and AI inference runs only on Australia-geo-fenced endpoints in Sydney and Melbourne. AI tutoring runs on Anthropic's Claude via Amazon Bedrock — verified with live inference calls in our own AWS account — and your content and learners' data are never used to train AI models, a contractual Amazon Bedrock commitment.
The platform launches with the eight-course library, all built through a structured authoring pipeline — the same pipeline that will power a staff-facing authoring workflow on the 2027 roadmap. Early-access partners help shape that workflow, and course localisation for your context is part of onboarding today.
Early-access partnerships are scoped per cohort — this is a conversation, not a checkout. One pricing principle is fixed from day one: AI usage is metered and passed through at cost, including the ~10% premium Australian-resident endpoints carry — never hidden inside a flat licence. First cohorts run late 2026 to early 2027; register interest above and a human replies.
