# Do Not Put All Our Eggs in One Basket This is a working Draft of 8,625 words. *A Protopian Gambit for Federated Sovereign AI, Community Compute and Civic Stewardship from Now to 2032 Olympics to the Solar System* Submission to the Senate Environment and Communications References Committee Inquiry: Artificial intelligence and data centres Submitter: Luke Nathan Hayes Location: Minjerribah / North Stradbroke Island, Redland City, Queensland, Australia Date: 21 August 2026 I make this submission as a private citizen. It extends the argument I previously made in Submission #2 to the federal inquiry into Local Government Funding and Fiscal Sustainability: https://www.aph.gov.au/Parliamentary_Business/Committees/House/Regional_Development_Infrastructure_and_Transport/LocalGovernmentFunding/Submissions This submission offers an architecture, a pilot pathway and a strategic proposition for public consideration. ### Executive summary Australia should not treat artificial intelligence infrastructure as synonymous with hyperscale data centres. Data centres are one valuable layer of a rapidly evolving compute ecology. Australia also needs self-sovereign personal AI, neighbourhood edge systems, serious community-scale compute, state and Commonwealth capability, commercial clouds, research supercomputers, emerging quantum systems and pathways to future orbital compute. The central proposition of this submission is: Australia should build a federated sovereign AI architecture in which personal AI, neighbourhood edge systems, community-scale compute, state and national systems, hyperscale data centres, quantum computers and future orbital systems can interoperate without any one layer becoming an unavoidable point of dependency. This is not an argument against data centres, global AI companies, foreign investment or frontier systems. It is an argument against putting all our eggs in one basket. Australia’s National AI Plan already identifies compute and resilient digital connectivity as critical infrastructure and recognises distributed processing as part of the national digital backbone. The Commonwealth’s Data Centre Expectations address sovereignty, Australian skills, research capability, community benefit, energy and water while distinguishing major data centres from smaller edge and on-site infrastructure. CSIRO’s Vetra platform at Pullenvale already demonstrates compact, modular, sovereign AI computing close to robots and sensors as part of an integrated edge-core-cloud system. Australia should deliberately fill the missing middle between the phone and the hyperscale facility. I propose using approximately one serious local compute node per 10,000 people as a starting planning ratio, adapted to geography, Country, population density, language, universities, industry, energy, disaster exposure and existing infrastructure. These should not be identical boxes bought from a single company. Some locations may use current GB200 or GB300-class NVIDIA systems. Others may use AMD Helios or Instinct systems, H100, H200 or A100-generation equipment, refurbished hardware, Australian-designed systems, Chinese or other international equivalents, or modular clusters suited to their actual workloads. The public value is not merely the number of floating-point operations. It is the capability transferred to people. The first national cohort should be recruited through existing networks that already convene, educate and spread practice: local government associations, chambers of commerce, high schools, TAFEs, universities, hospitals and health services, libraries, maker communities, Indigenous developer pathways, girls-in-code organisations, creative industries, research institutions and technologically capable community organisations. A few deliberately different early adopters in every state and territory should receive: 1. a serious rack or equivalent cluster; 1. professional power, cooling, networking and security; 1. a visible community AI laboratory; 1. a digital twin of the installation; 1. just-in-time learning and XR training; 1. access to Australian and international models; 1. a requirement to document what was learned and train the next cohort. While the first sites are discovering what one rack can do, the next sites should already be training and preparing. The rollout cycle should be: identify → qualify → train → install → experiment → document → teach the next cohort The racks would support education, local software development, small business, creative production, digital twins, environmental sensing, public planning, health research, citizen science, emergency management, infrastructure modelling and community deliberation. They would also create a disaster-resilience layer between personal devices and distant clouds. A community should retain useful local intelligence when a cyclone, fibre cut, cyberattack, grid failure, telecommunications outage or other disruption breaks the mainland tether. The deeper purpose is AI alignment through human self-alignment. Australians should not be asked to trust that government, a technology company or one national model permanently knows what “Australian values” are. People should be given tools to reflect on the values, virtues and lessons they believe we inherited from the past, require in the present and wish to cultivate in the future. Communities should be able to explore what Joyful Responsible Abundance means at the scales of self, family, community, Country, town, city, state, nation, company, artificial intelligence, planet and future generations. Joyful Responsible Abundance is not the predetermined finishing line. It is an opening orientation for a self-directed choose-your-own-adventure process: observe → reflect → choose → build → test → learn → adapt → share what works Aura of Intelligence is my proposed personal sovereignty layer. GAJRA Earth-the Global Association for Joyful Responsible Abundance-is the voluntary international federation through which communities, cultures, language groups and individuals could share what resonates without being forced into one global values vector. The Braided Economy provides a way to connect this infrastructure to community contribution without collapsing every human activity into a market transaction. The market braid pays for racks, energy, cooling, buildings, licensed engineering, professional operation and paid work. The reciprocity braid records verified community contribution through the Community-Hour: One verified hour of contribution equals one C-Hour. There is no dollar equivalent and no conversion back into the market braid. Communities decide which activities qualify and which non-market rewards they make available. The flagship Earn-to-Heal loop would allow C-Hours to support access to clinically governed HBOT, sauna and Aura Geode research and wellbeing infrastructure. Other locally selected benefits might include queue priority, VIP community-event access, community-grown produce, shared equipment, workshops, local experiences and minor privileges people voluntarily nominate outside the market economy. Brisbane 2032 creates a fixed mobilisation horizon. Instead of allowing the skills, relationships and civic energy of tens of thousands of Olympic and Paralympic volunteers to dissolve after the Closing Ceremony, Australia could build a permanent Civic Reserve trained in AI, digital twins, logistics, communications, citizen science, disaster resilience, cultural stewardship and space-weather awareness. The volunteers who help Queensland welcome the world could become long-term civic stewards of Earth and the solar system. Queensland already contains the beginning of an unusual capability corridor: CSIRO Vetra at Pullenvale → universities and health networks → Brisbane 2032 → PsiQuantum at Moreton Bay → Redlands and Minjerribah community compute → Gold Coast and South East Queensland → Australian national and international networks. PsiQuantum broke ground at Moreton Bay Central in June 2026 on its proposed utility-scale fault-tolerant photonic quantum computer, following the opening of its Test and Validation Lab at Griffith University. Australia should connect that specialised capability with classical supercomputing, community compute, schools, hospitals, researchers and public-interest applications rather than allowing each layer to develop as an isolated island. Using an all-in policy-planning allowance of A$6 million per frontier-class local node, a 32-site first cohort-four varied sites in each state and territory-would represent approximately A$192 million. A 100-site learning network would represent approximately A$600 million. Australia’s population of approximately 27.8 million produces a simple starting envelope of roughly 2,780 node-equivalents at one per 10,000 people. A fully built frontier-class network at the same allowance would be approximately A$16.68 billion. Using the world-cities dataset accompanying this submission, approximately 41,549 listed places have a recorded population of at least 9,000. At A$6 million per location, that produces a global planning thought experiment of approximately A$249.3 billion. Using a rounded 44,000 locations produces A$264 billion. That is in the same broad order of magnitude as the public A$268-368 billion life-cycle estimate for Australia’s nuclear-powered submarine program under AUKUS. I make that comparison deliberately. It is not an accounting claim that submarines and civic AI racks perform the same task. It is a strategic and moral question about what humanity is prepared to count as security infrastructure. Would a global network of locally governed intelligence, education, science, cultural understanding, disaster resilience and human self-alignment create more peace and shared capability than placing the equivalent resources into a single deterrence program? Australia has also shown that it can mobilise enormous resources quickly when a shock arrives. More than A$90 billion was committed through the Australian response to the Global Financial Crisis, and Commonwealth and state COVID-era support packages reached approximately A$327 billion under broad Parliamentary Budget Office accounting. The next market, climate, cyber, health or geopolitical shock will again produce rapid public expenditure. The question is whether we invent that spending program in the panic-or prepare a shelf-ready program of productive, sovereign, decentralised infrastructure beforehand. The computing horizon is also extending beyond terrestrial data centres. SpaceX’s published architecture now connects TERAFAB advanced semiconductor manufacturing, reusable Starship launch, Starlink V3 communications and large orbital AI-compute satellites. Its filings describe a long-term TERAFAB objective of producing one terawatt of compute hardware per year, including hardware optimised for orbital use. Starlink V3 is designed for approximately one terabit per second of downlink capacity per satellite, with up to 60 V3 satellites carried on a Starship deployment. SpaceX has also filed application SAT-LOA-20260630-00264 for a proposed third-generation Starlink architecture of up to 100,000 satellites. The orbital AI programme being publicly described as Starmind takes the trajectory further: solar-powered data centres in space, optical networking and potentially enormous constellations of compute satellites. SpaceX’s own public filings describe an architecture that could begin deploying large orbital AI systems from 2028 and eventually scale into millions of satellites. SpaceX’s United States launch system already spans four primary pads, while Starship infrastructure is expanding across Starbase in Texas and launch facilities in Florida. Its published design objective is full and rapid reuse approaching commercial-aviation turnaround, eventually supporting multiple launches per day. The trajectory I see by 2028 is daily and eventually multiple-daily Starship activity supporting Starlink V3, Artemis and Gateway missions, lunar infrastructure, scientific payloads, commercial activity and orbital compute. NASA’s current Artemis architecture places the Gateway and a second crewed lunar surface expedition around the September 2028 Artemis IV mission window. Australia should not respond by moving every egg from a terrestrial data-centre basket into a SpaceX basket. It should build the interfaces between all the baskets. The next three total solar eclipses crossing Australia provide a beautiful public timeline for this work: 1. 22 July 2028: totality through Australia and Sydney; 1. 25 November 2030: totality across southern Australia, ending just northwest of Brisbane; 1. 13 July 2037: totality across Brisbane, the Gold Coast and South East Queensland, including Quandamooka Country. These should become milestones for national citizen science, art, music, space-weather education, distributed sensors, public simulation and civic participation. Between them sits Brisbane 2032. Australia is already known for producing some of the world’s finest athletes. The Protopian Gambit is to use this period to also produce the world’s most capable generation of civic stewards-Australians of every age who understand AI, data, energy, Country, community, disaster resilience, science, space and the responsibilities that accompany powerful tools. We can be more than spectators of the intelligence age. We can be its civic architects. We go beyond. ### 1. Continuing the local-government fiscal sustainability argument My earlier submission to the inquiry into Local Government Funding and Fiscal Sustainability argued that councils operate inside an Incomplete Ledger. Local governments are expected to maintain infrastructure, community wellbeing, environmental assets, social cohesion, public spaces, disaster readiness and civic trust. Yet much of the work that sustains those systems-caregiving, mentoring, volunteering, local knowledge, cultural stewardship and ecological maintenance-is poorly represented in conventional economic accounting. That submission proposed a Braided Economy, a Community-Hour, community sovereign wealth mechanisms, sovereign digital twins, fractal governance and a Minjerribah pilot incorporating First Nations data sovereignty. This submission asks the next physical-infrastructure question: Where should the intelligence that helps operate those civic systems live? Who should own it? Who should understand it? Who should be capable of changing it when it is wrong? If every council, school, health service and community becomes a permanent customer of a distant proprietary cloud, AI may simply become another recurring cost and another form of capability dependency. If Australia builds locally accessible infrastructure, open models, transferable skills and interoperable systems, local government can become a competent participant in the intelligence age rather than merely buying reports and subscriptions from it. The data-centre inquiry therefore reaches directly into local-government fiscal sustainability. The policy choice is not simply how many data centres Australia approves. It is whether communities retain enough infrastructure, intelligence and skill to participate in the systems that increasingly shape their services, finances and futures. ### 2. AI infrastructure is an ecology, not a building type The phrase “AI and data centres” can accidentally place the building at the centre of the story. The intelligence ecology is much broader. A modern phone contains neural-processing hardware capable of running local models. A personal computer or workstation can run private language, image, video and scientific models. A neighbourhood node can coordinate sensors, emergency information and local communications. A community rack can run large models, digital twins, fine-tuning, rendering, simulation and local research. A state or Commonwealth system can run larger public-interest and strategic workloads. A hyperscale data centre can train frontier systems and elastically serve millions of users. A quantum computer may accelerate specialised problems that remain difficult for classical systems. An orbital system may eventually combine abundant solar power, optical networking and large-scale compute beyond Earth’s atmosphere. None of these layers makes the others obsolete. Australia needs the whole stack. #### Proposed federated compute continuum | Layer | Principal role | Sovereignty principle | | --- | --- | --- | | L0 - Personal and household | Personal AI, living memoirs, private RAG, health and ageing support, household sensors, local creative tools and self-directed learning | My data, my keys, my permissions and my ability to disconnect | | L1 - Neighbourhood and edge | Disaster kiosks, communications meshes, local sensors, schools, libraries, small shared clusters and community information | Explicit sharing, graceful degradation and local continuity | | L2 - Community, postcode or bioregional compute | One serious rack or suitable cluster supporting local development, models, simulations, digital twins, citizen science, creative industries, SMEs and training | Shared national infrastructure physically accessible to communities | | State and territory | Health, education, infrastructure, environment, emergency coordination, larger research and public-service workloads | State capability federated with local and Commonwealth systems | | Commonwealth and national research | Australian foundation models, scientific compute, strategic workloads, model evaluation and national coordination | National capability without technological isolation | | Hyperscale commercial | Frontier training, elastic inference, global services and extremely large workloads | Competition, portability, community benefit and avoidance of permanent lock-in | | Quantum and specialised accelerators | Chemistry, materials, optimisation, cryptography and other specialised computational structures | Australian skill, access and interoperability with classical compute | | Orbital and future systems | Communications, global sensing, specialised scientific workloads and future AI compute | Australian-controlled applications, permissions, protocols and ground infrastructure | First Nations and Native Nations should not be placed beneath local government in this diagram. They form a sovereignty plane crossing every layer. A sovereign Quandamooka model, Digital Twin of Country or cultural knowledge system may use personal devices, community racks, national research compute and international models while remaining governed by Quandamooka authority. The same principle can extend to Māori, Pacific and other Native Nations. The purpose of federation is not to send everything upwards. It is to make choices possible. ### 3. Filling the missing middle The weakest layer in the present Australian stack is the human-scale infrastructure between personal devices and major data centres. That missing middle is where community-scale sovereign compute belongs. I use one serious node per approximately 10,000 people as a starting ratio because it is understandable and large enough to support shared capability. It is not intended to erase geography, Country or real communities. A dense metropolitan area may organise several nodes through schools, universities, hospitals, libraries and industry precincts. A remote region may use a smaller cluster with satellite communications and stronger energy resilience. An island may prioritise offline operation. A First Nation may structure its node around cultural authority and collective data governance. A university town may host more specialised training and research. A hospital precinct may combine clinical research, private inference and heat reuse. A creative region may emphasise film, games, XR, virtual production and cultural preservation. The architecture should be vendor-neutral and generation-flexible. Current reference points include NVIDIA GB200 and GB300-class systems and AMD Helios or Instinct rack-scale systems. NVIDIA specifies power of up to approximately 142 kilowatts for GB300 NVL72 systems, while AMD is developing Helios around open rack-scale standards and a broad partner ecosystem. Previous-generation H100, H200, A100 and comparable systems remain useful for inference, fine-tuning, rendering, education, simulations and research. Australia should design a cascading public compute fleet. As national research facilities and larger urban nodes adopt new generations, earlier systems can move into regional and community use rather than being discarded or permanently concentrated in private facilities. The hardware generation should follow the workload. The sovereignty requirement is that communities retain a meaningful local layer. ### 4. A transparent A$6 million planning unit For strategic planning, I use an all-in allowance of A$6 million per frontier-class community node. That package should purchase more than GPUs. | Component | Indicative planning allocation | | --- | --- | | Rack-scale compute, storage and orchestration hardware | A$4.00 million | | Grid connection, liquid cooling, heat exchange, UPS and energy integration | A$650,000 | | High-speed networking, cybersecurity and backup communications | A$300,000 | | Secure machine room, fire systems, access control and monitoring | A$250,000 | | Visible community laboratory and XR just-in-time learning package | A$350,000 | | Initial spares, specialist onboarding and operational support | A$250,000 | | Program integration and contingency | A$200,000 | | Total planning allowance | A$6.00 million | This becomes a common planning unit rather than a locked procurement specification. Locations with existing high-capacity power, fibre and secure facilities may direct more of the allowance into compute or training. Remote and disaster-prone sites may direct more into microgrids, batteries, satellite connectivity and physical resilience. #### National scale envelopes | Stage | Sites or node-equivalents | Planning envelope | | --- | --- | --- | | First state and territory cohort | 32 | A$192 million | | National early-adopter network | 100 | A$600 million | | One node in every local-government-scale network | approximately 500-550 | approximately A$3.0-3.3 billion | | Population starting ratio of one per 10,000 Australians | approximately 2,780 | A$16.68 billion | If all 2,780 node-equivalents operated current full-rack systems simultaneously, the national full-load envelope would be approximately 334 megawatts at 120 kilowatts per rack, or approximately 395 megawatts at 142 kilowatts. That equates to approximately 2.92 to 3.46 terawatt-hours per year at continuous full load. The important architectural difference is distribution. Instead of concentrating hundreds of megawatts, backup generation and heat in one fenced industrial precinct, the load is spread across Australia and attached to useful community facilities, generation assets and public heat demand. A rack at an aquatic centre can contribute heat to pools and hot-water systems. A rack at a hospital or wellbeing precinct can connect to year-round thermal loads. A school, university or TAFE can turn the machine into a live curriculum. A community resilience hub can combine the rack with batteries, solar, communications and emergency facilities. The electricity is not free. The heat has already been purchased with the electricity. The design question is whether that heat becomes waste or a public utility. ### 5. Build the early-adopter network through networks that already exist Australia should not begin by selecting a handful of random councils and leaving them alone with complicated hardware. The Commonwealth should invite early adopters through institutions already capable of convening people and transferring practice. These include: 1. the Australian Local Government Association and state local-government associations; 1. chambers of commerce and regional economic-development networks; 1. high schools, TAFEs and universities; 1. hospitals, health services and medical-research networks; 1. libraries and community centres; 1. Indigenous businesses, youth programs and developer pathways; 1. Code Like a Girl, Tech Girls Movement and other girls-in-code and women-in-technology networks; 1. maker spaces, robotics clubs and citizen-science groups; 1. film, screen, game, music, XR and creative-industry networks; 1. emergency-management and environmental organisations; 1. research institutions including CSIRO; 1. industry partners willing to provide hardware, models, engineering and knowledge transfer. The host qualification process should ask: 1. Does the organisation already contain people capable of learning quickly and teaching others? 1. Can it provide or partner for suitable power, cooling, networking, security and physical facilities? 1. Does it have a credible plan for public, educational, research or community use? 1. Will it publish operating lessons, failures and reusable learning material? 1. Will it mentor the next cohort? 1. Can it support data sovereignty, privacy and First Nations governance where relevant? 1. Can the site remain useful during normal life as well as emergencies? 1. Is the proposal designed to circulate capability locally rather than merely host private equipment? Each state and territory should begin with several deliberately different sites. One may be metropolitan. One regional. One remote or disaster-prone. One may be First Nations-governed, hospital-based, educational, industrial, creative or island-based. The first cohort should not be expected to know every answer. Its purpose is to discover the answers. Every pilot should investigate: 1. what one rack actually does well; 1. which workloads should remain personal; 1. which workloads belong locally; 1. which workloads should move to state, national or commercial systems; 1. how people schedule and govern shared compute; 1. how much capacity is actually used; 1. which models work best; 1. which data should never leave; 1. what power and cooling profiles emerge; 1. which skills were missing; 1. what broke; 1. what surprised people; 1. which community outcomes justified the infrastructure; 1. what the next site should do differently. While cohort one is operating, cohort two should already be completing training, engineering and community design. The learning system therefore compounds rather than repeatedly restarting. ### 6. Just-in-time learning comes with the rack The rack should arrive with a learning environment, not merely an instruction manual. Every installation should have a live digital twin that represents: 1. compute nodes and accelerators; 1. power consumption; 1. coolant temperatures and flow; 1. storage and network traffic; 1. workload scheduling; 1. model deployment; 1. cyber events; 1. component faults; 1. battery state; 1. heat-reuse systems; 1. backup communications; 1. emergency and degraded modes. Students, technicians, council employees, health workers, local developers and community members should be able to enter an XR version of the facility before working with the live system. They should be able to practise: 1. starting and stopping workloads; 1. allocating compute; 1. deploying local models; 1. responding to cooling faults; 1. isolating compromised services; 1. moving workloads to another node; 1. entering disaster mode; 1. restoring from known-good images; 1. monitoring energy and heat; 1. inspecting model and dataset provenance; 1. applying cultural and privacy permissions; 1. maintaining the surrounding community laboratory. The training should be modular, visual and updated with each hardware generation. It should support people from secondary-school level through to specialist engineering. A Year 10 student should be able to see the architecture. A TAFE student should be able to maintain parts of it. A university researcher should be able to federate experiments through it. A local developer should be able to deploy useful models. A hospital should be able to train approved private workloads. A council planner should be able to run a simulation rather than merely commission a report. A Traditional Owner organisation should be able to connect a sovereign model under its own governance. The most important output of the first racks may not be compute. It may be thousands of Australians becoming familiar with how intelligence infrastructure actually works. ### 7. Everyday capability and extraordinary resilience Distributed compute earns its place by remaining useful every day. Potential everyday uses include: 1. local-government digital twins; 1. planning and infrastructure simulation; 1. environmental and coastal sensing; 1. water, waste, energy and transport modelling; 1. health and ageing research; 1. local software and AI development; 1. language and cultural preservation; 1. agricultural and food-system intelligence; 1. virtual production, film, games and XR; 1. citizen science; 1. secondary-school, TAFE and university education; 1. SME access to advanced models; 1. robotics and advanced manufacturing; 1. public-interest legal and policy analysis; 1. locally hosted open models; 1. Australian model evaluation; 1. community arts, music and storytelling. The same infrastructure can enter reduced or specialised modes during: 1. flood; 1. bushfire; 1. cyclone; 1. pandemic; 1. cyberattack; 1. grid failure; 1. telecommunications failure; 1. supply-chain disruption; 1. severe space weather; 1. other geophysical or celestial events. A local node can preserve maps, emergency directories, sensor feeds, resource inventories, local identity, translation and trusted communications when external services are congested or unavailable. The node should not attempt to reproduce the entire internet. It should keep the civic nervous system alive. A diversity of vendors, models, runtimes and locations also reduces common-mode cyber risk. One fault, price increase, political decision, software failure or ownership change should not disable every layer simultaneously. The objective is not an indestructible machine. It is infrastructure with multiple useful operating states rather than one giant on/off switch. ### 8. Sovereignty means more than locating a server in Australia A rack is not sovereign merely because it sits on Australian soil. Sovereignty should include: 1. data custody; 1. encryption keys; 1. identity; 1. permissions; 1. model weights; 1. model preferences; 1. software and runtime; 1. workload portability; 1. network dependence; 1. energy dependence; 1. hardware supply; 1. local skills; 1. maintenance; 1. recovery; 1. governance; 1. the practical ability to change providers. A foreign-owned system can increase Australian sovereignty when it expands our capability and preserves alternatives. An Australian-located proprietary system can weaken sovereignty if nobody else can understand, repair, move or replace it. The national objective should be: Sovereignty without isolation. Australia should use the best systems available while retaining the capability to leave, substitute, recover, interoperate and build its own. #### Personal sovereignty through Aura of Intelligence AI sovereignty begins with the person. Persistent AI systems may eventually know a person’s memories, relationships, health history, aspirations, fears, politics, finances, movement, voice, face, correspondence and values. A one-size-fits-all terms-and-conditions agreement is not an adequate constitutional framework for that relationship. Aura of Intelligence proposes a user-owned digital twin and a machine-readable permissions layer through which the person can express: 1. what may leave their device; 1. what may be used for inference; 1. what may be used for training; 1. whether derived representations may be retained; 1. who may access health, family or cultural data; 1. whether information may leave Australia; 1. what may be shared during an emergency; 1. what happens when permission is withdrawn; 1. which models and values the person wants their AI to use. The person should own their history even when they withdraw another organisation’s permission to use it. This is not merely privacy. It is personal constitutional agency. #### First Nations sovereignty and Digital Twins of Country First Nations data sovereignty must be a load-bearing architectural principle. The Commonwealth’s Framework for Governance of Indigenous Data already seeks to place Aboriginal and Torres Strait Islander peoples at the centre of data governance and strengthen their authority across the data lifecycle. A Minjerribah pilot should explore a self-sovereign Quandamooka AI and Digital Twin of Country governed with the relevant Quandamooka people and institutions. Different permission zones could apply to different knowledge. Some material may be public. Some may be available to community members. Some may require a cultural role, relationship, age, gender, family connection or other permission determined by custodians. Some may permit a particular research query without permitting general model training. Some may permit an aggregate answer while protecting the underlying material. Some may remain within Quandamooka-controlled infrastructure. Some may remain closed. Special data warrants special permissions. The architecture should make it possible for a First Nation to say: These are our models. These are our data. These are the permissions under which you may interact with them. Access does not silently convert custody into ownership. The preferred direction is: Models visit data more often than data has to visit models. That principle can become one of Australia’s most valuable international contributions. ### 9. AI alignment begins with human self-alignment Much of the global AI-alignment debate asks how artificial intelligence can be made to behave according to human values. The prior question is: Which human values, expressed by whom, under what conditions, at what scale and with what right of revision? Humanity has never held one permanent values vector. Individuals disagree with their families. Communities disagree internally. Cultures use different words for similar virtues and similar words for different ideas. Values change through experience, reflection and time. A centralised AI-alignment process risks encoding the preferences of whoever owns the largest model, dataset, committee or government. A federated alignment process can preserve plurality while making common ground visible. I propose recurring participatory exercises in which people identify: 1. values and lessons worth carrying from the past; 1. values needed in the present; 1. virtues they hope to cultivate in the future; 1. failures they do not want repeated; 1. hopes they want represented; 1. how they arrived at those conclusions; 1. what Joyful Responsible Abundance means in their context. These reflections can occur at multiple scales: self → family → community → Country or tribe → town → city → bioregion → state → nation → company → AI system → planet → future generations The output should not be one flattened global answer. The valuable dataset includes the differences: 1. Where do cultures converge? 1. Where do they diverge? 1. Which apparent conflicts come from different definitions? 1. Which principles remain stable across language and geography? 1. Which responsibilities change with scale? 1. What does abundance mean to a remote island, a dense city, a family, a corporation and an AI? 1. How do communities wish to resolve disagreement without erasing one another? #### Joyful Responsible Abundance as an opening orientation Joyful asks whether a possibility expands life, connection, creativity, meaning, curiosity and flourishing. Responsible asks about consequences, obligations, consent, stewardship, externalities and effects on others. Abundance asks whether knowledge, capability, health, energy, care, time, food, creativity and opportunity can be expanded rather than artificially restricted. The three terms constrain and enrich one another. Joy without responsibility can become short-horizon indulgence. Responsibility without joy can become grim control. Abundance without responsibility can become extraction. Their overlap provides an orientation for exploration. It does not impose a destination. #### GAJRA Earth GAJRA Earth-the Global Association for Joyful Responsible Abundance-is my proposed international, volunteer-led connective layer. It is not intended to become a global authority. It is intended to help communities find one another, share experiments, exchange art and music, build civic and scientific projects, test distributed systems and participate in global reflection without surrendering local sovereignty. A locally governed rack can become a GAJRA node. A school can participate. A First Nation can participate under its own protocols. A university can connect research. An island can preserve its language and environmental data. A city can model futures. An individual can participate through their Aura while keeping private material private. The network grows by resonance. A community sees something useful, adapts it and shares what it learned. That is how the infrastructure supports AI alignment without turning alignment into another top-down imposition. ### 10. Bridging the market and reciprocity braids The community compute program requires both professional markets and community participation. They should remain distinct. | Market braid | Reciprocity braid | | --- | --- | | Hardware procurement | Verified community contribution | | Electricity and network services | Mentoring and teaching | | Cooling and building works | Citizen science | | Licensed engineering | Environmental mapping | | Cybersecurity professionals | Community data stewardship | | Clinical and research staff | Disaster exercises | | Paid operators and technicians | Cultural and language projects | | Insurance, compliance and maintenance | Open-source development | | Commercial products and services | Community events and local care | The market braid uses Australian dollars and conventional contracts. The reciprocity braid uses the C-Hour. One verified hour equals one C-Hour. The C-Hour has no fiat equivalent. It is not purchased, traded for profit or translated into an hourly wage. Each community determines: 1. what forms of contribution qualify; 1. how contribution is verified; 1. which benefits are offered; 1. how many C-Hours particular access arrangements require; 1. which benefits are limited by capacity; 1. how decisions can be reviewed or changed. #### Earn-to-Heal The flagship application is Earn-to-Heal. Community contribution could provide access to: 1. clinically governed HBOT; 1. sauna and thermal wellbeing infrastructure; 1. Aura Geode and Aura Genesis research pathways; 1. community wellbeing programs; 1. preventative-health education; 1. approved local health-cooperative services. Other community-decided benefits could include: 1. queue priority; 1. VIP community-event tickets; 1. access to better community-grown produce; 1. workshops; 1. shared tools; 1. maker-space time; 1. local transport privileges; 1. equipment access; 1. community experiences; 1. recognition and credentials. These are benefits created and stipulated within the reciprocity braid. They are not market wages in disguise. This is how volunteers can share in the social infrastructure they help build rather than simply donating free labour to a government or technology supplier. ### 11. Brisbane 2032: from volunteers to civic stewards The Brisbane 2032 Olympic and Paralympic Games should be treated as a capability deadline, not merely an event deadline. Queensland’s official legacy framing already emphasises participation, belonging and benefits extending beyond the event, while the state’s volunteering strategy seeks to strengthen volunteering through 2032 and beyond. The Games will require a vast temporary workforce. The usual pattern is that volunteers are recruited, trained, mobilised and then dispersed after the Closing Ceremony. That is an enormous loss of civic energy. I propose that the volunteer system become the seed of a permanent Civic Reserve. Participants could receive just-in-time training through local racks and XR environments in: 1. event logistics; 1. crowd movement; 1. accessibility; 1. translation; 1. digital literacy; 1. local AI systems; 1. emergency communications; 1. disaster resilience; 1. community health; 1. citizen science; 1. digital twins; 1. environmental monitoring; 1. cybersecurity awareness; 1. media and creative production; 1. space-weather observation. Their verified contribution could be recognised through C-Hours and the Earn-to-Heal loop. After 2032, the network could continue supporting: 1. local events; 1. festivals; 1. disaster response; 1. environmental projects; 1. community technology; 1. health and ageing programs; 1. schools and mentoring; 1. civic deliberation; 1. scientific observations; 1. planetary and space projects. The volunteers become more than an event workforce. They become civic stewards of Earth. And as Australian space, quantum, robotics, satellite and scientific capability expands, some of those same people can become civic space stewards. They need not all become aerospace engineers. A solar-system civilisation also needs teachers, artists, coders, translators, carers, historians, medical professionals, growers, builders, communicators, organisers, sensor maintainers, citizen scientists and people capable of providing human oversight to complex AI and robotic systems. Australia is already excellent at mobilising people around sport. The next step is mobilising people around shared capability. ### 12. A Queensland frontier-compute corridor Queensland is unusually positioned to test the whole stack. #### Pullenvale: sovereign edge AI CSIRO’s Vetra at the Queensland Centre for Advanced Technologies demonstrates compact, modular computing located near sensors and physical systems, integrated with larger core and cloud infrastructure. #### Moreton Bay: utility-scale quantum ambition PsiQuantum’s Moreton Bay Central project aims to construct a utility-scale fault-tolerant photonic quantum computer using tens of thousands of photonic chips, large-scale cryogenic infrastructure and a growing Australian technical ecosystem. Its Griffith University Test and Validation Lab provides a nearby research and training anchor. #### Brisbane: event, university, health and creative capability Brisbane brings universities, hospitals, research, creative industries, major infrastructure and the 2032 mobilisation horizon. #### Redlands and Minjerribah: community and island edge testing Redland City can test how the architecture reaches local government, schools, businesses, community organisations, Country, environmental systems and disaster resilience. Minjerribah provides a practical island setting for testing: 1. local operation during network disruption; 1. sensor and digital-twin continuity; 1. ferry and transport logistics; 1. environmental and coastal models; 1. community communications; 1. renewable energy and batteries; 1. local model hosting; 1. First Nations data governance; 1. creative and tourism uses; 1. community access and training. Ready SET Co. is intended to make sustainable employment and training as important as the rack itself. #### Gold Coast and South East Queensland: regional federation The Gold Coast, Brisbane, Moreton Bay, Redlands and surrounding communities can connect multiple styles of node through one regional learning network. The opportunity is not simply to announce that Queensland contains a quantum computer, an edge system or Olympic infrastructure. It is to connect them into a public capability ecosystem. ### 13. Prepare the productive program before the next shock Governments often appear fiscally constrained until a crisis arrives. Then the constraint changes almost overnight. Australia committed more than A$90 billion through its Global Financial Crisis response over five years. COVID-era Commonwealth and state packages reached approximately A$327 billion under broad Parliamentary Budget Office accounting. Those interventions served their circumstances. The lesson is that Australia possesses the fiscal and administrative capacity to act at enormous scale when political urgency aligns. Financial markets are also becoming faster, more automated, more tokenised and increasingly capable of operating continuously across borders. Debt repricing, cyber events, climate shocks, geopolitical disruption, automated trading and new digital market structures can transmit stress more quickly than twentieth-century bureaucratic procurement systems can respond. Australia should therefore maintain a shelf-ready portfolio of productive infrastructure capable of absorbing counter-cyclical investment. A distributed civic-compute program is particularly suitable because it can: 1. be deployed across every state and territory; 1. use schools, hospitals, universities and public facilities; 1. generate skilled work; 1. strengthen local supply chains; 1. improve national resilience; 1. support small business and research; 1. create permanent public assets; 1. accelerate AI literacy; 1. increase competition and reduce dependency; 1. leave communities with capability after the stimulus ends. The question is not whether another large intervention will ever occur. The question is what we will wish we had prepared when it does. ### 14. The Protopian Gambit The Australian starting envelope is already significant. Approximately 2,780 frontier-node equivalents at A$6 million each produces a national thought experiment of approximately A$16.68 billion. That is large enough to transform national capability while remaining much smaller than past crisis interventions or the full AUKUS submarine estimate. The global scale is more provocative. The accompanying cities dataset contains 44,691 records. Approximately 41,549 have recorded populations of at least 9,000. At A$6 million per location: 41,549 × A$6 million = A$249.294 billion Using a rounded 44,000 locations: 44,000 × A$6 million = A$264 billion The public estimate for Australia’s AUKUS nuclear-powered submarine program is approximately A$268-368 billion across its life. I know these programs are not alike. That is why the comparison matters. AUKUS asks what Australia is willing to invest in deterrence, sovereign engineering and strategic capability. The Protopian Gambit asks what humanity might invest in: 1. distributed intelligence; 1. education; 1. language; 1. community resilience; 1. citizen science; 1. AI alignment; 1. disaster preparation; 1. cultural understanding; 1. local creativity; 1. open models; 1. data sovereignty; 1. peaceful international cooperation. I would like humanity to be willing to place a A$250-264 billion gambit on helping tens of thousands of communities understand themselves, understand one another and participate meaningfully in the intelligence age. Australia would not have to fund the entire world. It could begin the pattern. Australian local-government networks can connect with counterparts overseas. Chambers of commerce can identify capable businesses. Universities can connect through the accompanying dataset of thousands of institutions. First Nations can connect with Native Nations. Schools can twin projects. Hospitals can share approved research. GAJRA Earth can help communities find one another. Vendors can provide hardware at different generations and price points. Philanthropy, national governments, development banks, cities and communities can contribute according to their own capacity and sovereignty. The objective is not one Australian-controlled global network. It is a federation of locally governed networks that choose to interoperate. ### 15. From civic stewards of Earth to civic stewards of the solar system The next compute architecture will not stop at the walls of terrestrial data centres. SpaceX’s emerging architecture joins four previously separate industrial layers: 1. semiconductor fabrication through TERAFAB; 1. rapidly reusable mass-to-orbit through Starship; 1. high-capacity global communications through Starlink V3; 1. solar-powered AI compute in orbit through the programme being described as Starmind. #### TERAFAB SpaceX’s public filings describe TERAFAB as a vertically integrated advanced semiconductor initiative involving logic, memory, masks and advanced packaging, with chips designed for terrestrial and orbital workloads. Its long-term objective is manufacturing capacity equivalent to one terawatt of compute hardware per year. That is not merely a chip factory. It is an attempt to connect semiconductor design directly to rockets, satellites, AI models, robotics and energy infrastructure. #### Starlink V3 and the 100,000-satellite application Starlink V3 is designed for approximately one terabit per second of downlink capacity per satellite. SpaceX states that a Starship V3 deployment could carry up to 60 satellites and add roughly twenty times the downlink capacity of a Falcon 9 V2 deployment. Application SAT-LOA-20260630-00264 seeks authority for a third-generation Starlink network of up to 100,000 satellites. That application shows the scale at which future communications infrastructure is now being conceived. #### Starmind and orbital data centres SpaceX’s orbital-compute architecture proposes large solar-powered satellites containing substantial AI hardware and communicating through optical links. The broader Starmind ambition has been publicly described at scales reaching one million orbital compute satellites. SpaceX’s filings describe the possibility of beginning large orbital AI deployment from 2028 and eventually scaling to millions of spacecraft. #### Starship launch cadence SpaceX’s United States launch network operated from four primary pads in 2025, while Starship infrastructure is expanding through Starbase and Florida. The company describes Starship as a fully and rapidly reusable transport system with turnaround increasingly resembling commercial aviation and an eventual capacity for multiple launches per day. That changes the arithmetic of satellites, lunar infrastructure, science, manufacturing and compute. The trajectory points toward daily and then multiple-daily Starship activity as Starlink V3, Artemis, Gateway, lunar missions, commercial payloads and Starmind-scale infrastructure grow. #### Artemis, Gateway and lunar infrastructure NASA’s current Artemis architecture places Artemis IV around September 2028, combining Gateway assembly with another crewed lunar surface expedition and Starship Human Landing System operations. The Moon is moving from occasional destination toward sustained infrastructure. That will require: 1. communications; 1. power; 1. robotics; 1. materials; 1. health systems; 1. navigation; 1. digital twins; 1. autonomous logistics; 1. surface science; 1. Earth observation; 1. space-weather forecasting; 1. human-in-the-loop oversight; 1. civic and legal systems capable of governing shared activity. Australia should be preparing people for those responsibilities now. #### The Australian role Australia does not need to build a copy of SpaceX. We should build sovereign interfaces to whatever succeeds. Australian community nodes can contribute to: 1. distributed space-weather sensing; 1. optical and radio ground networks; 1. satellite data processing; 1. environmental comparison between Earth and other worlds; 1. materials research; 1. robotics; 1. autonomous-system testing; 1. digital twins; 1. citizen science; 1. public education; 1. cultural and ethical deliberation; 1. planetary-defence modelling; 1. human oversight of AI systems. A young person who learns to maintain a local sensor, train a model, inspect provenance and operate a digital twin is already taking the first steps toward civic space stewardship. The same is true of an elder preserving oral history, an artist visualising space-weather data, a nurse contributing to remote-health systems, a farmer working on closed-loop food production or a community group testing emergency communications. The solar system will not be stewarded by rocket engineers alone. ### 16. Three eclipses, one Olympic Games and a public timeline to 2037 Australia has an extraordinary series of total solar eclipses approaching. #### 22 July 2028 - Sydney and Australia The 2028 total eclipse crosses Australia and includes Sydney in the path of totality. This could become the first national rehearsal for: 1. distributed citizen-science sensors; 1. space-weather education; 1. community astronomy; 1. local AI interpretation; 1. synchronised art and music; 1. crowd and transport digital twins; 1. resilient communications; 1. national and international data sharing. #### 25 November 2030 - totality nearly reaches Brisbane The 2030 total eclipse crosses southern Australia and its path ends just northwest of Brisbane. South East Queensland could use the event as another operational rehearsal for the Games and for a wider Earth-Space-AI public network. #### 2032 - Brisbane Olympic and Paralympic Games By 2032, Australia could have trained a large network of volunteers, Digital Custodians, community developers, citizen scientists and local AI operators. The world would arrive not only at new venues but inside a region learning how to operate a federated intelligence and resilience system. #### July 2035 - GAJRA Earth’s global convergence GAJRA Earth proposes using the 50th anniversary of Live Aid as a horizon for globally synchronised art, music, civic science and participatory reflection across past, present and future. Community compute could provide the translation, media, simulation and values-deliberation infrastructure without requiring every participant to surrender their data or culture into one central platform. #### 13 July 2037 - Brisbane, Gold Coast and Quandamooka Country The 2037 total eclipse crosses Brisbane, the Gold Coast and South East Queensland, including Quandamooka Country. That provides an extraordinary culmination point: Sydney eclipse 2028 → near-Brisbane eclipse 2030 → Brisbane 2032 → GAJRA 2035 → Quandamooka and South East Queensland totality 2037 This is not merely an events calendar. It is a public learning arc for a generation. A child entering high school in 2028 could be an experienced AI developer, scientist, cultural custodian, community organiser or space-systems operator by 2037. An adult displaced by automation could enter a Try Everything Once learning pathway and develop a new civic role. A retiree could contribute knowledge, mentoring, history and citizen science. Australia could show the world that frontier technology is not reserved for executives, laboratories and fenced industrial compounds. It can become part of ordinary civic life. ### 17. Recommendations to the Committee #### Recommendation 1 - Adopt a balanced national compute doctrine Australian policy should explicitly plan for a mixture of: 1. self-sovereign personal AI; 1. neighbourhood and edge systems; 1. community-scale racks; 1. state and territory capability; 1. Commonwealth and research compute; 1. hyperscale commercial data centres; 1. quantum systems; 1. future orbital systems. The objective is a coherent ecology rather than competition between layers. #### Recommendation 2 - Establish a National Federated Sovereign AI Architecture The Commonwealth should develop standards for identity, permissions, data locality, encryption, model provenance, workload portability, revocation, open interfaces, degraded operation and interoperability across the whole stack. The core test should be: Can Australians continue to understand, use, recover and move an important capability when one vendor, network, facility or computational paradigm becomes unavailable? #### Recommendation 3 - Launch a 32-site national early-adopter cohort Fund at least four deliberately different pilots in every state and territory through host organisations capable of learning, teaching and sharing. The first cohort should include metropolitan, regional, remote, island, hospital, educational, creative, disaster-prone and First Nations-governed contexts. #### Recommendation 4 - Recruit through existing civic and learning networks Use local-government associations, chambers of commerce, schools, TAFEs, universities, hospitals, libraries, Indigenous developer pathways, girls-in-code organisations, maker spaces, creative networks and research institutions to identify capable early adopters and fan knowledge outward. #### Recommendation 5 - Make just-in-time XR training part of every installation Every funded node should include a digital twin, XR training environment, curriculum, operational simulations, public learning laboratory and obligation to teach the next cohort. #### Recommendation 6 - Use A$6 million as a transparent frontier-node planning unit Test an all-in public-infrastructure package covering compute, storage, power, cooling, heat reuse, networking, cybersecurity, secure facilities, XR training, initial support and community access. The program should preserve vendor and generation choice. #### Recommendation 7 - Develop a cascading national compute fleet As frontier systems are replaced in national, university and larger urban facilities, previous-generation systems should be redeployed into regional and community nodes where they remain useful. Public procurement should consider the full civic life of the hardware. #### Recommendation 8 - Treat local AI capability as shared national infrastructure The Commonwealth should investigate ownership and funding structures that prevent every council, school, hospital and community from separately purchasing the entire AI stack as a recurring service. Local capability belongs in the same conceptual family as libraries, communications, research facilities and emergency infrastructure. #### Recommendation 9 - Build an Australian Open Model Commons Australia should produce Australian-designed, trained and deployed models, with open architecture, weights, inference tools, evaluation tools and reproducible deployment wherever security, privacy, copyright and cultural authority permit. Open models and sovereign data are compatible. #### Recommendation 10 - Create a portable personal AI permissions standard Australians should be able to express machine-readable and plain-English rules governing training, inference, retention, sharing, commercial use, international transfer, emergency access and revocation. #### Recommendation 11 - Establish a First Nations sovereign-model pathway Support First Nations that choose to create their own models and Digital Twins of Country with authority spanning data, model training, inference, cultural permissions and derived outputs. First Nations sovereignty should cross every layer of the national architecture. #### Recommendation 12 - Pilot the Braided Economy and Earn-to-Heal alongside community compute Keep the market and reciprocity braids separate. Use conventional finance for infrastructure and professional work. Use C-Hours solely for verified community contribution and locally governed non-market benefits, including the proposed Earn-to-Heal access loop. Never assign the C-Hour a fiat or shadow-price equivalent. #### Recommendation 13 - Use Brisbane 2032 as a civic-capability deadline Connect local racks, XR training, digital twins and community learning to the Olympic and Paralympic volunteer program. Retain the trained network after the Games as a permanent Civic Reserve. #### Recommendation 14 - Develop the Queensland frontier-compute corridor Deliberately connect CSIRO Vetra, PsiQuantum Moreton Bay, universities, hospitals, Brisbane 2032, Redlands, Minjerribah, the Gold Coast and regional Queensland through interoperable classical, edge, quantum and community systems. #### Recommendation 15 - Integrate local compute with all-hazards resilience Require funded nodes to demonstrate offline and degraded operation, local maps and models, backup communications, cyber isolation, microgrid integration, restoration and workload migration. #### Recommendation 16 - Prepare a shelf-ready counter-cyclical deployment program Complete site qualification, standard designs, procurement pathways, training packages and partner networks before the next major economic shock creates pressure for rapid national spending. #### Recommendation 17 - Conduct an annual compute-horizon review Review accelerator generations, model efficiency, Australian semiconductor capability, quantum systems, robotics, energy, optical networking, satellite communications, launch economics and orbital compute every year through at least 2037. #### Recommendation 18 - Explore the global Protopian Gambit Work with local governments, universities, Native Nations, cities, international partners and GAJRA Earth to model a federated global network of locally sovereign civic AI nodes. Treat the A$249-264 billion global-town estimate as an invitation to expand the meaning of peace, resilience and planetary security infrastructure. #### Recommendation 19 - Use the 2028, 2030 and 2037 eclipses as national civic-science milestones Build a coordinated program of sensors, education, art, music, digital twins, community participation and international scientific collaboration around the eclipses and Brisbane 2032. #### Recommendation 20 - Prepare Australians to become civic stewards of Earth and the solar system Make AI, science, digital twins, data sovereignty, disaster resilience and space systems accessible to Australians across ages, professions and communities. Do not define national excellence only by the medals we win. Measure it also by the knowledge we share, the systems we steward, the communities we strengthen and the futures we become capable of building together. ### Conclusion Australia is making infrastructure choices during one of the fastest technological transitions in human history. The pace of change in AI, semiconductors, robotics, quantum computing, satellites, launch systems, energy and science is about to go bonkers. That is not a reason to surrender the future to whichever company currently has the largest facility. It is a reason to build optionality. We can have hyperscale data centres and community racks. We can work with Microsoft, Google, Amazon, Oracle, Anthropic, OpenAI and future frontier laboratories while building Australian open models. We can deploy NVIDIA, AMD and future accelerators without allowing any one vendor to become unavoidable. We can invest in PsiQuantum and preserve classical alternatives. We can connect to Starlink, Starship, TERAFAB and Starmind-scale systems without transferring the whole Australian intelligence stack into one corporate orbit. We can keep personal data personal. We can govern collective data collectively. We can give First Nations a genuine sovereign layer. We can train school students, retirees, developers, health workers, artists, carers, scientists, business owners, volunteers and local-government employees to understand and use the infrastructure. We can make community compute useful on an ordinary Tuesday and indispensable during a cyclone. We can use the Brisbane 2032 mobilisation to create civic capability that remains long after the final medal ceremony. We can ask people what they value instead of silently allowing model training and corporate terms of service to answer for them. We can build a global association around resonance, cultural difference, science, creativity and Joyful Responsible Abundance without creating another central authority. We can prepare Australians to become civic stewards of Earth. And then we can go beyond. The choice before Australia is not centralised AI or decentralised AI. It is whether we deliberately design the relationship between them. Do not put all our eggs in one basket. Build the interfaces between the baskets. Give people a sovereign layer. Give communities a sovereign layer. Give First Nations a sovereign layer. Give local government the infrastructure to participate rather than merely purchase. Give Australia the capability to cooperate with the world without becoming permanently dependent upon it. Give the next generation a civic workbench powerful enough to understand its home, model possible futures and take part in humanity’s expansion into the solar system. Australia is already one of the world’s great sporting nations. Let us also become one of its great learning, scientific and civic-stewardship nations. We Go Beyond.