Nine real builds, all shipped with AI co-piloting.
The Hub doesn't just demo AI — we ship it. Filter by category, scroll for more. Each card opens the live site.

Gamified life-OS for habits, projects, and personal growth.

Member portal for a Melbourne beekeeping community — events, dues, store.

AI-assisted hardware pipeline: idea → BOM → schematic → PCB → 3D enclosure.

Working website for a real Melbourne cleaning business — quotes, booking, content.

Multi-tenant member-ops platform — tenants, AI agents, workflow automation, Stripe.

Eight companies, twenty-four agents, one task-dispatch loop. The Paperclip company OS, live.

Rust rewrite of the Supabase stack — self-host Postgres + auth + storage. Includes a 35-screen studio prototype.
Cocktail discovery + inventory. Tell it what you've got; it tells you what you can pour.
Partner-friendly. If you're a frontier-model lab and want a quiet space in Melbourne to see what builders do with your tools, hello@melbourneaihub.com.au.
Open weights, new tools, a few honest retractions.
What the editors are actually reading. Each card links straight to the primary source — no third-party hot takes.
Carbon — genomic LLMs at 500M, 3B, and 8B.
Claude 4.x family — Opus, Sonnet, Haiku.
What's shipping on the OpenAI side.
Gemini's multimodal trajectory.
Copilot agent mode in the wild.
Grok and live-data search.
Curated by the editors until the auto-fetch from official RSS feeds ships. Spot something we should run? hello@melbourneaihub.com.au.
Founder-funded until demand proves itself.
No outside capital, no opaque budgets. The Hub is paid for by the people who run it; every dollar in and out is logged in public. Workshop tickets and member dues cover what they can; the rest comes from the founders.
The full ledger — milestones, decisions, vendor list, AI compute spend — lives on the Transparency page.
Our stack, in the open.
The Hub runs on a deliberately broad mix — frontier models from every major lab, plus the open-weight ecosystem. Members get the full set in workshops; we test what works, publish what doesn't.