AppStore2031
Evidence register
Each record preserves its publisher, date, geography, evidence class, intended use and important limitations. A source supports only the claims linked to it.
Sources do not validate the fictional products or ranks. They ground current conditions and directional pressures at the evidence cut-off.
measured-trend · AC-06
Supported use: Defines 50% and 80% task horizons and publishes data/code.
Important limit: The metric is human task-equivalent difficulty, not elapsed autonomous runtime; suite coverage is limited
measured-trend · AC-09
Supported use: Best systems passed only 26.7–28.9% end to end and overestimated their own success by 66.6–97.8 points; supports semantic-contract and self-validation limits.
Important limit: Preprint, one task family
measured-trend · AC-21
Supported use: Finds high concentration, capital barriers, vertical integration, state intervention and bottlenecks across chips, cloud, energy and networks.
Important limit: Some market shares rely on third-party estimates
policy-intent · AC-25
Supported use: Policy intent across deregulation/innovation, domestic infrastructure and full-stack exports.
Important limit: Political plan, not evidence of implementation or outcomes
policy-intent · AC-27
Supported use: Policy intent for AI integration and stated agent/smart-terminal penetration above 90% by 2030.
Important limit: The percentage is a government target whose definition and measurement are not yet independently established
policy-intent · AC-30
Supported use: Observed phased EU obligations and July 2026 simplification/timeline changes.
Important limit: Implementation and enforcement outcomes remain unobserved at cutoff
measured-trend · AC-33
Supported use: Reports shared facility above 45,000 GPUs, public compute access, labs and skills programmes.
Important limit: Government-reported counts do not equal frontier-equivalent or fully utilised compute
policy-intent · AC-36
Supported use: Policy intent for an Africa-centric, development-focused, inclusive and responsible path, including skills, data and infrastructure.
Important limit: Delivery depends on diverse national capacities and finance
measured-trend · AC-38
Supported use: Observes high adoption but major talent, investment and infrastructure gaps; 90% of regional supercompute in Brazil.
Important limit: Composite indicators and country reporting have limitations
measured-trend · AC-39
Supported use: Evidence of energy/state-capital-backed compute strategy; UAE 5 GW is announced campus scale, while Saudi reports over 440 MW operating in 2025.
Important limit: Promotional sources; announced partners and capacity are not assured delivery
measured-trend · C01
Supported use: Temperature, ocean heat, energy imbalance and extreme events.
Important limit: Single-event attribution requires separate studies
measured-trend · C03
Supported use: Risk escalation, adaptation limits and mitigation pathways.
Important limit: It is a synthesis with an evidence cut-off earlier than 2026
modelled-projection · C04
Supported use: Adaptation-finance need and flow gap.
Important limit: Needs estimates vary with modelling scope and do not measure effectiveness of every funded action
modelled-projection · C05
Supported use: Direct and cascading disaster costs.
Important limit: The full-cost estimate is model dependent and broader than recorded losses
measured-trend · C10
Supported use: Insurance coverage gap by peril and country.
Important limit: Historical coverage does not predict future availability or price
observed-deployment · E03
Supported use: Installed capacity through 2025.
Important limit: Capacity does not measure usable output, congestion or firm supply
modelled-projection · E05
Supported use: Investment, connection queues, lead times and supply-chain constraints.
Important limit: Queue totals include projects that will not all be built
scenario-assumption · E08
Supported use: Data-centre electricity demand and uncertainty.
Important limit: Workload growth, hardware efficiency and utilisation make the range unusually wide
scenario-assumption · E13
Supported use: Concentration, price shocks, investment and projected supply gaps.
Important limit: Commodity forecasts are sensitive to demand, substitution and new supply
scenario-assumption · E18
Supported use: Demand, import exposure, renewables and access.
Important limit: Published during an acute Middle East fuel shock, which may not persist
scenario-assumption · E21
Supported use: Reported connections versus remaining access gap.
Important limit: Connection counts do not establish reliability, affordability or productive use
modelled-projection · EWF-02
Supported use: Global public debt was just under 94% of GDP in 2025 and is projected to reach 100% in 2029; social, defence, strategic-autonomy and interest costs compete for fiscal space.
Important limit: Aggregate hides large country differences; projection is conditional on policy and macro assumptions.
measured-trend · EWF-04
Supported use: One in four workers is in an occupation with some GenAI exposure; 3.3% of global employment is in the highest-exposure gradient; transformation is more likely than wholesale automation.
Important limit: Task exposure is not displacement, adoption, productivity or wage effect.
measured-trend · EWF-05
Supported use: Developing economies may face automation pressure on connected formal jobs before augmentation gains reach less-connected workers; 66.9m of 441.8m augmentation-exposed jobs in detailed data lacked internet access.
Important limit: Internet access is a necessary but insufficient proxy for productive use; occupational structures will change.
measured-trend · EWF-07
Supported use: Sixteen experienced developers completed 246 tasks 19% more slowly when early-2025 AI tools were allowed, despite believing they were faster.
Important limit: Narrow sample, repositories and tool vintage; the follow-up says newer-tool estimates were selection-biased and only very weak evidence about the size of any improvement.
observed-deployment · EWF-14
Supported use: More than half the world had at least one social-protection benefit, while 3.8bn people remained entirely unprotected; coverage is weakest in climate-vulnerable states.
Important limit: “At least one benefit” does not measure adequacy, portability or shock responsiveness.
measured-trend · EWF-15
Supported use: Two billion people were uncovered or inadequately covered; strong delivery infrastructure enabled faster shock response; current progress would take many years to close gaps.
Important limit: Coverage estimates use heterogeneous and sometimes lagged national data.
policy-intent · EWF-18
Supported use: About 8m young people enter the regional labour market annually and fewer than 1m secure wage jobs; the programme intends skills and job investment through 2034.
Important limit: Programme targets are intent, not outcomes; region is not all Africa.
measured-trend · EWF-20
Supported use: Major cloud/AI partnerships include equity, revenue-sharing, cloud-spend commitments, IP/data access and control rights that may raise switching and input-access concerns.
Important limit: Examines a small set of partnerships through 2025-01; identifies risks, not adjudicated violations.
measured-trend · EWF-22
Supported use: Initial evidence reports that AI patenting correlates with faster markup growth, while smaller firms may also gain from access; AI start-ups are dynamic but often acquired.
Important limit: Correlation is not causation and the paper is an initial cross-source analysis, not a long adoption history.
modelled-projection · F03
Supported use: Non-linear and climate-zone-specific yield responses.
Important limit: Results inherit process-model and site-data limitations
measured-trend · HBD-05
Supported use: Service coverage and financial-hardship baseline.
Important limit: Indicators were revised; older report series are not directly comparable and national aggregates hide exclusion.
measured-trend · HBD-08
Supported use: Regional distribution, conflict exposure and 2030 workforce pressure.
Important limit: A regional leadership statement summarises underlying data; it is not a country-by-country workforce account.
measured-trend · HBD-11
Supported use: Scale of public digital health accounts, linked records, immunisation tracking and teleconsultation.
Important limit: Government administrative counts do not by themselves establish unique active users, clinical quality, outcome gains or inclusion.
measured-trend · HBD-12
Supported use: Ageing, unpaid care and gender implications.
Important limit: An institutional policy essay drawing on ECLAC data, not a new causal evaluation.
measured-trend · HBD-13
Supported use: Legally staged exchange and secondary-use timetable through 2031.
Important limit: Entry into force is not operational interoperability; implementing acts, national readiness and enforcement remain consequential.
measured-trend · HBD-16
Supported use: International convergence around representative data, human-AI teams and lifecycle monitoring.
Important limit: Principles require jurisdiction-specific implementation and do not prove enforcement or performance.
measured-trend · HBD-47
Supported use: End-to-end pathogen sequencing, sharing and public-health-use goal.
Important limit: Strategy does not prove sample representativeness, timely sharing or national analytical capacity.
measured-trend · PC-S01
Supported use: Treat electricity, grid connection, location, cooling and supply chains as real limits on abundant digital cognition. Model regional differences instead of assuming one global compute market. Use scenarios and ranges rather than one certain electricity number.
Important limit: The 2030 and 2035 figures are modelled scenarios, not promises. Data-centre electricity is not the whole environmental footprint of AI. The report does not show how capacity will be allocated to individual people or firms.
measured-trend · PC-S02
Supported use: Stress-test a 2031 scenario where software capability grows faster than physical capacity. Shows why transparent capacity reservation, flexible execution and recoverable handoffs matter. Keep efficiency and rebound effects together: cheaper tasks do not guarantee lower total use.
Important limit: Some 2026 values are estimates and the report says not all announced projects will be built. A task-energy comparison does not decide whether a use is socially valuable. This source does not establish that compute must be tokenised or put on a blockchain.
modelled-projection · PC-S03
Supported use: Represent large uncertainty in US load rather than a single precise forecast. Test whether services can schedule, relocate or reduce energy-intensive work.
Important limit: This is a US-only bottom-up model, not a global forecast. Shipment, lifetime, utilisation and cooling assumptions drive the range. It does not predict local tariffs, queue rules or who receives service.
measured-trend · PC-S04
Supported use: Keep water and cooling visible beside electricity. Ask for location-specific resource disclosure when physical infrastructure affects a community.
Important limit: Its future range ends in 2028 and was superseded for electricity by the 2026 update. Water impact is highly local and cannot be inferred from a national total alone. The existence of water use does not justify private control of public water allocation.
policy-intent · PC-S05
Supported use: Represent China as an active shaper of compute infrastructure and data-market rules. Test cross-region interoperability and jurisdiction-aware execution.
Important limit: A target is not verified delivered or generally available capacity. EFLOPS measures are not automatically comparable across all hardware and workloads. The plan does not show that an individual can buy or move capacity freely.
measured-trend · PC-S06
Supported use: Model compute placement as linked to power location, carbon rules and network transport. Show that China-specific operating evidence can differ from US and EU signals.
Important limit: The 80% figure is a policy objective, not independently verified achievement. PUE does not include all electricity-system, water, construction or chip impacts. This Chinese-language source requires careful translation and should not be paraphrased as a global rule.
policy-intent · PC-S07
Supported use: Represent EU capacity-building and trust rules together. Test whether a service can explain where work runs and which legal or sustainability constraints apply.
Important limit: This is a policy plan, not proof that gigafactories or tripled capacity are operating. EU-wide goals hide national grid and permitting differences. The source does not establish open public access to supercomputing capacity.
measured-trend · PC-S08
Supported use: Show a shared-access model rather than assuming capacity belongs only to hyperscalers. Test price, hardware fit, eligibility and availability as separate parts of a compute request.
Important limit: Empanelled capacity is not the same as continuously available capacity. GPU counts do not capture performance differences, networking, storage or software readiness. Government reporting is not evidence that every person can access every listed resource.
policy-intent · PC-S09
Supported use: Support the idea that work may move to where power and networks can support it. Represent location and transmission as linked constraints.
Important limit: The report states policy direction, not delivered capacity or a public scheduling interface. Moving computation is not always possible because latency, data residency and security matter. Private software cannot replace grid, telecom or land-use planning.
policy-intent · PC-S10
Supported use: Keep procurement, ownership structure, power connection and location in the build story. Treat policy ambition and execution risk as separate facts.
Important limit: Most capacity numbers are planned procurement or construction, not delivered service. A public-private centre does not imply unrestricted person-level access. The source is one national plan and should not be generalised to all of Asia.
measured-trend · PC-S11
Supported use: Treat factory robotics as a large, geographically uneven installed base. Supports analysis of assigning, supervising, accepting and recovering physical work done through machines.
Important limit: Industrial robots are mainly bounded machines in factories; they are not proof of general-purpose home robots. Installations do not measure autonomy, reliability, task breadth or person-level access. The 2028 number is an industry forecast and could change with investment and trade conditions.
measured-trend · PC-S12
Supported use: Shows current growth in time-limited access to specialised machines. Distinguish controlled-site execution from open streets, homes and emergency settings.
Important limit: The data comes from a changing sample of 294 suppliers and IFR says it must not be projected to the whole industry. Year-to-year comparisons across changing samples are discouraged. Sales and fleet growth do not prove that a robot can safely carry out a person's custom request.
measured-trend · PC-S13
Supported use: Keep fabrication, packaging, yield, site construction and geography in the causal chain behind compute supply. Show that more chips require multi-year physical investment, not only better software.
Important limit: This is one company's disclosure and forward-looking plan. It does not reveal all customer allocations or supply-chain constraints. A scheduled technology ramp is not guaranteed availability or affordability to end users.
modelled-projection · PC-S14
Supported use: Corroborate that data-centre load growth was treated as a national electricity-planning issue.
Important limit: This short announcement adds little beyond the underlying LBNL report. Its 2028 scenario was later updated by LBNL's 2026 report.
policy-intent · PC-S15
Supported use: Test physical-execution ideas across several settings without assuming one universal robot. Keep safety, testing and sector rules as prerequisites rather than optional features.
Important limit: A policy page is not deployment data. It does not establish that open-world autonomy will be reliable by 2031. Regulated care, transport, safety and security duties cannot be delegated to private software merely because a robot is involved.
measured-trend · RSM-02
Supported use: Reports robots per 10,000 manufacturing employees for 2024.
Important limit: Density is a stock-to-employment ratio, not a measure of autonomous task coverage or productivity
measured-trend · RSM-13
Supported use: ASIMOV tests semantic safety beyond collision avoidance; OopsieVerse distinguishes nominal task success from safe success when property damage or hazardous execution occurs.
Important limit: Benchmarks expose measurement gaps but do not yet provide a universal safety score for deployed robots
measured-trend · SC-AI01
Supported use: Support mixed short-term associations between AI use and psychosocial outcomes Distinguish brief or bounded use from heavy voluntary use Keep both augmentation and substitution branches open
Important limit: Short study period, one service context and selective user samples; heavy voluntary use was not randomly assigned.
measured-trend · SC-AI02
Supported use: Report the four-week randomized study design Support the absence of a simple interaction-mode effect Support associations between heavier voluntary use, attachment and worse outcomes without claiming causality
Important limit: Preprint; 28-day duration; no global representation; post-randomization voluntary usage intensity limits causal interpretation.
policy-intent · SC-AI03
Supported use: Show regulatory concern about companion-chatbot safety, children and data practices Identify questions asked of providers
Important limit: An inquiry shows concern and information gathering; it is not a finding of violation or a measure of harm prevalence.
measured-trend · SC-AI04
Supported use: Show that one jurisdiction legally distinguishes companion chatbots Identify disclosure, crisis and minor-protection duties Support an emerging control-and-exit responsibility surface
Important limit: One jurisdiction; legal requirements show an emerging duty surface but do not establish typical effects or global adoption.
measured-trend · SC-G01
Supported use: Distinguish loneliness from social isolation Report WHO global prevalence estimates Identify documented risk factors and intervention levels
Important limit: Global estimates combine surveys with different wording, years and country coverage; associations do not establish a single causal path.
measured-trend · SC-G02
Supported use: Support the public-health scale of loneliness and isolation Support age and income-group differences Identify social-infrastructure and policy responses considered by WHO
Important limit: Country coverage and definitions are uneven; much of the health and driver evidence is observational.
policy-intent · SC-G03
Supported use: Show that social connection became a formal global public-health policy priority Document planned WHO reporting through 2031
Important limit: A resolution proves policy attention and future reporting commitments, not intervention effectiveness.
measured-trend · SC-G04
Supported use: Show that most surveyed people retain strong ties while material minorities lack support or close friends Support OECD contact and subgroup trends Counter a universal social-collapse claim
Important limit: Mainly covers OECD members; measures and time series are not identical between countries and do not represent the world.
measured-trend · SC-G05
Supported use: Compare mental-health workforce capacity by income group Support global budget and community-care capacity constraints Separate mental-health system capacity from social-connection prevalence
Important limit: Country reporting quality varies; workforce totals do not show geographic distribution, affordability or service quality within countries.
measured-trend · SC-G06
Supported use: Report the estimated global mental-disorder population Summarize treatment, workforce and financing gaps Support the need for careful clinical handoff boundaries
Important limit: Summarizes larger reports; global and income-group medians hide national and local variation.
measured-trend · SC-G07
Supported use: Frame insecurity, inequality and institutional trust as social-cohesion pressures Document the limits of comparable global cohesion measurement
Important limit: Trust and cohesion measures have weaker global comparability than demographic indicators; the report does not establish uniform decline.
policy-intent · SC-G08
Supported use: Report low government trust among the covered countries Support the need to distinguish interpersonal, community and institutional trust
Important limit: Trust evidence covers only countries with available surveys and varies by institution, question and year.
modelled-projection · SC-G09
Supported use: Support the direction of population ageing Identify projected timing of major global age-structure changes
Important limit: Population projections depend on fertility, mortality and migration assumptions and are not observed 2031 outcomes.
measured-trend · SC-R01
Supported use: Show US public-health prioritization of loneliness and isolation Report health-risk associations with social disconnection Identify a US social-infrastructure response frame
Important limit: Many reported health relationships are observational associations; United States evidence is not globally representative.
measured-trend · SC-R02
Supported use: Report changes in US household composition Support a living-alone demographic signal while keeping it separate from loneliness
Important limit: Household composition is not a measure of loneliness, relationship quality or voluntary solitude.
measured-trend · SC-R03
Supported use: Report EU survey prevalence bands Compare reported loneliness across participating EU countries with sampling caveats
Important limit: Online 2022 survey; response and sampling differences limit exact comparisons between countries.
measured-trend · SC-R04
Supported use: Report changes in EU household composition Support older single-adult household planning without treating it as loneliness
Important limit: Household form does not measure loneliness or belonging; member-state patterns vary.
measured-trend · SC-R05
Supported use: Report household size, older-population share and floating-population scale Support demographic and transition pressure without claiming loneliness prevalence
Important limit: Demographic, household and migration indicators do not directly measure loneliness, belonging or relationship quality.
policy-intent · SC-R06
Supported use: Document China's 2030 community mutual-aid eldercare target Support policy attention to community care as a possible counterforce
Important limit: A target is not proof of implementation, equitable access, service quality or social-connection outcomes.
measured-trend · SC-R07
Supported use: Report India's older-population projection for 2031 Support state variation in ageing pressure
Important limit: 2031 figures are projections; national totals conceal large state, gender, rural and income differences.
measured-trend · SC-R08
Supported use: Support feminisation and ruralisation of ageing Report older women's living arrangements in covered states Keep living alone separate from loneliness
Important limit: Living arrangements and vulnerability vary greatly by state and gender; living alone is not a loneliness measure.
measured-trend · SC-R09
Supported use: Support the 2015-16 mental-health treatment and system-capacity baseline Report severe workforce and access constraints in surveyed states Mark the baseline as ageing rather than current
Important limit: The baseline is old and may not reflect current capacity; state variation is substantial.
measured-trend · SC-R10
Supported use: Report 2024 Japanese loneliness frequency bands Identify age patterns and reported life-event triggers Counter the assumption that loneliness is only an older-person issue
Important limit: Self-reported cross-sectional survey; the response rate and Japanese wording matter; it cannot establish causes.
modelled-projection · SC-R11
Supported use: Report projected growth in one-person and older households Support household-planning pressure without claiming loneliness
Important limit: Projection assumptions may change; one-person household counts do not measure isolation or loneliness.
policy-intent · SC-R12
Supported use: Document the first national lonely-death prevention plan Identify intended measurement, detection and social-connection support actions
Important limit: Policy intent and monitoring design do not demonstrate service reach, causal impact or completed outcomes.
measured-trend · SC-R15
Supported use: Report average Saudi household size and single-person household share Counter automatic transfer of OECD solo-household assumptions to the Gulf
Important limit: Household size and type do not reveal loneliness, autonomy, relationship quality or within-household exclusion.
measured-trend · SC-R16
Supported use: Report self-rated satisfaction with social and family relations in Abu Dhabi Provide a Gulf countercase to assumed universal decline
Important limit: Self-reported, emirate-specific results; the public summary does not expose every sampling and item-level detail.
measured-trend · SC-R17
Supported use: Report regional mental-health treatment-gap and primary-care integration signals Support capacity risk without assigning it to every Gulf state
Important limit: The WHO region extends beyond the Gulf; regional treatment-gap statements cannot be assigned to every member state.
measured-trend · SC-R18
Supported use: Report regional mental-health spending and primary-care training progress Support continued specialist and access constraints Avoid presenting capacity as static
Important limit: Regional averages hide large country differences; service inputs do not prove access, quality or outcomes.
modelled-projection · SC-R19
Supported use: Support the direction and scale of population ageing through 2030 Mark Africa as ageing while remaining comparatively youthful
Important limit: Older source and projection; Africa is highly heterogeneous and remains younger than other world regions.
measured-trend · SC-R20
Supported use: Compare older-person living arrangements across broad settings Counter universal solo-household assumptions Keep co-residence separate from supportive relationships
Important limit: Census years, household definitions and coverage differ; co-residence does not prove supportive relationships.
policy-intent · SC-R21
Supported use: Report regional ageing scale Support unequal unpaid-care pressure Frame care-system strain without claiming loneliness prevalence
Important limit: Regional aggregates conceal country and subnational variation; care pressure is not a direct loneliness measure.
measured-trend · SC-R22
Supported use: Report regional mental-health treatment-gap and spending patterns Support a mental-health capacity constraint while preserving country uncertainty
Important limit: Regional treatment-gap and spending summaries are not current service-capacity estimates for each country.
policy-intent · SC-R23
Supported use: Frame inequality, institutional mistrust, violence and unrest as cohesion pressures Keep cohesion and trust multidimensional Reject uniform country-level decline claims
Important limit: Social cohesion is multidimensional; the evidence does not prove that every country or form of trust is declining.
measured-trend · SC-WORK-01
Supported use: Report occupational exposure to generative AI Support transformation as more likely than replacement in the baseline Reject observed mass displacement as a current fact
Important limit: Measures occupational exposure, not actual displacement, worker meaning, realized productivity or 2031 employment outcomes.
measured-trend · W01
Supported use: WASH access and gendered governance burden.
Important limit: National data timeliness and quality vary
measured-trend · W02
Supported use: Dependence on mountain water and changing flow timing.
Important limit: Local basin effects differ
scenario-assumption · W03
Supported use: Water-demand/supply gaps at warming levels.
Important limit: Global results are not local allocations and remain model sensitive
measured-trend · horizon.ai-agents-compute.ac-01
Supported use: More than 100 independent experts with nominees from over 30 countries; synthesises evidence only through Dec 2025. Supports capability, autonomy, safeguard, diffusion and uncertainty baselines.
Important limit: Scenarios and trend continuations are not predictions
measured-trend · horizon.ai-agents-compute.ac-02
Supported use: Cross-benchmark and deployment baseline, including jagged performance, agents, robotics and US–China comparison.
Important limit: Aggregates heterogeneous sources; leaderboard movement is volatile
measured-trend · horizon.ai-agents-compute.ac-03
Supported use: Model production, compute capacity, data-centre and transparency baselines.
Important limit: Several environmental values are estimates rather than operator-metered data
measured-trend · horizon.ai-agents-compute.ac-04
Supported use: Investment, adoption and productivity evidence.
Important limit: Corporate surveys and private-investment data undercount state finance and do not establish causality
measured-trend · horizon.ai-agents-compute.ac-05
Supported use: Incident, transparency, governance and safety-evaluation baseline.
Important limit: Incident reporting is voluntary and incomplete
measured-trend · horizon.ai-agents-compute.ac-07
Supported use: Strongest tested agent scored 56% across 25 black-box reimplementation projects, including one 16,000-line system.
Important limit: Precise behavioural specifications, high inference budgets and software-only tasks limit generalisation
measured-trend · horizon.ai-agents-compute.ac-08
Supported use: Best tested agent completed 24% of tasks in a simulated software company; supports a realistic-work gap.
Important limit: Model generation is now dated, but the environment exposes coordination and long-horizon failures
measured-trend · horizon.ai-agents-compute.ac-10
Supported use: Early tools made 16 experienced developers 19% slower across 246 tasks; later data weakly suggests speed-up but has severe selection effects.
Important limit: Narrow population, fast model drift
measured-trend · horizon.ai-agents-compute.ac-11
Supported use: AI acts as an organisational amplifier; higher adoption is associated with throughput and instability.
Important limit: Observational, self-reported and not proof of causality
measured-trend · horizon.ai-agents-compute.ac-12
Supported use: Two years of testing over 30 frontier systems: autonomy, cyber, bio, safeguards, self-replication precursors and scaffold effects.
Important limit: Some task detail is withheld; report explicitly is not a forecast
measured-trend · horizon.ai-agents-compute.ac-13
Supported use: Multi-agent hypothesis generation with limited expert comparison and several experimental validations.
Important limit: Small samples, vendor-developed system, substantial expert-in-the-loop involvement
measured-trend · horizon.ai-agents-compute.ac-14
Supported use: Verified algorithm optimisation deployed in Google infrastructure and applied across domains; evidence of a narrow AI-R&D feedback loop.
Important limit: Best suited to objectively scoreable problems, not open science generally
measured-trend · horizon.ai-agents-compute.ac-15
Supported use: Leading systems achieved only 9.39% accuracy on deep literature discovery and 9.31% IoU on wide research; supports an open-ended research bottleneck.
Important limit: Preprint and benchmark-specific
measured-trend · horizon.ai-agents-compute.ac-16
Supported use: Demonstrates a system generating ideas, code, experiments, analysis, manuscripts and reviews in bounded computational research.
Important limit: Output validity, novelty, review quality and responsible operation remain limiting
modelled-projection · horizon.ai-agents-compute.ac-17
Supported use: Projects global data-centre electricity demand rising from about 415 TWh to about 945 TWh in 2030, with AI the main growth driver.
Important limit: Scenario-sensitive energy model, not a capacity commitment
modelled-projection · horizon.ai-agents-compute.ac-18
Supported use: Updates 2030 central estimate to about 950 TWh from 485 TWh in 2025 and stresses grid, chip, financing and project-pipeline bottlenecks.
Important limit: Central case is not a ceiling or floor
modelled-projection · horizon.ai-agents-compute.ac-19
Supported use: Bottom-up estimate that US data centres could use 11.8% of electricity in 2030, with a 9.5–15.3% scenario range.
Important limit: Dependent on shipment and facility assumptions
measured-trend · horizon.ai-agents-compute.ac-20
Supported use: Documents geographic concentration, specialised inputs and trade dependencies.
Important limit: Mostly structural data; capacity changes rapidly
weak-signal · horizon.ai-agents-compute.ac-22
Supported use: Observed leading-edge and advanced-packaging investment and stated A14 volume-production plan for 2028.
Important limit: Company demand outlook and roadmap are management expectations, not assured supply
policy-intent · horizon.ai-agents-compute.ac-23
Supported use: Observed US work on identity, authorisation, interoperability and agent security.
Important limit: Voluntary initiative; deliverables and adoption were incomplete at cutoff
policy-intent · horizon.ai-agents-compute.ac-24
Supported use: Taxonomy and experiments establish indirect prompt injection/agent hijacking as a live failure mode.
Important limit: Does not provide a universal mitigation or population-level incidence rate
measured-trend · horizon.ai-agents-compute.ac-26
Supported use: Observed move to case-by-case licensing for specified advanced chips under conditions.
Important limit: Rules can change and do not reveal delivered volume or effective compute
policy-intent · horizon.ai-agents-compute.ac-28
Supported use: Policy intent to coordinate compute with clean energy, storage, grids, data and domestic hardware by 2030.
Important limit: Targets and pilots do not prove operational capacity
policy-intent · horizon.ai-agents-compute.ac-29
Supported use: Policy intent for infrastructure, open cooperation, capacity building, safety and sovereignty.
Important limit: Diplomatic position, not a binding regime
policy-intent · horizon.ai-agents-compute.ac-31
Supported use: Observed documentation, downstream information, copyright, training-summary and systemic-risk duties; enforcement was scheduled for 2026-08-02, one day beyond cutoff
Important limit: unknown
measured-trend · horizon.ai-agents-compute.ac-32
Supported use: Reports 19 operational AI Factories and 13 antennas; gigafactory interest is not built capacity
Important limit: unknown
policy-intent · horizon.ai-agents-compute.ac-34
Supported use: Policy path combining hardware/AI industrial support with interpretation of existing civil law, including agent cases.
Important limit: Guidance is not adjudicated precedent
policy-intent · horizon.ai-agents-compute.ac-35
Supported use: Observed innovation-plus-trust framework with transparency, safety, operator duties, voluntary verification and impact assessments.
Important limit: Implementation quality is not established by enactment
policy-intent · horizon.ai-agents-compute.ac-37
Supported use: Policy intent for R$23bn, national supercomputing, public-service use and inclusion.
Important limit: Budget envelope mixes public/private sources; execution not established
measured-trend · horizon.ai-agents-compute.ac-40
Supported use: Estimates one in four jobs potentially exposed, with transformation more likely than full replacement.
Important limit: Task exposure is not realised automation, job loss or wage impact
measured-trend · horizon.ai-agents-compute.ac-41
Supported use: Evidence on new-skill demand, youth/polarisation risks and adjustment mechanisms.
Important limit: Staff views and historical analogy do not forecast net employment
weak-signal · horizon.ai-agents-compute.ac-42
Supported use: Observed delegation, task mix, success and user expectations within one vendor's products.
Important limit: Not representative of all workers, firms, countries or models; classifiers are model-generated
measured-trend · horizon.ai-agents-compute.ac-43
Supported use: Observed universal forum on safety, AI divides, interoperability, rights and oversight.
Important limit: Dialogue does not itself create enforceable rules or compute redistribution
measured-trend · horizon.ai-agents-compute.ac-44
Supported use: More than 50 organisations pledged reporting; signal of interoperable risk disclosure.
Important limit: Voluntary participation, self-reporting and uneven verification limit assurance
scenario-assumption · horizon.economy-work-finance.ewf-01
Supported use: Potential growth has slowed; an illustrative AI scenario can lift productivity, but economies lacking physical, digital and institutional foundations risk being left behind.
Important limit: The AI result is an illustrative model, not a forecast of realised adoption or distribution.
modelled-projection · horizon.economy-work-finance.ewf-03
Supported use: Sovereign and corporate bond markets totalled $109tn; governments and firms were expected to borrow $29tn in 2026; shorter maturities raise refinancing exposure.
Important limit: Bond-market lens is not total global debt and weights market-financed economies.
measured-trend · horizon.economy-work-finance.ewf-06
Supported use: AI assistance raised issues resolved per hour by 14% on average and 34% for novice/low-skilled workers in one deployment.
Important limit: One firm and workflow; does not establish economy-wide gains or durable wage/employment effects.
measured-trend · horizon.economy-work-finance.ewf-08
Supported use: The inference cost of a model performing at GPT-3.5 level fell more than 280-fold from 2022-11 to 2024-10; hardware cost and energy efficiency also improved.
Important limit: Benchmark-equivalent inference price is not the cost of reliable end-to-end work; proprietary price series are among the inputs.
modelled-projection · horizon.economy-work-finance.ewf-09
Supported use: Measured 50%-reliability task horizon doubled about every seven months historically; the authors explicitly say current agents still fail on many substantive real projects.
Important limit: Extrapolation has major external-validity and trend-break risk and is concentrated on digital tasks.
observed-deployment · horizon.economy-work-finance.ewf-10
Supported use: Provides high-frequency applications and projected employer formations; the 2026 methodology revision excludes internet-sales applications from some high-propensity series.
Important limit: Applications are not firms, survival, jobs or AI-enabled formation; definition changes break naive comparisons.
policy-intent · horizon.economy-work-finance.ewf-11
Supported use: Digital registration, identity verification and interoperable public services can reduce formal-entry friction; B-READY measures regulation, public services and operational efficiency separately.
Important limit: Interim rollout; measures a standardised company profile, not informal or agent-founded entities.
modelled-projection · horizon.economy-work-finance.ewf-12
Supported use: Around 40% of employment is classed as AI-exposed in its framework; strong complementarity with high-income work and higher capital returns can widen income and wealth inequality.
Important limit: Staff note, not IMF policy; static occupations and scenario assumptions limit longer-horizon interpretation.
policy-intent · horizon.economy-work-finance.ewf-13
Supported use: Social spending, transition support and capital-income taxation matter; a world of radically reduced labour demand would require transfer infrastructure, while a crude automation tax could suppress productive adoption.
Important limit: Policy analysis, not evidence that radical labour reduction will occur.
measured-trend · horizon.economy-work-finance.ewf-16
Supported use: Global account ownership reached 79%; Sub-Saharan Africa reached 58%, with the world’s highest mobile-money use; digital access, safety and gender/income gaps remain material.
Important limit: Account ownership does not prove regular use, resilience, low fees or autonomous-payment readiness.
measured-trend · horizon.economy-work-finance.ewf-17
Supported use: Informality was 46.7% in 2025 H1; youth informality was 56%, and youth unemployment nearly three times the adult rate.
Important limit: Regional average masks national differences; informal work is measured imperfectly.
measured-trend · horizon.economy-work-finance.ewf-19
Supported use: Estimates 26% of Indian workers in high-AI-exposure occupations: 14% more complementary and 12% more displacement-prone; 43% of the labour force remained in agriculture.
Important limit: Model-based exposure, not observed losses or gains.
measured-trend · horizon.economy-work-finance.ewf-21
Supported use: AI compute has high concentration and entry barriers across chips, fabrication, memory and cloud, plus vertical integration and state intervention.
Important limit: Fast-moving market; several concentration measures rely on third-party estimates.
measured-trend · horizon.economy-work-finance.ewf-23
Supported use: 91% explored retail or wholesale CBDC; wholesale work was more advanced; only three live retail CBDCs existed at end-2024; designs diverged on access, limits, interest, DLT and programmability.
Important limit: Central-bank exploration is not issuance or adoption; survey reports intentions as of end-2024.
measured-trend · horizon.economy-work-finance.ewf-24
Supported use: Stablecoins demonstrate programmable payments but raise singleness, elasticity, integrity, reserve, dollarisation and fragmentation concerns; BIS favours tokenisation anchored in central-bank money.
Important limit: Institutional viewpoint and preferred architecture; competing designs may evolve.
weak-signal · horizon.economy-work-finance.ewf-25
Supported use: Prototype combined tokenised central-bank reserves and commercial-bank deposits for atomic multi-currency settlement and embedded compliance/conditional logic.
Important limit: Prototype success is not legal production readiness, cost advantage at scale or adoption.
measured-trend · horizon.economy-work-finance.ewf-26
Supported use: Tokenised government-bond issuance was about $8bn, with lower observed bid-ask spreads and comparable issuance costs in the sample.
Important limit: Tiny, selected, early market; liquidity results may reflect issuer/platform selection.
measured-trend · horizon.economy-work-finance.ewf-27
Supported use: Adoption remained nascent; interoperability, credible settlement assets, legal certainty, market integrity and investor protection constrained scale.
Important limit: Mostly early use cases and stakeholder evidence; does not predict the winning infrastructure.
policy-intent · horizon.economy-work-finance.ewf-28
Supported use: At 2025-08, implementation had significant gaps and inconsistencies; global-stablecoin regulation lagged broader crypto rules.
Important limit: Jurisdictions can change quickly after review; alignment does not ensure enforcement.
measured-trend · horizon.economy-work-finance.ewf-29
Supported use: Gross cross-border flows in the two largest dollar stablecoins rose from an estimated $12bn in 2020 Q1 to $316bn in 2025 Q1; EM use correlates with remittance, trade and macro-instability channels.
Important limit: On-chain geographic attribution is estimated; gross flow is not net adoption or payment volume.
policy-intent · horizon.economy-work-finance.ewf-30
Supported use: More than 250 stablecoins and over $300bn market capitalisation existed by mid-2025; FATF documents illicit use and enforcement gaps involving P2P/unhosted and cross-chain activity.
Important limit: The headline illicit-volume estimate is supplied by a blockchain analytics firm and does not measure the illicit share of all stablecoin activity.
policy-intent · horizon.economy-work-finance.ewf-31
Supported use: Creates a payment-stablecoin issuer regime with one-to-one permitted reserves, disclosures, redemption and AML obligations.
Important limit: Implementation depends on rules and supervision; a framework does not prove mass payment use or zero run risk.
policy-intent · horizon.economy-work-finance.ewf-32
Supported use: MiCA stablecoin provisions applied from 2024-06 and the framework fully from 2024-12, with issuer/service-provider rules and technical standards.
Important limit: Harmonised regulation does not imply global interoperability or adoption.
policy-intent · horizon.economy-work-finance.ewf-33
Supported use: If legislation is adopted in 2026, a 12-month pilot is planned from 2027 H2 and the Eurosystem aims to be ready for possible issuance in 2029; issuance remains a later decision.
Important limit: Explicitly conditional intent; readiness is not issuance, demand or displacement of private payments.
policy-intent · horizon.economy-work-finance.ewf-34
Supported use: Retail CBDC pilots added offline and programmable uses, including a state-benefit channel used by about 88,000 beneficiaries; cross-border pilots were being explored.
Important limit: Pilot count and use cases do not demonstrate repeat demand, cost advantage, privacy acceptance or national rollout.
policy-intent · horizon.economy-work-finance.ewf-35
Supported use: China intends to expand CIPS, faster-payment/QR interoperability and exploration of e-CNY in cross-border payments under macroprudential controls.
Important limit: No inference of global e-CNY adoption is warranted.
observed-deployment · horizon.economy-work-finance.ewf-36
Supported use: Digital Dirham is legal tender; a wallet was developed and the first live government transaction occurred in 2025; open finance also went live.
Important limit: One government transaction is not retail adoption or cross-border scale.
policy-intent · horizon.economy-work-finance.ewf-37
Supported use: Saudi Arabia joined the multi-wholesale-CBDC mBridge MVP to study cross-border bank settlement.
Important limit: Participation and MVP status do not prove deployment.
measured-trend · horizon.economy-work-finance.ewf-38
Supported use: Tests programmable transactions involving receivables, government bonds, trade finance, FX, real assets and public-network connections.
Important limit: Simulated/pilot services, not a launched retail currency or validated business benefit.
observed-deployment · horizon.economy-work-finance.ewf-39
Supported use: Pix transaction volume grew 52% in 2024 and reached 47% of non-cash transactions in Q4, showing that public instant-payment rails can scale without tokenisation.
Important limit: Does not include cash and does not establish transferability to other institutional contexts.
policy-intent · horizon.economy-work-finance.ewf-40
Supported use: Nearly half of African Union member states had instant-payment systems, but non-bank licensing, compliance cost and interoperability constrained inclusion and scale.
Important limit: Continental overview; maturity and use vary sharply by country.
policy-intent · horizon.economy-work-finance.ewf-41
Supported use: Provides technology-neutral recognition of automation, AI, code and dynamic information in contract formation/performance, including attribution and unexpected outputs.
Important limit: Model law requires national enactment; it does not resolve AI ethics, liability or every consumer/employment issue.
policy-intent · horizon.economy-work-finance.ewf-42
Supported use: Smart contracts used for data-sharing agreements must support robustness, access control, safe termination/interruption, auditability and consistency with legal terms.
Important limit: Scope is data-sharing agreements, not all smart contracts or agent transactions.
policy-intent · horizon.economy-work-finance.ewf-43
Supported use: Confirms that a thing may be personal property even if outside the traditional things-in-possession/action categories, reducing uncertainty for some digital assets.
Important limit: Property status alone does not settle title transfer, custody, insolvency, tax or cross-border conflicts.
policy-intent · horizon.economy-work-finance.ewf-44
Supported use: Tokenising a bank-deposit claim does not by itself change its nature; identified cases were very few, with possible automation benefits and consumer, operational and AML risks.
Important limit: Early and EU-specific; does not prove that deposits will beat stablecoins or CBDCs.
measured-trend · horizon.economy-work-finance.ewf-45
Supported use: AP2 uses typed, signed intent/payment mandates, guardrails and receipts to bind human or organisational authority to agent purchases.
Important limit: Specification and samples prove implementability, not neutral governance, interoperability, merchant acceptance or adoption.
policy-intent · horizon.economy-work-finance.ewf-46
Supported use: Defines signed agent identity/commerce intent so merchants can distinguish authorised agents from malicious automation over existing web infrastructure.
Important limit: Corporate specification with payment-scheme onboarding; claimed traffic/benefits are not independent adoption evidence.
policy-intent · horizon.economy-work-finance.ewf-47
Supported use: Announces permissioned, high-frequency, low-value machine payments across a network, with 30+ named early supporters.
Important limit: Announcement and support list do not establish production volume, economics, liability outcomes or standards convergence.
weak-signal · horizon.economy-work-finance.ewf-48
Supported use: Implements accountless pay-before-response flows over HTTP 402 with extensible settlement schemes, including on-chain exact and batched payments.
Important limit: Repository activity is a weak signal, not evidence of safe scale, consumer suitability or a durable standard.
policy-intent · horizon.economy-work-finance.ewf-49
Supported use: Pillar Two is moving from rulemaking to administration; CARF/GIR machine-readable exchange formats make cross-border tax reporting more executable.
Important limit: Participation, domestic law, US treatment and enforcement remain uneven; schemas do not eliminate avoidance.
policy-intent · horizon.economy-work-finance.ewf-50
Supported use: Establishes a parallel multilateral negotiation path, including cross-border services in an increasingly digitalised economy.
Important limit: Negotiation mandate is not agreed convention text or implementation; geopolitical conflict can delay or dilute it.
modelled-projection · horizon.economy-work-finance.ewf-51
Supported use: Decentralised-consensus costs can generate chain proliferation and fragmentation that weakens money’s network effects.
Important limit: Model result and author view, not an observed universal outcome or necessarily the view of the BIS; architecture, layer design and regulation can alter outcomes.
measured-trend · horizon.education-social-culture.s01
Supported use: 272 million children and youth were estimated out of school in 2023; countries were already projected to be 75 million off their own 2025 benchmarks.
Important limit: Modelled estimates were revised upward as data improved, so apparent change is not all deterioration
measured-trend · horizon.education-social-culture.s02
Supported use: UNESCO estimates 44 million additional primary and secondary teachers are needed by 2030.
Important limit: Staffing need is not a forecast of actual recruitment or teacher quality
measured-trend · horizon.education-social-culture.s03
Supported use: 19% reported a formal AI policy and 42% guidance in development; some were redesigning assessment.
Important limit: UNESCO Chairs/UNITWIN respondents are a self-selected institutional sample, not a global prevalence estimate
measured-trend · horizon.education-social-culture.s04
Supported use: synthesis finds generative AI can support learning when governed by clear pedagogy, while unguided task completion can weaken learning.
Important limit: Member-country evidence is not globally representative
measured-trend · horizon.education-social-culture.s05
Supported use: a six-week randomized secondary-school intervention reported a 0.31 SD improvement on an assessment spanning English and AI knowledge.
Important limit: One short, facilitated programme does not establish scaled, independent or durable effects
measured-trend · horizon.education-social-culture.s06
Supported use: randomized crossover study in a Harvard undergraduate physics course found larger immediate learning gains in less time with a carefully scaffolded tutor.
Important limit: Single institution, subject, short duration and high design support limit transfer
measured-trend · horizon.education-social-culture.s07
Supported use: a privacy-aware, cryptographically verifiable web credential standard reached Recommendation status.
Important limit: Standardization proves technical maturity, not issuer, employer or government adoption
measured-trend · horizon.education-social-culture.s08
Supported use: the treaty entered into force in 2023 and had 41 States Parties by May 2026; it covers non-traditional learning and procedures for displaced learners.
Important limit: Ratification does not prove consistent recognition practice
measured-trend · horizon.education-social-culture.s09
Supported use: the methodology can assess qualifications when documents are missing; 183 passports had been awarded.
Important limit: This is a meaningful but very small implementation base and the passport is not automatic formal recognition
measured-trend · horizon.education-social-culture.s10
Supported use: 20% of young people were NEET in 2023, two-thirds of them women; Africa's youth labour force is projected to grow by 76 million by 2050 while other regions contract.
Important limit: Definitions and modelled estimates conceal national informality and care work
measured-trend · horizon.education-social-culture.s11
Supported use: 20.8 million children were newly internally displaced in 2023; the report models divergent 2050 childhood scenarios around demography, climate and frontier technology.
Important limit: Its long horizon is useful for direction, not a 2031 point forecast
measured-trend · horizon.education-social-culture.s12
Supported use: about six billion people were online, 2.2 billion remained offline, and 82% of 15-24-year-olds used the internet.
Important limit: Connectivity does not imply affordability, safety, skill or meaningful participation
measured-trend · horizon.education-social-culture.s13
Supported use: online play, video and social participation can build digital skills; particular harmful experiences are associated with poor mental-health outcomes.
Important limit: It rejects a simple screen-time-only causal account and notes major cross-country data gaps
measured-trend · horizon.education-social-culture.s14
Supported use: 114 systems, 58%, had a national school-phone ban by March 2026, up from 24% in June 2023.
Important limit: Policy count does not measure enforcement, learning effects or displacement of use outside school
measured-trend · horizon.education-social-culture.s15
Supported use: under-16 account restrictions took effect in December 2025; the regulator reported that platforms had removed access to about 4.7 million accounts and began a two-year impact evaluation.
Important limit: Restricted access is implementation evidence, not proof of well-being or safety outcomes
measured-trend · horizon.education-social-culture.s16
Supported use: rules in force from March 2026 impose child-risk mitigation, age-assurance, parental-supervision, advertising and games obligations.
Important limit: Outcomes and enforcement consistency are not yet established
measured-trend · horizon.education-social-culture.s17
Supported use: the Commission presented minor-protection guidance and a privacy-preserving over-18 proof prototype tied to forthcoming identity wallets.
Important limit: This establishes policy direction and architecture, not adoption or exclusion rates
measured-trend · horizon.education-social-culture.s18
Supported use: estimated loneliness prevalence in 2014-2023 was 16% globally, 20.9% for ages 13-17 and 24.3% in low-income countries; WHO proposed a global Social Connection Index.
Important limit: Definitions, sparse country coverage and observational studies limit causal attribution
measured-trend · horizon.education-social-culture.s19
Supported use: repeated interactions reduced loneliness immediately after use for a week, but effects did not persist between sessions; authors explicitly do not support replacing human relationships or care.
Important limit: Long-term dependency and substitution remain unknown
measured-trend · horizon.education-social-culture.s20
Supported use: four-week preregistered RCT found no simple modality/task effect on loneliness; heavier voluntary use correlated with worse loneliness, socialization, dependence and problematic use.
Important limit: It is a preprint, the duration is short, and usage intensity was not randomly assigned
measured-trend · horizon.education-social-culture.s21
Supported use: AI-written deep-talk responses produced strong perceived closeness when labelled human; AI labelling reduced but did not eliminate it.
Important limit: Scripted short interactions do not establish real relationship formation or long-term outcomes
measured-trend · horizon.education-social-culture.s22
Supported use: online groups can provide companionship, information and support, while homogeneous interests and harmful norms can amplify risk.
Important limit: A conceptual review cannot quantify a single net effect across community types
measured-trend · horizon.education-social-culture.s23
Supported use: cross-national Global Flourishing Study analysis links childhood experiences with adult belonging.
Important limit: Coverage is broad but not universal, and retrospective observational associations are not causal
measured-trend · horizon.education-social-culture.s24
Supported use: trust in news fell to 37%, 42% sometimes or often avoided news, 27% got news from news-focused creators and 46% from creators of any kind.
Important limit: Online samples in lower-connectivity markets skew more educated and cannot be read as whole-population global estimates
measured-trend · horizon.education-social-culture.s25
Supported use: 62% did not systematically fact-check before sharing and 73% wanted training.
Important limit: Small, non-probability sample and self-report limit prevalence claims
measured-trend · horizon.education-social-culture.s26
Supported use: fewer than 30% report that most people can be trusted; more than half report little or no trust in government; later birth cohorts show lower institutional trust.
Important limit: Survey availability, question wording and regime context vary substantially
measured-trend · horizon.education-social-culture.s27
Supported use: maps provenance, watermarking, detection, labelling and testing approaches.
Important limit: It does not show that any one technique determines truth or works universally against adaptive attacks
measured-trend · horizon.education-social-culture.s28
Supported use: binding rules effective 2025-09-01 require explicit notices and implicit metadata across text, image, audio, video and virtual scenes, with duties for distribution services.
Important limit: enforcement outcomes are not supplied.
measured-trend · horizon.education-social-culture.s29
Supported use: law requires disclosure of deepfakes and certain public-interest AI text; the voluntary implementation code was published ahead of obligations applying from 2026-08-02.
Important limit: At this scan's cutoff the main application date was one day away, so impact is unknown
measured-trend · horizon.education-social-culture.s30
Supported use: interoperable signed provenance manifests reached version 2.2.
Important limit: A provenance record describes origin and edits when present; it does not prove the semantic truth of a claim, and ecosystem coverage is not universal
measured-trend · horizon.education-social-culture.s31
Supported use: AI labels changed perceptions of whether media was generated, but simple AI-origin labels had little effect on stated engagement.
Important limit: Experimental posts and stated intentions do not establish field behaviour
measured-trend · horizon.education-social-culture.s32
Supported use: identifies risks to cultural diversity, discoverability, creator remuneration, cultural data sovereignty and intellectual-property protection from digital concentration and AI.
Important limit: Member-State reporting is uneven and partly self-reported
measured-trend · horizon.education-social-culture.s33
Supported use: more than half of languages are described as endangered; UNESCO is monitoring state implementation of its multilingualism recommendation.
Important limit: Language vitality estimates and digital availability are incomplete for many communities
measured-trend · horizon.education-social-culture.s34
Supported use: affiliation changed through both demography and religious switching, with strong regional variation.
Important limit: The observation window ends in 2020; affiliation is only one dimension of identity, practice and meaning
measured-trend · horizon.education-social-culture.s35
Supported use: global fertility was 2.25 births per woman in 2024; the population is projected to keep ageing and to peak around 10.3 billion in the mid-2080s, with sharply divergent national trajectories.
Important limit: Medium-variant demographic assumptions are not destiny
measured-trend · horizon.education-social-culture.s36
Supported use: 39% said financial limits had affected or would affect desired family size; one in five expected not to realize their fertility intentions.
Important limit: The survey is diverse but not a census of the world, and intentions can change
measured-trend · horizon.education-social-culture.s37
Supported use: at end-2025 there were 41.6 million refugees and 68.7 million people internally displaced by conflict or violence; 39% of refugees were children.
Important limit: Return counts include adverse or involuntary conditions and do not imply durable reintegration
measured-trend · horizon.education-social-culture.s38
Supported use: up to 216 million internal climate migrants by 2050 in a high-impact pathway, with hotspots possible by 2030; strong mitigation and resilient development could reduce the figure by as much as 80%.
Important limit: It is a scenario model, not a count or prediction
measured-trend · horizon.education-social-culture.s39
Supported use: the latest underlying UN estimate was 281 million international migrants, 3.6% of world population, while most migration remained safe, regular and development-linked.
Important limit: The stock estimate is lagged and must not be conflated with forced displacement
measured-trend · horizon.education-social-culture.s40
Supported use: about 800 million people lacked official ID and at least 2.8 billion lacked access to government-recognized digital ID for online transactions.
Important limit: Availability does not equal ownership, use, consent, inclusion or trust
measured-trend · horizon.energy-climate-food-water.c02
Supported use: The headline five-year temperature probabilities cover 2026-30; regional predictive skill is uneven and the report does not predict individual disasters
Important limit: unknown
measured-trend · horizon.energy-climate-food-water.c06
Supported use: 36 material climate risks and urgency.
Important limit: Findings are European, not globally transferable
policy-intent · horizon.energy-climate-food-water.c07
Supported use: Eight core heat-health capabilities.
Important limit: Publication of guidance is not evidence of local implementation
scenario-assumption · horizon.energy-climate-food-water.c08
Supported use: Cooling demand, refrigerants and passive/efficient-cooling pathway.
Important limit: BAU and sustainable-cooling totals are scenarios, not forecasts
measured-trend · horizon.energy-climate-food-water.c09
Supported use: Insured natural-catastrophe losses and secondary perils.
Important limit: Proprietary methods; losses reflect exposure, inflation and insurance penetration as well as hazard
modelled-projection · horizon.energy-climate-food-water.e01
Supported use: Demand, generation, emissions, grid investment and regional outlook.
Important limit: Forecasts are conditional, not commitments
modelled-projection · horizon.energy-climate-food-water.e02
Supported use: Updates 2026-27 demand and generation after fuel-price and weather shocks.
Important limit: Two-year volatility should not be extrapolated mechanically to 2031
measured-trend · horizon.energy-climate-food-water.e04
Supported use: New-project levelised costs.
Important limit: LCOE comparisons do not by themselves price networks, curtailment, balancing or financing differences
modelled-projection · horizon.energy-climate-food-water.e04a
Supported use: Solar/wind/storage configurations and cost ranges.
Important limit: Results are modelled for favourable locations and should not be generalized to every grid
modelled-projection · horizon.energy-climate-food-water.e06
Supported use: Demand response, battery additions and costs.
Important limit: Global aggregates conceal market-design and local-network differences
scenario-assumption · horizon.energy-climate-food-water.e07
Supported use: Storage needs and battery cost/supply-chain pathways.
Important limit: The 1,500 GW storage figure is a net-zero scenario requirement, not a forecast
scenario-assumption · horizon.energy-climate-food-water.e09
Supported use: Updates data-centre growth and infrastructure bottlenecks.
Important limit: Corporate investment plans are not equivalent to completed load
modelled-projection · horizon.energy-climate-food-water.e10
Supported use: Reactor fleet, construction concentration, life extensions and small-modular-reactor plans.
Important limit: Cost, schedule and financing risks remain large
modelled-projection · horizon.energy-climate-food-water.e11
Supported use: Nuclear-capacity range.
Important limit: High case is explicitly a scenario rather than a delivery forecast
modelled-projection · horizon.energy-climate-food-water.e12
Supported use: Demand, low-emissions production, offtake and final-investment status.
Important limit: Announced-project volume greatly exceeds committed capacity
policy-intent · horizon.energy-climate-food-water.e14
Supported use: Current traceability coverage and data barriers.
Important limit: Surveyed firms may not represent the whole supply chain
measured-trend · horizon.energy-climate-food-water.e15
Supported use: Capacity additions, cumulative capacity and generation share.
Important limit: Capacity growth is not evidence that congestion and thermal reliance have disappeared
policy-intent · horizon.energy-climate-food-water.e16
Supported use: Planning assumptions for demand, capacity, storage and transmission.
Important limit: Treat as policy/planning intent until delivery is observed
measured-trend · horizon.energy-climate-food-water.e17
Supported use: Demand growth, generation and data-centre effects under existing laws and assumptions.
Important limit: Not a prediction of future federal or state policy
scenario-assumption · horizon.energy-climate-food-water.e19
Supported use: Hydro-rich power mix, resource potential and transition pathways.
Important limit: Older than other regional sources; live signposts need refreshing
scenario-assumption · horizon.energy-climate-food-water.e20
Supported use: Cooling, desalination, peak load and supply pathways.
Important limit: Regional aggregate masks energy-exporter and importer differences
measured-trend · horizon.energy-climate-food-water.f01
Supported use: Hunger and healthy-diet affordability.
Important limit: Estimates are modelled where surveys are incomplete
modelled-projection · horizon.energy-climate-food-water.f02
Supported use: Production, demand, trade, emissions and income volatility.
Important limit: Baseline assumes broadly normal weather and continuation of stated policy
policy-intent · horizon.energy-climate-food-water.k01
Supported use: Coverage, revenue and credit prices.
Important limit: Coverage does not establish an effective or harmonized marginal carbon price
measured-trend · horizon.energy-climate-food-water.k02
Supported use: Rules for cooperative approaches and the Paris Agreement crediting mechanism.
Important limit: Agreed architecture is not evidence of liquid, high-integrity trade
scenario-assumption · horizon.energy-climate-food-water.k03
Supported use: Current conventional and novel removal, scenarios and policy.
Important limit: Taxonomy and future need vary across models
scenario-assumption · horizon.energy-climate-food-water.k04
Supported use: Separates current removal from Paris-consistent pathways and near-term commitments.
Important limit: Scenario spread is substantial
scenario-assumption · horizon.energy-climate-food-water.m01
Supported use: Extraction, impacts and resource-efficiency pathways.
Important limit: Forward totals depend on economic and policy assumptions
modelled-projection · horizon.energy-climate-food-water.m02
Supported use: Building emissions, material demand and efficiency investment.
Important limit: Required investment is pathway analysis, not pledged capital
measured-trend · horizon.energy-climate-food-water.w04
Supported use: Atmospheric evaporative demand's contribution to observed drought severity.
Important limit: Drought metric and period matter
measured-trend · horizon.energy-climate-food-water.w05
Supported use: Large-scale groundwater recovery can follow abstraction control, transferred water and recharge.
Important limit: It is not automatically transferable
modelled-projection · horizon.energy-climate-food-water.w06
Supported use: Projects weaker warm-season agricultural-drought predictability at higher warming levels.
Important limit: This is a future-model result, not an observed 2031 outcome
measured-trend · horizon.geopolitics-governance-security.s01
Supported use: Research paper; global trade-bloc analysis using monthly goods data.
Important limit: Authors' research, not an official WTO position; sample ends February 2024 in the later update described by the WTO.
modelled-projection · horizon.geopolitics-governance-security.s02
Supported use: Official outlook; global trade, bilateral diversion and data-policy fragmentation.
Important limit: Forecasts and model estimates are not observed outcomes.
measured-trend · horizon.geopolitics-governance-security.s03
Supported use: Official dataset/report; 129 economies, rules current to 31 October 2025.
Important limit: Measures recorded regulatory restrictions, not enforcement quality or political motivation.
policy-intent · horizon.geopolitics-governance-security.s04
Supported use: transfer exemptions, thresholds and security assessment.
Important limit: unknown
policy-intent · horizon.geopolitics-governance-security.s05
Supported use: equipment, software, HBM and entity restrictions.
Important limit: Government rationale; effectiveness is not established by the announcement.
policy-intent · horizon.geopolitics-governance-security.s06
Supported use: case-by-case review under conditions.
Important limit: A licensing-policy change, not evidence of realised exports or strategic effect.
policy-intent · horizon.geopolitics-governance-security.s07
Supported use: Primary policy action; China/global critical-material flows.
Important limit: unknown
measured-trend · horizon.geopolitics-governance-security.s08
Supported use: Research paper; global, data-driven grouping and decoupling analysis.
Important limit: Working-paper findings do not represent IMF policy; bloc definitions are analytical constructs.
measured-trend · horizon.geopolitics-governance-security.s09
Supported use: state-based, interstate and fatality counts.
Important limit: Uses UCDP definitions and preliminary/updated event coding; counts are not interchangeable with other conflict datasets.
measured-trend · horizon.geopolitics-governance-security.s10
Supported use: Independent research institute; global security, armaments and nuclear forces.
Important limit: Estimated arsenals and stockpiles; secrecy creates unavoidable uncertainty.
modelled-projection · horizon.geopolitics-governance-security.s11
Supported use: UN administrative and modelled statistics; global forced displacement.
Important limit: End-2024 total is an estimate and later provisional updates can revise it.
observed-deployment · horizon.geopolitics-governance-security.s12
Supported use: Official global dataset using 2024 Global Findex and administrative sources; identity coverage and use.
Important limit: Digital-ID definitions depend on government recognition and remote authentication; country data quality varies.
policy-intent · horizon.geopolitics-governance-security.s13
Supported use: safety, inclusion and rights principles.
Important limit: Voluntary safeguards, not evidence of adoption or compliance.
policy-intent · horizon.geopolitics-governance-security.s14
Supported use: internet, data governance, DPI and AI cooperation.
Important limit: Political commitment with nationally mediated implementation; not self-executing law.
policy-intent · horizon.geopolitics-governance-security.s15
Supported use: wallets, relying parties, selective disclosure and safeguards.
Important limit: Delivery depends on implementing acts, national wallets, certification and adoption.
policy-intent · horizon.geopolitics-governance-security.s16
Supported use: proofing, authentication and federation.
Important limit: Guidance is not universally mandatory and does not establish deployment.
measured-trend · horizon.geopolitics-governance-security.s17
Supported use: public/private service connection through Singpass and Corppass.
Important limit: First-party description; user-control and service claims are not independent impact evaluation.
measured-trend · horizon.geopolitics-governance-security.s18
Supported use: digital-government development.
Important limit: Composite indices compress institutional and distributional differences.
measured-trend · horizon.geopolitics-governance-security.s19
Supported use: Official benchmark; 197 economies, 48 indicators.
Important limit: Measures maturity proxies and reported capabilities, not service quality experienced by every group.
measured-trend · horizon.geopolitics-governance-security.s20
Supported use: trust, debt, capability and service pressures.
Important limit: OECD-heavy coverage; cannot stand in for lower-income or fragile states.
measured-trend · horizon.geopolitics-governance-security.s21
Supported use: standards, trade and implementation capability.
Important limit: Its estimates aggregate unlike sectors; standards can enable or exclude depending on capability.
policy-intent · horizon.geopolitics-governance-security.s22
Supported use: coordinated, Africa-centred and responsible AI.
Important limit: Continental strategy; national implementation and finance will vary.
policy-intent · horizon.geopolitics-governance-security.s23
Supported use: harmonised governance, data flows and rights.
Important limit: Predates the cut-off by four years but remains the relevant continental baseline; implementation is uneven.
policy-intent · horizon.geopolitics-governance-security.s24
Supported use: GOV.BR, national identity, consent and service delivery.
Important limit: Targets are intentions, not achieved outcomes.
policy-intent · horizon.geopolitics-governance-security.s25
Supported use: sovereign cloud, automation, AI and spending intent.
Important limit: First-party target language; no independent proof of completion or value.
policy-intent · horizon.geopolitics-governance-security.s26
Supported use: high-impact and generative AI duties.
Important limit: a grace period delays the evidential test of enforcement.
measured-trend · horizon.geopolitics-governance-security.s27
Supported use: staged application and amendments.
Important limit: Timelines have changed; verify the law and later omnibus measures before treating any duty as active.
policy-intent · horizon.geopolitics-governance-security.s28
Supported use: designation and enforcement.
Important limit: Commission account of its own programme; remedies and market effects need independent evaluation.
policy-intent · horizon.geopolitics-governance-security.s29
Supported use: service governance, content and security.
Important limit: unknown
policy-intent · horizon.geopolitics-governance-security.s30
Supported use: explicit/implicit labels, distribution and platform duties.
Important limit: effectiveness is untested here.
policy-intent · horizon.geopolitics-governance-security.s31
Supported use: innovation, infrastructure, exports and security.
Important limit: Policy intent from the administration, not enacted law or measured outcome.
policy-intent · horizon.geopolitics-governance-security.s32
Supported use: rights assessments and remedies.
Important limit: Opened for signature in 2024; entry into force depends on ratifications. Defence is outside scope and national-security treatment is qualified.
measured-trend · horizon.geopolitics-governance-security.s33
Supported use: biometric databases and surveillance safeguards.
Important limit: Foundational but older; legal standards remain relevant while technology and national law have moved.
policy-intent · horizon.geopolitics-governance-security.s34
Supported use: impact, transparency and severe-harm tests.
Important limit: Normative guidance, not evidence that states comply.
policy-intent · horizon.geopolitics-governance-security.s35
Supported use: digital borders, profiling, due process and remedy.
Important limit: Normative interpretation; domestic effect and enforcement vary by jurisdiction.
policy-intent · horizon.geopolitics-governance-security.s36
Supported use: signatures, ratifications and entry threshold.
Important limit: A signature is not ratification or implementation; status can change after the cut-off.
measured-trend · horizon.geopolitics-governance-security.s37
Supported use: compute, cloud, chips and vertical relations.
Important limit: Market shares are compiled from third-party estimates and differ by market definition and geography.
measured-trend · horizon.geopolitics-governance-security.s38
Supported use: UN trade/development analysis; global and emerging-economy competition capacity.
Important limit: Aggregated multinational-enterprise measures do not describe every digital market.
policy-intent · horizon.geopolitics-governance-security.s39
Supported use: Final standards and transition direction; US with global standards influence.
Important limit: Standard availability is not migration completion; schedules can change with cryptanalysis and implementation evidence.
policy-intent · horizon.geopolitics-governance-security.s40
Supported use: inventory by 2028, priority migration by 2031, completion by 2035.
Important limit: Target dates are guidance, not a measured national inventory or mandate for all organisations.
policy-intent · horizon.geopolitics-governance-security.s41
Supported use: synchronised migration.
Important limit: Roadmap is policy intent and coordination, not completed replacement of cryptography.
measured-trend · horizon.geopolitics-governance-security.s42
Supported use: discovery, prioritisation and migration.
Important limit: Advisory roadmap; organisation-level adoption is not measured by publication.
measured-trend · horizon.geopolitics-governance-security.s43
Supported use: Joint national cyber-authority advisories; US and allied telecom/critical infrastructure.
Important limit: Attribution and threat framing are government assessments; disclosure is selective.
measured-trend · horizon.geopolitics-governance-security.s44
Supported use: ransomware, DDoS, cloud and geopolitical drivers.
Important limit: Incident visibility and reporting vary; observed events undercount total compromise.
measured-trend · horizon.geopolitics-governance-security.s45
Supported use: legal, technical, organisational, capacity and cooperation commitments.
Important limit: Measures commitment/capability proxies rather than breach prevention or operational resilience.
measured-trend · horizon.geopolitics-governance-security.s46
Supported use: crime burden and enforcement capability.
Important limit: Combines member reporting and private-sector data; recording rates and definitions differ.
measured-trend · horizon.geopolitics-governance-security.s47
Supported use: International infrastructure evidence and governance initiative; global connectivity.
Important limit: Repair counts include accidents and natural causes; they must not be presented as sabotage counts.
measured-trend · horizon.geopolitics-governance-security.s48
Supported use: biometric travel records and rollout.
Important limit: Operational status does not establish accuracy, fairness, proportionality or traveller experience.
policy-intent · horizon.geopolitics-governance-security.s49
Supported use: border, asylum, Eurodac and safeguards.
Important limit: Legal application does not establish uniform national capacity or rights outcomes.
policy-intent · horizon.geopolitics-governance-security.s50
Supported use: biometrics, credentials and access.
Important limit: Programme description; benefits and harms depend on context, consent, data sharing and remedy.
policy-intent · horizon.geopolitics-governance-security.s51
Supported use: prohibition/restriction case.
Important limit: ICRC recommendation is not a treaty or evidence of state agreement.
measured-trend · horizon.geopolitics-governance-security.s52
Supported use: cascading effects across water, heat, health, education and communications.
Important limit: One conflict context; demonstrates a mechanism but not global incidence.
modelled-projection · horizon.geopolitics-governance-security.s53
Supported use: market growth, R&D concentration and inequality.
Important limit: Projection, not observed market size; firm and R&D estimates use historical underlying data.
policy-intent · horizon.geopolitics-governance-security.s54
Supported use: data use in individualised pricing.
Important limit: An information request is evidence of regulatory concern, not proof that unlawful or harmful conduct occurred.
measured-trend · horizon.geopolitics-governance-security.s55
Supported use: non-consensual intimate imagery, deepfakes and takedown duties.
Important limit: CRS identifies constitutional questions; enactment does not establish platform compliance or net rights effect.
measured-trend · horizon.health-biotech-demography.hbd-01
Supported use: Population, fertility, longevity, migration and population-peak baselines.
Important limit: Medium- and long-range demographic projections are model-dependent; 2031 is less uncertain than 2050/2100 but migration and fertility shocks still matter.
measured-trend · horizon.health-biotech-demography.hbd-02
Supported use: Fertility distribution and regional divergence.
Important limit: National averages conceal subnational, income, ethnic and parity differences. Replacement fertility is not an immediate population-growth threshold.
measured-trend · horizon.health-biotech-demography.hbd-03
Supported use: Gap between desired and realised fertility; reproductive agency.
Important limit: Survey countries cover over one-third, not all, of the global population; attitudes and realised births differ.
measured-trend · horizon.health-biotech-demography.hbd-04
Supported use: Life expectancy, mortality and health-workforce trajectory.
Important limit: Latest year varies by indicator; pandemic-era revisions affect trend comparability.
measured-trend · horizon.health-biotech-demography.hbd-05
Supported use: Service coverage and financial-hardship baseline.
Important limit: Indicators were revised; older report series are not directly comparable and national aggregates hide exclusion.
measured-trend · horizon.health-biotech-demography.hbd-06
Supported use: Global 2030 workforce shortfall and regional concentration.
Important limit: Projection depends on member-reported workforce accounts, definitions and pre-existing trend assumptions.
measured-trend · horizon.health-biotech-demography.hbd-07
Supported use: Shortage, unemployment, training, migration and fiscal-absorption paradox.
Important limit: Needs-based model has wide ranges; data quality and cadre coverage vary by country.
measured-trend · horizon.health-biotech-demography.hbd-09
Supported use: Long-term-care need, cost and affordability.
Important limit: High- and upper-income member coverage limits transferability to informal-care-dominant systems.
measured-trend · horizon.health-biotech-demography.hbd-10
Supported use: Regional long-term-care need and policy direction.
Important limit: Regional estimate does not resolve large differences in coverage, informality and fiscal capacity.
measured-trend · horizon.health-biotech-demography.hbd-14
Supported use: Expected health uses and governance risks of multimodal foundation models.
Important limit: Guidance is non-binding and explicitly notes that broad general capability is not proven.
measured-trend · horizon.health-biotech-demography.hbd-15
Supported use: Total-product-lifecycle evidence, transparency and change-management direction.
Important limit: Draft, non-binding recommendations; US device jurisdiction does not cover all clinical AI or other countries.
measured-trend · horizon.health-biotech-demography.hbd-17
Supported use: Regulatory expectations for AI from discovery through post-authorisation.
Important limit: Reflection paper captures current regulator thinking, not a complete binding rulebook.
measured-trend · horizon.health-biotech-demography.hbd-18
Supported use: Support and standards work for AI devices, medical robots, biomaterials and brain-computer interfaces.
Important limit: delivery, approvals and safety outcomes require separate evidence.
measured-trend · horizon.health-biotech-demography.hbd-19
Supported use: Sandbox testing of evolving AI, diagnostics and post-market surveillance questions.
Important limit: Sandbox reports are explicitly not formal guidance; small selected cohorts are not market-wide evidence.
measured-trend · horizon.health-biotech-demography.hbd-20
Supported use: Regulatory clearance of a less invasive amyloid-assessment aid.
Important limit: Studied symptomatic adults; it is not a population-screening test and must be interpreted with other clinical information.
measured-trend · horizon.health-biotech-demography.hbd-21
Supported use: Diagnostic-capacity and equitable-access gap in antimicrobial resistance.
Important limit: Programme intent; laboratory capability and access vary sharply within countries.
measured-trend · horizon.health-biotech-demography.hbd-22
Supported use: Priority-resistant pathogens and R&D direction.
Important limit: Prioritisation is not a forecast of local incidence; the pipeline and resistance patterns change.
measured-trend · horizon.health-biotech-demography.hbd-23
Supported use: Cardiovascular-outcome evidence for pharmacological obesity treatment.
Important limit: Manufacturer-funded; secondary prevention population with existing cardiovascular disease, not universal primary prevention.
measured-trend · horizon.health-biotech-demography.hbd-24
Supported use: Conditional global clinical recommendation plus explicit affordability, supply and system constraints.
Important limit: A guideline does not establish national reimbursement, manufacturing sufficiency, adherence or long-term population effects.
measured-trend · horizon.health-biotech-demography.hbd-25
Supported use: State of human senolytic evidence and biomarker/stratification problem.
Important limit: Perspective on small early trials; no demonstrated general anti-ageing efficacy in humans.
measured-trend · horizon.health-biotech-demography.hbd-26
Supported use: Global mandate for equitable precision-medicine readiness, governance and capacity.
Important limit: Resolution is policy intent; national financing, data representation and delivery capacity are not supplied by adoption.
measured-trend · horizon.health-biotech-demography.hbd-27
Supported use: National whole-genome reference resource exceeding 10,000 individuals.
Important limit: Government description; representation quality, research access, clinical utility and consent governance need independent assessment.
measured-trend · horizon.health-biotech-demography.hbd-28
Supported use: Genomic underrepresentation, diversity and African-led data governance.
Important limit: Project expands but does not make African genomic coverage comprehensive; many ethnolinguistic groups remain unsampled.
measured-trend · horizon.health-biotech-demography.hbd-29
Supported use: Population-scale research of newborn whole-genome screening for 200+ treatable rare conditions.
Important limit: Research study, not routine national screening; clinical utility, cost, consent, false positives and lifetime data governance remain under study.
measured-trend · horizon.health-biotech-demography.hbd-30
Supported use: Feasibility and clinical outcomes in 1,000 newborns using 605 genes.
Important limit: Single-state cohort; consent model, ancestry mix, follow-up and health-system integration constrain generalisation.
measured-trend · horizon.health-biotech-demography.hbd-31
Supported use: First approved CRISPR/Cas9 therapy.
Important limit: Approval for a defined population at specialist centres; access, conditioning burden, durability and long-term safety are separate questions.
measured-trend · horizon.health-biotech-demography.hbd-32
Supported use: First reported bespoke lipid-nanoparticle base-editing treatment for one infant.
Important limit: Single-patient report with short follow-up; it proves technical possibility, not a repeatable regulatory, manufacturing or reimbursement model.
measured-trend · horizon.health-biotech-demography.hbd-33
Supported use: Governance baseline for somatic, germline and heritable editing.
Important limit: Predates several clinical advances; implementation remains nationally heterogeneous.
measured-trend · horizon.health-biotech-demography.hbd-34
Supported use: Expanded biomolecular-structure prediction across proteins, nucleic acids, ligands and complexes.
Important limit: Benchmark accuracy is not equivalent to biological mechanism, drug efficacy, safety or reduced clinical-development time.
measured-trend · horizon.health-biotech-demography.hbd-35
Supported use: Early human efficacy/safety signal for an AI-generated target and molecule.
Important limit: Short, small trial with withdrawals; does not show improved industry-wide success rates or phase 3 benefit.
measured-trend · horizon.health-biotech-demography.hbd-36
Supported use: Transition from foundational transfer toward 2026–2030 GMP-grade, commercially sustainable regional production.
Important limit: Programme milestones and partner count do not prove licensed products, demand, procurement or viable utilisation.
measured-trend · horizon.health-biotech-demography.hbd-37
Supported use: Funding-linked screening expectations for synthetic nucleic-acid orders and benchtop synthesis.
Important limit: The source says a May 2025 executive order requires agencies to revise or replace the framework; successor requirements, compliance, international coverage and effectiveness were unresolved at cutoff.
measured-trend · horizon.health-biotech-demography.hbd-38
Supported use: Early basket-trial evidence for immune reset in refractory autoimmune disease.
Important limit: Small, specialised, non-randomised early trial; durability, comparative benefit, toxicity, capacity and cost are unresolved.
measured-trend · horizon.health-biotech-demography.hbd-39
Supported use: Sustained physiological function in a human decedent model.
Important limit: Decedent evidence is not a living-recipient clinical outcome; rejection, zoonosis, ethics and scalable husbandry remain material.
measured-trend · horizon.health-biotech-demography.hbd-40
Supported use: Near-real-time speech synthesis from cortical activity in severe paralysis.
Important limit: Single participant, invasive research system; reliability, surgery, maintenance, privacy and generalisation are unresolved.
measured-trend · horizon.health-biotech-demography.hbd-41
Supported use: Market authorisation of adaptive stimulation for specified Parkinson's disease use.
Important limit: Narrow indication and device configuration; does not establish broad closed-loop neurotechnology adoption.
measured-trend · horizon.health-biotech-demography.hbd-42
Supported use: First global normative standard addressing mental privacy, dignity, autonomy and discrimination.
Important limit: Non-binding recommendation; national law, technical definitions and enforcement may diverge.
measured-trend · horizon.health-biotech-demography.hbd-43
Supported use: Lifetime infertility prevalence and access need.
Important limit: Heterogeneous study methods and sparse data in some regions; prevalence is not treatment demand or access.
measured-trend · horizon.health-biotech-demography.hbd-44
Supported use: Professional conclusion that polygenic embryo screening is not ready for clinical use.
Important limit: US professional opinion, not law; commercial availability and rules differ internationally.
measured-trend · horizon.health-biotech-demography.hbd-45
Supported use: Updated oversight for stem-cell-based embryo models.
Important limit: Professional self-governance; legal embryo definitions and research limits remain national.
measured-trend · horizon.health-biotech-demography.hbd-46
Supported use: Scientific, clinical and governance hurdles for laboratory-derived gametes.
Important limit: Perspective; explicitly states human reproductive use is not yet feasible.
measured-trend · horizon.health-biotech-demography.hbd-48
Supported use: Cutoff-current status of the 2025-adopted Pandemic Agreement and its unfinished Pathogen Access and Benefit Sharing annex.
Important limit: At cutoff the agreement is not yet open for signature/ratification until the annex is adopted; implementation and compliance are unknown.
measured-trend · horizon.health-biotech-demography.hbd-49
Supported use: New pandemic-emergency alert and stronger equity/coordination provisions.
Important limit: Eleven States Parties rejected the 2024 amendments; WHO cannot compel national action.
measured-trend · horizon.health-biotech-demography.hbd-50
Supported use: Rapid-vaccine mission and risk of reproducing access inequity.
Important limit: Review, not proof that a safe, authorised and equitably deployed vaccine can be delivered in 100 days.
measured-trend · horizon.health-biotech-demography.hbd-51
Supported use: Heat, climate-sensitive disease, productivity and health-system pressure.
Important limit: Attribution and burden estimates use models with uncertainty; global indicators mask local adaptation.
measured-trend · horizon.institutional-scan-of-scans.iss-07c9f693
Supported use: current Pandemic Agreement legal/process status
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-0b701133
Supported use: adopted strategy
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-0dda4c31
Supported use: converging-technology assessment and governance
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-154538ce
Supported use: four AI scenarios
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-20f8b837
Supported use: AI development evidence and policy
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-2d077429
Supported use: estimates and probabilistic projections
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-394f06ea
Supported use: development signals
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-5722750b
Supported use: task/occupation exposure
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-58fb6a5a
Supported use: supply/demand, concentration and scenarios
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-592a4b98
Supported use: macro-financial assessment
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-5bf2f5eb
Supported use: macro forecast
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-5d935de1
Supported use: 35 plausible disruptions and stakeholder assessment
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-6ade368d
Supported use: global perception survey and foresight
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-6f7279a0
Supported use: platform-usage distribution and labour-cost-equivalent estimate
Important limit: working-paper author views.
measured-trend · horizon.institutional-scan-of-scans.iss-7371bf3f
Supported use: whole-of-government/parliament foresight
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-73c84b44
Supported use: near-term climate observations and forecast
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-75051685
Supported use: electricity forecast
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-78623d73
Supported use: patent publications through 2025
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-7f7b8f0e
Supported use: signals/scenarios to 2035
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-8a1c1289
Supported use: One Health policy intent and pathogen evidence
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-8b16b5e6
Supported use: patent/science evidence plus foresight
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-928bc9ca
Supported use: macro forecast and data
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-a2dfd6cf
Supported use: data-centre energy modelling
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-a33b5a27
Supported use: current-policy, stated-policy and normative energy scenarios
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-a749e931
Supported use: method
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-af654fd7
Supported use: tokenisation evidence and proposed architecture
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-b82c404b
Supported use: structural forces and five scenarios
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-c2e6d457
Supported use: EU strategic foresight
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-c526501b
Supported use: innovation inputs/outputs and clusters
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-c7882c4c
Supported use: labour exposure scenarios
Important limit: staff views, not IMF policy.
measured-trend · horizon.institutional-scan-of-scans.iss-ca53bedd
Supported use: planetary health foresight
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-cd4d54c8
Supported use: benchmark, investment, adoption, policy, incident and public-opinion index
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-ce6c2515
Supported use: cross-domain foresight and method
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-cee272ba
Supported use: AI macro/financial analysis
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-d5a0289e
Supported use: vision, Delphi, scenarios, horizon scan and technology convergence
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-dd0ce37c
Supported use: employer expectations
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-f03d3232
Supported use: connectivity estimates, gaps and policy targets
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-f73f115e
Supported use: method, drivers and wildcards
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-f952a70b
Supported use: strategic foresight and alternative pathways
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-feb5ea18
Supported use: AI macro scenarios
Important limit: unknown
measured-trend · horizon.institutional-scan-of-scans.iss-ffcacbfc
Supported use: development/AI
Important limit: unknown
measured-trend · horizon.robotics-spatial-mobility.rsm-01
Supported use: Reports 542,076 industrial robot installations in 2024, an operational stock of 4,663,698, and regional and national splits.
Important limit: IFR compiles producer and national-association survey data; installations do not measure generality, utilization, task success, jobs displaced, or profitability
measured-trend · horizon.robotics-spatial-mobility.rsm-03
Supported use: Reports professional, medical, consumer and robot-as-a-service unit sales and a producer census.
Important limit: Supplier participation and coverage vary; annual units are not an installed base and fast percentage growth can start from very small numbers
measured-trend · horizon.robotics-spatial-mobility.rsm-04
Supported use: Reports more than one million robots across more than 300 facilities and a 10% fleet-travel-efficiency gain from its DeepFleet model.
Important limit: This is a company statement about its own, highly structured network; definitions, counterfactuals and independent audit are not supplied
measured-trend · horizon.robotics-spatial-mobility.rsm-05
Supported use: Documents a national-health-service deployment and a reported reduction in some delivery times from as much as four hours to about 15 minutes.
Important limit: It is an older programme account, useful as longitudinal evidence but not proof of identical economics elsewhere
measured-trend · horizon.robotics-spatial-mobility.rsm-06
Supported use: RBC reports a 2025 pilot for malaria-drug delivery; RISA lists a 13.4 billion Rwandan-franc programme to build shared operational capacity.
Important limit: Programme pages do not provide a comparable cost-per-delivery series
measured-trend · horizon.robotics-spatial-mobility.rsm-07
Supported use: Sets out a scheme to equip 15,000 women’s self-help groups to provide agricultural drone services, with public financing, operator training and staged distribution.
Important limit: This is policy and partial implementation evidence, not an independent income or agronomic-impact evaluation
measured-trend · horizon.robotics-spatial-mobility.rsm-08
Supported use: Aggregates data from more than 20 institutions, 22 robot embodiments, more than one million episodes and over 500 skills.
Important limit: The corpus demonstrates cross-institution data pooling, not reliable open-world autonomy
measured-trend · horizon.robotics-spatial-mobility.rsm-09
Supported use: Reports zero-shot cross-embodiment and unseen-task results for a vision-language-action model.
Important limit: It is a company-authored preprint; benchmark gains do not establish field reliability, safety, cost, or maintainability
measured-trend · horizon.robotics-spatial-mobility.rsm-10
Supported use: Describes a generalist humanoid foundation model trained with real, simulated and synthetic data and reports simulation benchmark gains.
Important limit: Company research; no inference about commercial task economics is warranted
measured-trend · horizon.robotics-spatial-mobility.rsm-11
Supported use: Reports embodied reasoning, dexterity and spatial-reasoning capabilities across research platforms.
Important limit: These are developer-reported model results, not independent long-duration deployments
measured-trend · horizon.robotics-spatial-mobility.rsm-12
Supported use: Introduces a simulated whole-body humanoid benchmark and reports that contemporary reinforcement-learning methods struggled on most tasks.
Important limit: Simulation omits many field failure modes, but the result is a useful check on generalized-capability claims
measured-trend · horizon.robotics-spatial-mobility.rsm-14
Supported use: Requires specified manufacturers and operators to report qualifying crashes.
Important limit: NHTSA explicitly warns that reports can be incomplete, duplicated, misclassified and lack exposure denominators, so raw counts cannot rank system safety
measured-trend · horizon.robotics-spatial-mobility.rsm-15
Supported use: Announces a policy framework built around safety, domestic innovation and commercial deployment.
Important limit: It is policy intent; detailed implementation and state law remain material
measured-trend · horizon.robotics-spatial-mobility.rsm-16
Supported use: Reports more than 400,000 paid rides per week across six US metropolitan areas in early 2026.
Important limit: It is a company-reported activity metric; it does not disclose a common intervention, profitability or matched safety denominator
measured-trend · horizon.robotics-spatial-mobility.rsm-17
Supported use: Reports 3.2 million fully driverless Apollo Go rides in the quarter, a weekly peak above 350,000, more than 22 million cumulative rides by April, and operation in 27 cities by May.
Important limit: These are company statements and do not show equal service depth or operating design domains across cities
measured-trend · horizon.robotics-spatial-mobility.rsm-18
Supported use: Reports more than 16,000 autonomous-driving test licences and 32,000 kilometres of roads opened for testing by August 2024, and 20 city or cluster pilots for vehicle-road-cloud integration.
Important limit: Government reporting; comparable exposure and disengagement data are not supplied
measured-trend · horizon.robotics-spatial-mobility.rsm-19
Supported use: Establishes a framework for fully driverless systems including safety management, scenario testing, a safety case and in-service monitoring, subject to implementation by contracting parties.
Important limit: Adoption is not the same as national approvals or fleet deployment
measured-trend · horizon.robotics-spatial-mobility.rsm-20
Supported use: Provides type-approval procedures for fully automated vehicles in predefined areas, fixed hub-to-hub routes and automated valet parking, with safety-management, testing and in-use reporting requirements.
Important limit: Scope is bounded rather than permission for unrestricted autonomy
measured-trend · horizon.robotics-spatial-mobility.rsm-21
Supported use: Reports a driverless Class 8 freight launch on the Dallas–Houston corridor, initially in daytime clear-weather operation, and approximately 20,000 driverless commercial miles by the end of June 2025.
Important limit: This is a narrow operating domain and a company-reported metric
measured-trend · horizon.robotics-spatial-mobility.rsm-22
Supported use: Routine small-drone operations still require visual line of sight unless an operator obtains a waiver or uses another certification route; package delivery can proceed under Part 135; a broader BVLOS rule remained a proposal after its August 2025 publication
Important limit: unknown
measured-trend · horizon.robotics-spatial-mobility.rsm-23
Supported use: Defines designated airspace services including flight authorization, geo-awareness, network identification and traffic information; information-security requirements were incorporated in the current easy-access rules.
Important limit: Designation and service provision are implementation work, not evidence of uniform EU coverage
measured-trend · horizon.robotics-spatial-mobility.rsm-24
Supported use: Reports 1.27 million registered civil unmanned aircraft and 23.11 million flight hours in 2023; later official reporting lists more than 14,000 certified companies and 225,000 licensed operators. An official projection put the low-altitude economy above CNY 2 trillion by 2030.
Important limit: Registration and flight-hour counts are facts as reported; the economic value is a forecast whose definition is broad
measured-trend · horizon.robotics-spatial-mobility.rsm-25
Supported use: The national plan names robotics as a strategic emerging industry and embodied intelligence as a future industry; a Hangzhou national pilot base reported 130 robots across 30 vocational scenarios.
Important limit: The former is policy intent and the latter an operator/government deployment count, not independent task-performance evidence
measured-trend · horizon.robotics-spatial-mobility.rsm-26
Supported use: Sets a 2030 ambition for robots able to operate without human intervention under supervision in specified circumstances and for research automation.
Important limit: This is a research mission, not a capability forecast
measured-trend · horizon.robotics-spatial-mobility.rsm-27
Supported use: Promotes robot-friendly workflows and facilities and examines public implementation of medium-speed delivery robots.
Important limit: Policy intent; national adoption and unit economics remain unmeasured
measured-trend · horizon.robotics-spatial-mobility.rsm-28
Supported use: Targets at least 30% labour saving and 1.5-times productivity in construction by fiscal 2040 through automation, data linkage and management reform.
Important limit: These are ministry targets, not realized performance
measured-trend · horizon.robotics-spatial-mobility.rsm-29
Supported use: Announces 26 manufacturing projects with KRW 3.7 trillion of combined investment and target improvements in productivity, cost, defects and energy; separately coordinates robot companies, researchers and component suppliers around foundation-model work.
Important limit: Targets are intent, not measured outcomes
measured-trend · horizon.robotics-spatial-mobility.rsm-30
Supported use: Announces a 2030 physical-AI ambition, sovereign world-model work, a three-year foundation-model programme and full-stack demonstrations.
Important limit: Policy intent; no comparative capability baseline is supplied
measured-trend · horizon.robotics-spatial-mobility.rsm-31
Supported use: Dubai targets 25% of trips to be autonomous by 2030, has a licensing law, and reported more than 60 trial vehicles.
Important limit: The reported 20.4% current autonomous-transport share uses a broad multimodal definition that includes established automated public transport; it is not a road-robotaxi share
measured-trend · horizon.robotics-spatial-mobility.rsm-32
Supported use: Reports demonstrations across 15 construction categories and more than 20 robot types.
Important limit: Municipal pilot evidence; the page does not establish sustained productivity, paid utilization or safety performance
measured-trend · horizon.robotics-spatial-mobility.rsm-33
Supported use: Defines modifications for sensing using WLAN signals in sub-7.125 GHz and above-45 GHz bands.
Important limit: Publication proves technical standardization, not device penetration, permission for every use, or user acceptance
measured-trend · horizon.robotics-spatial-mobility.rsm-34
Supported use: Extends an industry interoperability standard to routers/access points, multi-admin operation and home-energy management.
Important limit: Alliance release; certification and installed-base data are needed before inferring practical interoperability
measured-trend · horizon.robotics-spatial-mobility.rsm-35
Supported use: Consolidates commonly used extensions into a cross-platform spatial-computing API.
Important limit: A stable interface can lower developer fragmentation, but it does not show headset adoption or a single spatial map across devices
measured-trend · horizon.robotics-spatial-mobility.rsm-36
Supported use: The Data Act has applied since 2025-09-12 and gives users rights over connected-product data; Cyber Resilience Act reporting obligations begin in 2026 and main obligations in 2027.
Important limit: Legal application does not prove easy portability, enforcement capacity or secure implementation
measured-trend · horizon.robotics-spatial-mobility.rsm-37
Supported use: Identifies trust, data integrity, identity, privacy and cyber-physical risks in digital twins and connected home-care environments.
Important limit: These are risk frameworks, not incident-rate estimates
measured-trend · horizon.robotics-spatial-mobility.rsm-38
Supported use: Two separate single-participant studies reported high-rate speech decoding, including approximately 62 words per minute in one study and roughly 32 words per minute with high accuracy after calibration in another.
Important limit: N=1, invasive surgery, training burden and long-term signal stability preclude consumer-interface generalization
measured-trend · horizon.robotics-spatial-mobility.rsm-39
Supported use: Release 17 introduced the first normative NTN specifications for 5G, including satellite access for handheld and IoT devices, while documenting delay, moving-cell, Doppler, GNSS and regulatory constraints.
Important limit: Standardization does not equal universal commercial service
measured-trend · horizon.robotics-spatial-mobility.rsm-40
Supported use: Establishes a regulatory framework for satellite operators to use terrestrial mobile spectrum through carrier partnership, with interim 911 conditions.
Important limit: It creates a permission path, not guaranteed coverage, capacity or affordable service
measured-trend · horizon.robotics-spatial-mobility.rsm-41
Supported use: A 12-year concession covers a planned multi-orbit constellation of more than 290 satellites; ESA schedules first launch for 2029, initial service in 2030 and full service in 2031.
Important limit: These dates and capabilities are contracted programme intent and remain delivery risks
measured-trend · horizon.robotics-spatial-mobility.rsm-42
Supported use: The 2026-07-31 statistics list about 46,100 regularly tracked objects, around 16,100 functioning spacecraft and more than 17,000 tonnes in orbit; models estimate much larger small-debris populations.
Important limit: Catalogued counts and model estimates must not be conflated
measured-trend · horizon.robotics-spatial-mobility.rsm-43
Supported use: Defines six usage scenarios, including immersive communication, ubiquitous connectivity, integrated sensing and communication, and AI-and-communication; technical requirements remained in the standards process in 2026.
Important limit: Capability values are targets for evaluation, not delivered network performance
measured-trend · horizon.robotics-spatial-mobility.rsm-44
Supported use: Reports more than 20 million electric-car sales in 2025, around one quarter of car sales, with China at approximately 55% and Chinese manufacturers around 60% of global electric-car sales; it also documents regional battery-price differences.
Important limit: EVs are an enabling industrial base, not a proxy for autonomous operation
measured-trend · horizon.robotics-spatial-mobility.rsm-45
Supported use: Reports high and in some cases rising concentration in refining, with the leading refiner averaging about 70% across key minerals in 2025 and China exceeding 90% in several specified materials. It also documents export-control and investment risks.
Important limit: The analysis combines official and commercial datasets; mineral exposure varies by machine design
measured-trend · horizon.robotics-spatial-mobility.rsm-46
Supported use: Projects data-centre electricity demand to roughly double to about 945 TWh by 2030 in its base case.
Important limit: This covers all data centres, not embodied AI alone, and is a scenario-model output rather than a measured 2030 result
measured-trend · horizon.robotics-spatial-mobility.rsm-47
Supported use: Finds that populations have already peaked in several large economies including China, Germany and Japan, and identifies very low fertility in China and Korea.
Important limit: Population projections are conditional and do not translate mechanically into robot demand
measured-trend · horizon.robotics-spatial-mobility.rsm-48
Supported use: Projects a global shortfall of 11 million health workers by 2030, concentrated largely in low- and lower-middle-income countries.
Important limit: A headcount gap does not show which tasks are technically or ethically automatable
measured-trend · horizon.robotics-spatial-mobility.rsm-49
Supported use: Projects 2.3 billion people will need care by 2030 and reports that most domestic work remains informal.
Important limit: This frames unmet and undervalued labour demand; it is not evidence that robotic care will be acceptable or cheaper
policy-intent · src-anthropic-frontier-roadmap-2026
Supported use: Supports the claim that one frontier developer plans for full or strong automation of top-tier research teams on a near-term horizon.
Important limit: The timing language is the company's planning belief, not a consensus forecast. Full research-team automation is a threshold to prepare for, not an observed result.
policy-intent · src-anthropic-rsp-2026
Supported use: Supports the existence and current wording of developer capability thresholds for automated research and development and related safeguards.
Important limit: This is a developer's current policy and can change. Capability thresholds show what the developer plans to monitor, not that the capabilities exist or will diffuse.
policy-intent · src-asean-ai-governance-guide-2024
Supported use: Supports the other-region variant's Southeast Asian emphasis on interoperable, voluntary, responsible AI governance across diverse member states.
Important limit: The guide is voluntary and focuses on responsible adoption rather than frontier capability. ASEAN member states differ greatly in infrastructure, regulation, income, and implementation capacity.
policy-intent · src-au-ai-priority-dialogue-2025
Supported use: Supports the Africa variant's current talent and funding concentration warning and the policy priority placed on inclusion and investment.
Important limit: Dialogue statistics are reported by the convening institution and should be checked against their underlying datasets before precise reuse. Continental aggregates hide large national differences.
policy-intent · src-au-continental-ai-strategy-2024
Supported use: Supports the Africa variant's continental direction on infrastructure, data, skills, inclusion, governance, and African-centred control.
Important limit: A continental strategy cannot establish delivery across highly varied states and infrastructure conditions. Compute, power, funding, language data, and implementation capacity remain uneven.
policy-intent · src-china-ai-plus-2025
Supported use: Supports the mainland China variant's state-coordinated AI Plus adoption direction and official 2027 and 2030 targets.
Important limit: Adoption percentages are official targets, not verified outcomes. Internationally comparable data on capability, incidents, and rights may remain limited.
policy-intent · src-deepmind-frontier-framework-2026
Supported use: Supports the existence of machine-learning research acceleration thresholds and developer monitoring of them.
Important limit: This is a developer's own risk framework and update. Tracked machine-learning research thresholds do not establish realised economy-wide acceleration.
modelled-projection · src-deepmind-robotics-er-card-2026
Supported use: Supports a current embodied-reasoning progress signal and the need to keep developer-reported robotics evidence separate from economy-wide reliability.
Important limit: This is developer-reported evidence that requires independent replication. Embodied reasoning demonstrations do not establish unattended household reliability or mass-market economics.
measured-trend · src-eclac-ilia-2025
Supported use: Supports the Latin America and Caribbean variant's cross-country differences in readiness, talent, investment, infrastructure, governance, and adoption.
Important limit: The index measures current readiness and gaps across 19 countries; it is not a 2031 forecast. Country-level index scores can hide subnational, informal-economy, and language differences.
policy-intent · src-eu-gpai-act-qa
Supported use: Supports the European Union variant's general-purpose-model obligations, systemic-risk threshold, documentation, and enforcement context.
Important limit: The official page did not expose a clear publication date; it was accessed on 2026-08-02. Legal systemic-risk thresholds and duties do not predict capability or enforcement capacity.
modelled-projection · src-iea-energy-and-ai-2025
Supported use: Supports the base-case 2030 data-centre electricity estimate, grid-delay counterforce, and energy-source variation.
Important limit: The 2030 figures are scenario-dependent estimates rather than fixed demand. The report does not establish that grid build-out will arrive where advanced compute demand appears.
modelled-projection · src-iea-key-questions-ai-energy-2026
Supported use: Supports the conditional 2030 rise in data-centre and AI-focused electricity demand and the claim that grid constraints are locational.
Important limit: Electricity demand estimates depend on deployment, efficiency, utilisation, and grid assumptions. Global totals can hide binding local connection, cooling, and water constraints.
measured-trend · src-ifr-world-robotics-2025
Supported use: Supports industrial-robot deployment scale and regional concentration while preserving the boundary between controlled-place automation and general physical intelligence.
Important limit: Installations of industrial and service robots do not measure general physical intelligence. Factory deployment does not establish safe, economical household or open-world autonomy.
measured-trend · src-ilo-exposure-indicators-2026
Supported use: Supports the measurement boundary that occupational exposure cannot stand in for profitable substitution, workflow redesign, employment, wages, or demand.
Important limit: Exposure indicators do not directly measure workflow redesign, profitability, labour demand, wages, or worker bargaining. The source is mainly a measurement warning rather than a numerical future estimate.
measured-trend · src-ilo-genai-jobs-2025
Supported use: Supports the current global occupational-exposure baseline and the distinction between task transformation and full job automation.
Important limit: Occupational exposure is not the same as job loss, wage change, or profitable automation. The analysis primarily reflects then-current systems, not the the high-capability discontinuity scenario threshold bundle.
measured-trend · src-imf-ai-future-work-2024
Supported use: Supports global and income-group exposure differences, complementarity versus displacement, and distribution risk from higher capital returns.
Important limit: The exposure model predates later agent capabilities and does not predict a discontinuity. It is an IMF staff discussion note, not an agreed forecast of employment or distribution.
policy-intent · src-india-ai-mission-2024
Supported use: Supports the India variant's public mission funding, compute, dataset, skills, and inclusion direction.
Important limit: Approved funding and compute targets are inputs, not proof of delivered capacity or inclusive use. State-level service quality, connectivity, cost, and language coverage can remain uneven.
measured-trend · src-international-ai-safety-report-2026
Supported use: Supports the current autonomy baseline, the observed short task horizon, rapid but uncertain progress, mixed AI-research results, deployment lags, and the need to retain counter-scenarios.
Important limit: The report synthesises evidence available through December 2025, so later capability claims require separate checks. Trend continuation and risk scenarios are not forecasts, and benchmark progress may not transfer to messy real work.
policy-intent · src-japan-geniac-robotics-2026
Supported use: Supports the Japan part of the Japan-Korea variant's manufacturing, robotics, foundation-model, and labour-shortage direction.
Important limit: A selected programme proves industrial policy effort, not successful general robotics. It does not establish affordable care or household autonomy.
policy-intent · src-korea-ai-basic-act-2024
Supported use: Supports the Republic of Korea part of the Japan-Korea variant's AI infrastructure, innovation, safety, and trust-law direction.
Important limit: The law sets governance and infrastructure direction; implementation and effects remain uncertain. It does not demonstrate a capability discontinuity.
measured-trend · src-metr-long-tasks-2025
Supported use: Supports the measured historical increase in software-and-research task horizons and the need for fresh, independently scored long-task evaluation.
Important limit: The task set is weighted toward software and research work. A historical doubling trend does not guarantee continuation or transfer to 30-day, multi-stakeholder assignments.
scenario-assumption · src-oecd-ai-trajectories-2026
Supported use: Defines a credible high-capability boundary case in which systems approach human performance across most cognitive abilities and professional digital tasks; supports scenario distinctness, not likelihood.
Important limit: The four trajectories are scenarios, not predictions or assigned probabilities. The strongest trajectory is useful as a boundary case but does not show that human-level cognitive performance will arrive by 2030.
scenario-assumption · src-oecd-powerful-ai-curriculum-2025
Supported use: Supports a conditional curriculum shift toward foundational knowledge, judgement, verification, and deciding what should be done under powerful AI.
Important limit: The paper explores curriculum choices under a conditional future, not a settled capability forecast. National education systems may respond slowly and unevenly.
measured-trend · src-oecd-social-connections-2025
Supported use: Supports the mixed current evidence on digital interaction and social connection and the refusal to assign a fixed the high-capability discontinuity scenario social outcome.
Important limit: Digital and chatbot evidence is mixed and often observational. OECD-country patterns should not be generalised to all regions.
policy-intent · src-openai-preparedness-v2-2025
Supported use: Supports a hashable planning threshold for severe AI self-improvement and long-range autonomy risk; does not verify that either threshold has been crossed.
Important limit: This is a developer's governance framework, not an independent capability evaluation. Its self-improvement thresholds are planning triggers rather than proof that a system has crossed them.
policy-intent · src-saudi-data-ai-strategy
Supported use: Supports the Gulf variant's sovereign investment, specialist-skills, priority-sector, and 2030 AI ambition.
Important limit: The page states 2030 goals and investment ambitions, not verified delivery. It does not resolve cooling, water, imported-labour, component, or rights constraints.
measured-trend · src-stanford-ai-index-2026
Supported use: Supports the current picture of rapid but jagged digital-agent progress, weak household-robot performance, investment concentration, and early entry-work pressure.
Important limit: The report combines many benchmark and market sources with different methods. High performance on one task does not imply robust performance across domains, and labour signals are not clean causal estimates.
policy-intent · src-uae-ai-strategy-2031
Supported use: Supports the Gulf variant's UAE 2031 leadership ambition and priority-sector framing.
Important limit: The official PDF did not expose a reliable publication date; it was accessed on 2026-08-02. Leadership goals and priority sectors are policy intent, not proof of outcomes.
policy-intent · src-un-sids-digital-foundations
Supported use: Supports the other-region variant's small-island constraints in connectivity, digital capacity, finance, resilience, scale, and dependence on imported technology.
Important limit: The official topic page did not expose a single reliable publication date; it was accessed on 2026-08-02. It establishes structural constraints and agreed priorities, not a specific AI adoption path.
policy-intent · src-unesco-genai-education-2023
Supported use: Supports human-centred education, privacy, agency, inclusion, and the need to redesign assessment rather than equate cheap answers with learning.
Important limit: This is human-centred policy guidance, not a measured learning-outcome study. Its original evidence base predates the most capable systems in this register.
policy-intent · src-us-ai-action-plan-2025
Supported use: Supports the United States variant's emphasis on private innovation, infrastructure, exports, and national security.
Important limit: The plan proves federal policy direction, not future implementation or capability. State law, litigation, infrastructure, and market concentration can produce divergent outcomes.
policy-intent · src-who-multimodal-health-guidance-2024
Supported use: Supports health use cases, risk classes, and the continuing need for defined tasks, audits, clinicians, treatment capacity, and equitable access.
Important limit: The guidance identifies health uses, risks, and governance needs; it does not measure clinical benefit from future systems. Health outcomes still depend on data, trials, approvals, clinicians, treatment capacity, and equitable access.
measured-trend · src-who-social-connection-2025
Supported use: Supports the global baseline that social connection is a health-relevant scarcity and prevents treating future machine companionship as already-proven benefit or harm.
Important limit: The report establishes a social-connection baseline, not a causal effect of future AI. It cannot show that machine companionship will improve or worsen loneliness.
measured-trend · src-world-bank-ai-foundations-2025
Supported use: Supports developing-country differences in connectivity, compute, context, competency, frontier concentration, and lower-cost adoption routes.
Important limit: The four-foundations framework and concentration snapshot do not predict adoption under the high-capability discontinuity scenario. Web traffic and investment indicators are incomplete proxies for useful, affordable capability.