A public-interest sensorium would let communities compare maps, data, law, history, infrastructure and scientific models without turning citizens into raw material.
Between a laptop and a giant corporate data centre sits the missing middle: serious shared computing people could see, learn to operate and use in their own community. The proposal connects personal AI, neighbourhood systems and local racks to larger research and cloud systems, with local authority over data and a practical choice of providers.
The missing middle is shared high-performance computing between personal devices and enormous data centres. Think of a library of serious tools: local people could learn, build and run models, then connect to larger systems when useful.
A willing library, TAFE, hospital, university, co-operative or community could host a properly engineered computing cluster linked to an accessible workshop and learning laboratory.
AI-assisted just-in-time learning could help local technicians, students, artists, businesses and researchers use it for real projects: food tools, legal source retrieval, simulations, film, environmental models and local services.
Local systems could share authorised results and model improvements rather than send everyone’s raw records away. The same assets could switch to essential maps, directories and communications during disruption; larger systems remain available when a workload needs them.
Everyday example
A Year 10 student could learn AI on serious equipment nearby; a small business could develop a tool; a community could model a flood. They would gain skills and options rather than only rent somebody else’s black box.
Who could act and how it could be funded
The Fair Go submission treats this as shared national infrastructure, with state and local partners. The AUKUS submission sketches a 32-site learning cohort at a $192 million planning allowance; the Fiji proposal explores a $100 million connected package. These are distinct brainstorming scenarios, not committed programmes.
Fair-go safeguards
People and cultural custodians choose permissions. Hardware remains replaceable, with trained operators, security, energy, water and maintenance included. A rack is a shared capability asset, not a claim of automatic privacy or disaster reliability.
What would show progress
Skills and useful work; local access; custody and provider exit; resource use; disaster continuity; shared learning; public value.
The AUKUS submission asks what a comparable investment in distributed intelligence, education, science, disaster resilience and peaceful cooperation might make possible. Its missing-middle network would spread capability through communities; Pillar Two partnerships could transfer civilian skills and open interfaces. The larger peace gambit invites willing towns and Pacific partners to learn and cooperate on their own terms.
A Protopian Gambit means a useful first move that leaves more options for the next one. The submission asks what distributed intelligence and cooperation could contribute to security beside a concentrated military investment.
Publish a portfolio comparison of commitments, delivery times, dependencies, alternative uses and credible failure scenarios.
Pilot civilian capability in research, disaster resilience, learning and shared compute with paid operators and public-interest governance.
Build partnerships that reward useful shared capability and de-escalation, and review the portfolio when assumptions fail.
Everyday example
A regional institution gains equipment, training and disaster capability with daily civilian use; that benefit is assessed separately from a submarine's military role.
Who could act and how it could be funded
Defence and foreign affairs are Commonwealth responsibilities; Victoria can act through research, industry, education and resilient infrastructure. The submission's figures remain proposals; obtain independent cost and capability assessments.
Fair-go safeguards
Civilian computers do not perform a submarine's job. No fabricated peace guarantee or effortless budget transfer. Protect sensitive information while publishing objectives and scrutiny arrangements.
The Web3 Sensorium is a draft research direction for a shared digital twin of civic and natural systems: open-source where possible, local-first where privacy matters, and clear about what is known, what is modelled and what is speculative.
Its job is not to make a giant public database of everyone. Its job is to let communities test claims, compare evidence, model risk and argue with assumptions in the open.
Community intelligence · October 2026
Serious compute. A local say.
Between a phone and a hyperscale centre sits the missing middle: serious computing controlled by towns, islands, libraries, universities, hospitals, cooperatives and communities. A school can learn on real equipment; a hospital can keep protected workloads close; filmmakers can render; researchers can model reefs, farms and public infrastructure. The machine follows the purpose.
That gives the Sensorium somewhere to live: local models, maps, legal memory and simulations held by people who can inspect, maintain and change the system. Sovereignty means custody, keys, skills, recovery and the ability to leave a provider, as well as where the machine sits.
Personal / householdPrivate AI, personal memory, keys and permissions on devices people control.
Edge / neighbourhoodSensors, local communications, disaster kiosks and fast nearby services.
Community missing middleSerious shared racks or clusters for larger models, simulation, rendering, training and public access.
State / nationalResearch, public-interest workloads, evaluation, coordination and sovereign backup.
HyperscaleFrontier training and elastic capacity, with portable workloads and meaningful provider choice.
Specialised horizonsQuantum and possible future orbital systems, connected through chosen interfaces. These are distinct capabilities, not a delivered local service.
This continuum comes from Senate proposal, which explicitly calls for interoperability and permissions standards like the Web3 global sensorium proposal. It does not rank people or require every workload to travel upwards.
Straddie’s L0-L2 labels describe scene briefs: private spaces, local places and bioregions. They are not these compute tiers. A bioregional scene brief can be written on a laptop; the chosen rendering job may need a different machine.
Community intelligence · October 2026
Ready S.E.T.: the node around the rack.
The Dunwich cultural intelligence node is a public working proposal. GB200 NVL72 supplies a concrete rack-scale reference, replaceable across vendors and generations. It is not evidence that a rack has been bought, installed or commissioned. Real workloads, a site pathway, power, liquid cooling, fire protection, secure networking, maintenance and lifecycle costs must determine the build.
Equipment + peopleA machine people can use and maintain
Secure infrastructure sits alongside a visible workshop, paid operators, creators, learners and technical support.
Custody + applicationsPermission stays attached
Viewing, storage, transcription, model work, publication and review are separate choices. Technical administration does not grant cultural authority.
Chosen bridgesShare selected work
A connection names its cargo: a public method, selected output, benchmark, tool or learning exchange. Local identity and custody remain visible.
The proposed compute node builds around work that already has public doors. The local brief builder runs in the browser tab and exports Markdown. The model workbench traces a selected question through sources, proposed rack work, outputs and human review. Neither requires pretending the physical rack is operational.
The catalogue distinguishes the working Quandamooka Country Events Engine, the Gumpi ferry evidence prototype, the public citation-first Australian Legal Engine and Straddie scene builders from proposed screen networks, place projects and rack workloads. Follow each project’s own status and source trail.
Start with one question: what service should stay available when the island loses its outside connection? Record the people responsible, minimum map or directory, permissions, power and cooling budget, recovery exercise and what may be shared afterwards.
Senate submission proposes a varied early-adopter network, public learning, open operating playbooks and published failures as well as successes. Its A$6 million all-in frontier-node allowance and approximately A$600 million for 100 sites are planning allowances, not vendor quotes or an approved budget. Smaller and reused equipment remain live choices.
Local rackMore local control and practical learning, with demanding maintenance, power, cooling and utilisation responsibilities.
Shared regional capacityPool scarce skills and equipment across hosts, while testing who gets access and what survives a network outage.
Personal + cloud firstLower initial local capital needs, but service dependency, recurring fees and exit costs must be visible.
Ask a library, university, hospital, council or cooperative to host a workload workshop. Use the node brief to name daily uses, paid roles, cultural permission, procurement alternatives and a stop-or-expand review. Publish measured utilisation, energy, cooling and water use, recovery results and community outcomes. Heat reuse needs an actual nearby heat demand and engineering; distribution alone does not erase resource impacts.
People, groups, councils and research teams should be able to hold their own data, work offline where needed, and share only what they choose or are lawfully authorised to share.
TwinDigital twin, not digital master
Maps, sensors, ledgers, legal constraints and scenarios can be modelled together, but the model remains a tool people can challenge.
DebateKnown science and wild ideas
Established models and speculative hypotheses can sit in the same arena if labels, evidence, uncertainty and prediction tests are visible.
External example
Straddie digital twin prompt builders.
The Straddie Digital Twin Builders site is a working example of the lighter side of this sensorium: plain-English forms that generate Markdown prompt packets for outside image, video, 3D, 4D and world-building tools.
It uses L0 for private spaces, L1 for local places and L2 for bioregion-scale briefs. This link leaves P4A. The pattern is useful here: describe the scene, choose the level, label the sources and keep sharing controls visible.
A future sensorium could bring together public data streams such as weather, space weather, earthquakes, infrastructure condition, environmental monitoring, historical maps, public records, community reports and legal sources.
Time-series memorySensor readings, events, alerts and observations need clean timestamps, units, locations and source labels.
Knowledge graphPeople, places, laws, assets, risks, documents and events need relationships, not just files in folders.
GraphRAGAI should answer from cited civic memory, not freewheel over law, science or public claims.
Civic use case
For P4A, the first practical version is grounded in everyday needs: public assets, council powers, climate and disaster risk, food resilience, energy, water, insurance pressure, state law and transparent project ledgers. The same source trails can support larger scientific questions while communities retain control of what they share.
Web3 without the grift
Decentralised identity, verifiable credentials, peer-to-peer sync, signed contributions and community governance can be useful. Tokens, markets and DAOs are not magic. Any civic Web3 layer must prove it improves consent, resilience, provenance or accountability before it earns a place.
Hard guardrail
Public-interest simulation must not become public surveillance. Private homes, health data, cultural knowledge, personal histories and sensitive community signals need consent, aggregation, access rules, human review and the right to stay out.
Last research run
3 October 2026. Updated against UNGA81 capability page, Ready S.E.T. public pages and 24 August Senate submission. Sensorium architecture and the physical node remain proposals; the linked brief builders and public project tools have their own working status. These options are open for public consideration, not adopted party policy.
Large data centres: inspect the actual bargain
Water demand depends on the cooling design, climate, operating load and water source; grid effects depend on connection and supply arrangements. Ask for project-specific estimates, measured operation and who pays for new infrastructure. A blanket claim that every data centre drinks more than its town cannot substitute for the numbers.
Conditions on large projectsRequire transparent costs, resource plans and community benefits, while keeping approvals and compliance clear.
Shared local capabilityThe missing-middle proposal adds public access, training and portable workloads. It needs a host, operators and a realistic budget.
Demand and siting choicesCompare efficient workloads, suitable cooling and locations against local constraints. Moving impacts elsewhere is not the same as reducing them.
PM&C’s September consultation covers large data centres and conditions for AI training, including Australian creators. Its published closing date is . Use the official page to check whether submissions are still accepted; after that date, follow published outcomes and later consultation rounds.
Earlier clip briefs and proposals are kept here alongside the current evidence and options above. These are exploratory ideas, not adopted policy.
Twinkle 11 - earlier seed and related proposals
A back-fence rant about the giant grey shed that landed up the road with its own water deal and grid deal, then a hard turn into local-first compute and towns that get a twin instead of a bill.
Open seed
The everyday gripe
Straight to camera: a hyperscale data centre lands on the edge of town and residents want the water source, cooling demand, grid costs and public benefits explained. The answers depend on the actual project and its agreements. The trapdoor is the town getting its own twin: local terms in public, local compute as part of the deal. This seed is open: film your own grey shed.
Description
BaitThese sheds are landing across the country right now, and every major party is courting the companies building them. Residents should be able to inspect the public terms and challenge the assumptions.
SwitchFrom "progress, I suppose" to "where is our slice": water, power and data are local resources, and extraction without local benefit has a long history here.
SystemThe Web3 Sensorium: local-first nodes and digital twins, so the place that hosts the compute also holds the terms, the data and a share of the benefit.
The one-breath drop
Said before the flush
"If it uses our water, our grid and our data, the town gets the terms in public and a slice of the compute. No twin, no deal." Second breath lives in the room.
