Resilience in practice to 2035

Executive summary

The proposed programme should begin with an individual service that works, then combine it with funded working-time arrangements, and scale only the components whose delivery and outcomes justify expansion. People retain freedom to learn by doing outside work, choose legal personal uses, decline evaluation and seek optional backing. Existing public institutions retain public-money accountability; a contracted operator administers access. The One World Community can develop and explain the proposal, but its staff, legal personality and finances cannot be assumed.

This report completes the six-report research draft with a staged demonstration, outcome matrix, funding envelope and route to 2035. The recommended sequence is D0 professional readiness and synthetic prototyping; D1 a six-month, 1,000-person service demonstration; D2 a 4,000-person combined demonstration with time arrangements for an illustrative 1,000 employees. Each stage needs its own actual authority and resources before execution. The costed envelopes are A$0.42m, A$1.86m and A$42.18m respectively in the cautious staffing case. They are design scenarios, not quotes or spending approvals.

A1–A3 now incorporate the historical review and verified Australian hours variation. The campaign may secure reductions in steps, while D2 retains its original full-year 20% resource scenario. Any staged D2 alternative must disclose its smaller time dose and be recosted and evaluated separately; a lower cost is not the same demonstrated benefit.

The design judgement remains DESIGN PLAUSIBLE; DELIVERY UNRESOLVED. Neither the programme nor a real demonstration is launch-ready. The next implementation stage would require a synthetic walkthrough and complete operational specifications. National rollout requires further institutional, fiscal and outcome evidence.

Reader orientation, premises and confidence

B1 establishes the individual opportunity model and overseas precedents; B2 develops funding and infrastructure. A1–A3 address actual shorter-week evidence, fair Australian delivery and broad public adoption. This report integrates their dependencies rather than retesting the starting proposition as a prerequisite.

ASSUMED FOR DESIGN — P1: Some people will use released time and no-cost AI to experiment. P2: Personal learning by doing is the best foundation for discovering useful AI applications. P3: That capability can support employment and social-cohesion resilience. Their acceptance does not prove a particular service, allocation mechanism or combination will work.

Confidence is high in reproducible conditional costs and directly observed publication dates; medium in selected mechanisms transferred from overseas; unresolved in national employment/cohesion effects, procurement feasibility, actual eligibility counts and total time-reform costs. No numerical success probability is assigned. WF4 names an achieved future outcome, not the evaluation verdict today.

Pass 0 — Start with the individual and a concrete outcome

Illustrative 2035 journey. A 57-year-old permanent resident sees tasks in his occupation being automated. His funded time arrangement preserves weekly ordinary earnings and reduces actual hours. He uses a personal account to explore a local repair service, asks an AI to challenge the idea and speaks to potential customers himself. He can abandon the idea, try something else or simply use AI for daily life. An AI assessment identifies weak demand evidence; a human panel declines development funding and explains why. His basic allowance continues. A 16-year-old citizen has a separately verified age-appropriate service and can experiment without enrolling in a course. Neither is required to use an employer’s account.

The base offer is the reviewed A$1,200 annual central case, citizens and permanent residents aged 16+, no upper age. Residency administration, including overseas residence and visa-category edge cases, still needs a legal specification. No school, employer, business, income, progress, OWC-membership or evaluation condition is added. Eligibility verification and lawful account security are administration, not a test of merit. All legal purposes remain within the proposal’s intent; actual providers must be assessed for content and age restrictions that would leave that promise unmet.

WF4 requires six things together by 2035: real time/pay/service protection; usable personal access; practical capability beyond assisted performance; better work continuity and adequate income; stronger social connection and resilience to deception without increased harm; and sustainable accountable delivery. The presence of an account or a business registration is insufficient. The numerical standards for national outcome improvement cannot be invented before baseline, measurement and affordability work.

Pass 1 — Transfer overseas mechanisms and retain their limits

The useful international lesson is to remove friction between the person, compute and an experiment. It is not to import institutional eligibility conditions that defeat the proposition.

A newly identified 16 March 2026 Longgang government-hosted account reports that a developer event offered a free token pool and announced three-month OpenClaw compute support, with 500 access keys taken up. Its wording includes developer and one-person-company support, and says detailed conditions are subject to publication. This is more concrete than the earlier draft policy, but is provider self-report of an event, not an audited universal benefit or durable outcome evaluation. Translate the operational mechanism—immediate access and low experimentation cost—into an Australian personal service. Do not translate it as “China funds everyone to use OpenClaw”. B3-01: Longgang event account, Chinese original

The earlier Longgang and Wuxi draft policies remain labelled as drafts where final implementation has not been verified. The Beijing immediate-settlement mechanism, India’s compute marketplace, Singapore’s course-linked access and independent personal project grants remain useful component precedents in B1. US education/child accounts show portable individual accounts and settlement, but education restrictions or locked savings do not meet immediate unrestricted personal AI use. These are comparative design inferences, not evidence of an overseas equivalent to the whole programme.

A practical Australian combination would use easy account activation, multi-provider settlement, optional peer help and small person-held validation grants. Any university, library or community organisation offering help supplies an optional service; it cannot select who receives the base account. No provider is entitled to the allowance merely because it owns local infrastructure. Buy usable service capacity and continuity, with domestic capacity where the actual resilience, cost and service case supports it.

Historical implementation and unequal starting hours

A1's postwar cases show that the legal standard, actual time and employment outcomes can diverge. A2 now distinguishes 35, 37.5 and 38-hour instrument examples and existing RDO/shift arrangements. B3 therefore treats the time intervention as a measured change from each person's established arrangement, with sector coverage and income protections. Earlier hours reforms do not establish the AI-learning or cohesion effects assumed here.

The national campaign's C1–C5 sequence, this report's D0–D2 demonstration stages and the WF1–WF4 alternative futures serve different purposes. An initial 10% time reduction can be useful delivery within WF2/WF3; it does not establish WF4 or replace the intended final benefit. Personal AI access proceeds independently of the timing of workplace reform.

The adverse evidence matters

A1’s selected employers do not represent every sector. B1’s learning studies distinguish assisted task performance from retained ability. More capable AI can also support deception and abuse; eSafety describes deepfake harms without establishing this programme’s effectiveness against them. Therefore the design includes optional verification practice, reporting routes and harm outcomes, while retaining personal choice. B3-02: eSafety deepfakes position

Pass 2 — Demonstrate the service before claiming national resilience

D0: readiness and synthetic prototype, approximately three months

Produce a legal eligibility and appropriation specification; operator/service requirements; verified youth route; full-redemption funding arrangements; provider exit/recovery design; sector time agreements; independent evaluation protocol; and a synthetic account-to-idea walkthrough. No live identity documents, real personal prompts or real money are needed for the synthetic prototype. Its output demonstrates a proposed process, not a working national service.

Pre-D0 specification work may use fictional data for a walkthrough, instruments and tender-style requirements. Professional legal opinions, supplier quotations and ethics decisions require appropriately authorised people. Those fields remain unresolved; no approval is inferred.

D1: personal service demonstration, 1,000 people for six months

Subject to actual funding and authority, offer 1,000 eligible individuals a six-month A$600 ceiling, equivalent to the annual A$1,200 offer. Reserve the full A$600,000. Use an openly specified allocation process across the eligible population, not selection by career promise or academic record. If capacity is oversubscribed before any national entitlement exists, a transparent lottery is a candidate allocation mechanism, with its legal/ethical acceptability resolved in D0. Evaluation consent must not decide selection or renewal. This capacity limit is a temporary demonstration boundary requiring explicit approval, not the national programme’s proposed eligibility rule.

The operator verifies eligibility, opens the account, lets the person choose a service, displays the balance, settles valid use and supports switching or appeal. Declining surveys leaves access intact. Optional help and devices/connectivity are separately resourced, so “free compute” does not conceal an unaffordable device requirement. The service can be used for an ordinary personal question or a speculative project; participants do not owe the programme a commercial result.

Invite a diverse service cohort, including young entrants, older workers, carers, people with disability and non-employed people, through accessible general recruitment. This is evaluation coverage, not electoral persuasion. Record who cannot access the service and why. Youth participation cannot begin until the actual service terms and consent arrangements are verified. A youth-route failure must be visible as incomplete delivery of the 16+ design, not silently redefined as an adult-only success.

D1 asks whether account activation, choice, settlement, support, portability and appeal work. Its optional A$100,000 project pool allows up to twenty A$5,000 validation grants. It is not powered to establish reduced national job losses or better national cohesion.

D2: combined opportunity and time, 4,000 people for twelve months

After D1 readiness findings are addressed, fund a new twelve-month account period for 4,000 individuals. Some may be returning D1 users; report unique people and funded person-periods separately. Within this cohort, the resource illustration assumes 40 workplaces averaging 25 participating employees: 1,000 people receiving an actual 20% hours reduction. The other 3,000 retain AI opportunity and receive no invented employer time benefit. Analyse their caring, study, unemployment and self-employment constraints separately.

The 40 workplaces are a capacity illustration, not a statistically justified trial size or a representative national sample. Sector selection must cover credible delivery constraints, including services needing maintained staffing. Employers must fund or secure funding for agreed coverage; ordinary/usual weekly pay, work intensity, overtime and service outcomes are measured. A roster that moves work into unpaid time fails the arrangement.

Before recruitment, record each workplace's instrument, ordinary and actual hours, cycle length, attendance days, RDO accrual, overtime/penalties, normal remuneration and proposed target. Sample constraints, not merely industries: include shorter existing standards, 38-hour/RDO arrangements, long actual hours and staffed services. A 20% reduction means 35 → 28, 37.5 → 30 or 38 → 30.4 ordinary hours in A2's examples. Measure the actual change independently; an ordinary-hours agreement with unchanged total work has failed to release the promised time. The illustrative payroll model assumes its baseline captures the work needed for existing output. Unrecorded overtime, extra coverage and its remuneration must be reconciled before applying that model to a real workplace; a smaller ordinary-hours denominator cannot make those costs disappear.

A staged alternative could deliver 10% first and 20% later, but it is a different D2 protocol, not a reinterpretation of the existing twelve-month 20% scenario. Pre-specify timing and achieved time dose, allocation/comparisons, staffing costs and analysis of mixed exposure before fieldwork. Existing capacity and precision illustrations do not establish statistical power for extra stages or subgroup comparisons. Report actual four-day attendance separately from equivalent reduced-hours rosters.

D2’s optional A$1.15m award commitment pool comprises fifty A$5,000 validations, ten A$50,000 development grants and two investments up to A$200,000. These are stage commitments, not necessarily 62 unique recipients or cash paid on day one. The larger national A$250m pool in B2 remains a separate scenario; do not add it to this demonstration budget.

AI assesses; people decide

For an optional proposal, AI prepares a structured assessment: the individual’s goal; problem and proposed benefit; evidence supplied; customer/user validation; costs and cash needs; relevant legal/operating risks; alternatives; uncertainties; and the smallest useful experiment. Separate evidence from generated estimates. A second model may challenge the assessment, but agreement is not independent confirmation. Do not reject an applicant automatically because a score falls below an AI threshold.

The applicant sees and can correct factual errors. A human panel with conflict declarations decides each grant or investment, records reasons and any departure from the assessment, and offers human reconsideration. Incorporation is not required to learn or seek a validation grant; the legal vehicle for an actual equity investment is resolved at that stage, with the individual remaining the intended beneficiary. Do not manufacture predicted returns, require personal guarantees in the base design or expose personal prompts to investors. Awards can use agreed milestones; the base allowance cannot. A failed project does not create base-access debt.

No real award assessments have occurred. The next synthetic walkthrough should include a promising idea, an incomplete idea, a mistaken AI assessment, a human refusal and a successful appeal. It should also show a person who never applies for funding.

Pass 3 — Evaluate delivery and outcomes without coercing learning

The Commonwealth Evaluation Policy supports proportionate, independent and useful evaluation planned with the programme. That is a relevant existing-government framework, not a reason to create a new agency. B3-03: Commonwealth Evaluation Policy

The 2025 National Statement took effect on 23 June 2026. A prospective research sponsor must determine the appropriate independent ethics review and consent arrangements, including youth participation and linked data. Older toolkit references to earlier editions do not establish the current standard. B3-04: NHMRC current National Statement

Comparison design

Maintain the four policy pathways as the national scenario framework. Do not label four convenient groups “WF1–WF4” and claim to have randomised Australia’s future. D1 is primarily a process evaluation. D2 should compare consenting workplace clusters receiving earlier and later time arrangements, only if a funded, ethically acceptable staggered rollout is feasible and all retain the personal AI offer. That comparison estimates the incremental effect of earlier time release within the study, not the total effect of universal AI access versus BAU.

If random timing is infeasible, use a prospectively specified matched comparison with pre-period outcome trends, explain remaining selection bias, and narrow causal claims. Observe BAU and incremental-policy context through compatible external data or separately consented comparison research, not by removing an established entitlement. Self-funded AI use in comparisons and spillovers between workplaces must be measured. Analysis by original assignment is the primary approach where randomisation exists; usage-based associations remain secondary. Treasury’s trial guidance identifies randomisation, power, spillover and missing-data issues; the detailed choices here are proposed applications. B3-05: ACE trial guidance

Before fieldwork, preregister outcomes, estimands, cluster allocation, follow-up periods, exclusions, missing-data handling and analysis. Independent statistical design must set sample size using meaningful effects, expected variation, intra-cluster correlation and attrition. No such power analysis has been completed. The supplied precision calculations are descriptive illustrations only: with 1,000 clustered workers, 80% response, original cluster size 25 and ICC 0.05, the design effect is 2.2 and effective sample size about 364. Counting all 1,000 as independent would overstate precision.

The time/pay dataset should also preserve contractual full-time status separately from ABS statistical full-time/part-time categories. A reduction below the statistical hours threshold can change the latter without a lost job. Track retained jobs, involuntary losses, earnings and wanted hours directly; also measure hiring and effects on coworkers, since historical studies show those margins can differ. SH10: ABS methodology and glossary

Outcome and decision matrix

All thresholds below are proposed operating rules, not validated scientific cut-offs. Outcomes are collected before access/time changes and, where appropriate, at months 3, 6, 12 and 24. Longer follow-up requires continuing authority and resources. Consent is separate from service delivery; routine minimum settlement records do not authorise collection of private prompt histories.

Scanlon’s 2025 material treats cohesion as multiple dimensions and reports associations with financial circumstances and local connections. This supports broad measurement rather than equating model use with cohesion; it does not show that AI causes those outcomes. Reuse validated items with appropriate permissions and unchanged wording where comparability matters. B3-06: Mapping Social Cohesion 2025

Set programme-level operational targets before recruiting: candidate D1 criteria are at least 95% of valid activation attempts completed within five business days, at least 99% correct sampled settlements, and every critical incident given a named human owner within one business day. These are negotiable service targets with disclosed denominators and audit samples. They do not prove employment impact. Fund and remedy owed access when delivery misses a target; do not penalise the individual. D2 expansion additionally requires agreed sector service margins, an adequate workforce and no unresolved material pay or safety breach. National outcome thresholds remain a pre-rollout decision informed by baseline and meaningful-benefit work.

The evaluator reports adverse, null and favourable results, attrition and subgroup uncertainty. Accessible consent, withdrawal and complaints must work outside employers. Employers receive aggregate findings rather than personal learning or survey records. Do not collect political preferences or build individual “cohesion scores”. A severe incident may require pausing a particular integration or award process; it does not justify cancelling everyone’s entitlement.

Pass 4 — Fund the complete experiment and compare pathways

Illustrative stage envelopes

All amounts are September 2026 gross planning dollars, without assumed investment returns or avoided-unemployment savings. The model reserves full allowance use rather than financing promises from expected non-use.

D2’s A$5.35m comprises administration/help/devices A$0.4m each, infrastructure A$0.5m, evaluation A$1m, management A$0.5m, awards A$1.15m and time redesign A$1m. Its assumed baseline payroll is A$100m for the 1,000 workers. At a 10% effective hourly gain, the total envelope is about A$28.54m; at 25%, A$12.18m. The latter is arithmetic, not a credible default for direct coverage. Actual skill mix, staffing availability, tax, insurance, supplier prices and participant support could change all envelopes.

For comparison only, a D2 alternative with six months at 10% and six months at 20%, no effective hourly gain and the same A$100m annual baseline payroll would require approximately A$18.06m additional payroll over the year, yielding an envelope of A$33.85m with the same other costs and contingency. It delivers less released time than twelve months at 20%. The calculation assumes proportional staffing costs and excludes extra stage-transition costs, whole-recruit constraints and any longer evaluation contract. It is not a quote or a selected replacement for D2. calculate-sector-hours.py reproduces this separate sensitivity.

D0–D2 total A$44.46m in the cautious case. A3’s separately managed A$1.2m campaign would make A$45.66m if all stages and that campaign were funded, without overlap. D2 is a total resource envelope; the split between public and employer contributions remains to be agreed. Treat it as a public budget only if government funds the whole amount. Expected 75% redemption leaves A$150k of D1 and A$1.2m of D2 entitlement reserved, not available to spend elsewhere while owed. The 20% contingency is additional.

The envelopes cover the stated service periods and contracted analysis provision, not an automatic ongoing entitlement after the trial. Before recruitment, disclose the funded period and transition plan. No abrupt cliff should be hidden: extension or national continuation needs its own funding; provide export, balance expiry notice and optional continuation routes. Continuing 24-month data collection must fit the evaluation contract or receive additional funding. Do not describe finite demonstration participation as the enacted national entitlement.

Four pathways and common stresses

National AI totals reuse B2’s 20m illustrative eligibility, 50% activity and 75% redemption. At full activity/redemption, WF3/WF4 AI costs become A$26.55bn under that model. T2/T3/T4 are still unpriced national time costs. Demonstration envelopes are separate stage resources, not additional perpetual national allocations. Avoid double counting them if absorbed into a future national budget.

All four pathways face the same shocks. Faster job displacement requires accessible help and honest income outcomes; it does not establish that entrepreneurship can replace all lost employment. Fiscal tightening requires openly revised scope before new promises. Provider exit requires funded migration and portability. Power/water constraints require actual infrastructure capacity and procurement choices, not sovereignty claims alone. A large abuse incident requires human-led response and evidence-based corrections. Under any shock, WF4 remains an endpoint to work toward; if its conditions fail, report the achieved pathway components and the unresolved gap.

Pass 5 — An accountable route from now to 2035

OWC’s realistic initial contribution is an openly reasoned proposal, reproducible scenarios, a practical demonstration design and public accountability questions. Public institutions decide taxation, expenditure and national eligibility. Employers and worker representatives negotiate lawful time arrangements; service authorities protect coverage. The contracted operator delivers accounts; an independent evaluator measures outcomes; people decide awards. These are proposed roles, not claims that organisations have agreed.

Objections, current judgement and handoff

A voluntary service may mostly attract people already advantaged. Address access barriers and measure offers, non-use and exclusion, without conditioning support on activity. AI assessment may reproduce bias or polished but false claims; provide correction, human decision and audit. A shorter week may intensify work or reduce earnings; measure actual hours, pay and service quality and repair affected arrangements. Financial pressure and housing can dominate social cohesion; the programme must not claim to solve every underlying cause.

DESIGN PLAUSIBLE; DELIVERY UNRESOLVED. All six reports are now drafted. There is sufficient design detail for human review of a later preparation package, but no appointed/funded operator, legal instrument, representative support survey, powered trial protocol or national time-cost account. These gaps prevent a launch-ready judgement; they do not require proving P1–P3 before preparing a concrete prototype.

The integrated direction remains subject to human acceptance or amendment. D0, legal authority, funding, provider terms and evaluation design remain prerequisites; spending, external contact, recruitment and campaign launch require separate decisions.

References and source locations

B3-01–B3-06 link directly beside supported claims. The public source register records publication/retrieval dates and limits. A1–A3 incorporate historical evidence and sector-specific hours. The public action inputs, calculator, budget scenarios and precision scenarios reproduce this report’s arithmetic.

Glossary

D0/D1/D2: preparation, service demonstration and combined demonstration, distinct from WF policy pathways. ICC: correlation between outcomes within a workplace. Estimand: the precise effect a study aims to estimate. Full redemption: use of the whole funded allowance. Transfer: applying capability to a new problem, including after assistance is removed. B0/T: existing policy expenditure and incremental working-time costs respectively.

Authorship and method

Rick Molony owns the proposition and decisions. Primary-source research, translation, synthesis and calculations were developed with AI assistance. The translation is not an official translation. No people were recruited, no prototype was run, no models assessed real funding applications, and no national effects were observed. An independent professional review is still required before field execution. Evidence checked 9 September 2026.

Suggested citation

How this page was derived: the full substantive source body was rendered after the declared public-context transformations, removal of private delivery/workflow material and omission of the private change log.