{"title":"AI, AGI, ASI, and Abundance Evidence Ledger","description":"A reviewed claim-level ledger separating observed findings, institutional definitions, projections, inferences, scenarios, and unresolved questions in the path from AI capability to universal access.","version":"1.0.0","published":"2026-07-26","updated":"2026-07-26","reviewCadenceDays":30,"methodologyUrl":"/research/methodology","entries":[{"id":"openai-charter-agi-definition","claim":"OpenAI defines AGI as highly autonomous systems that outperform humans at most economically valuable work.","dimension":"AGI definition","evidenceClass":"Attributed definition","status":"verified","source":{"id":"openai-charter","publisher":"OpenAI","title":"OpenAI Charter","url":"https://openai.com/charter/","published":null,"accessed":"2026-07-26","tier":1,"evidenceClass":"Attributed definition"},"firstAdded":"2026-07-26","lastVerified":"2026-07-26","geography":"Global institutional definition","metric":null,"methodology":"Direct reading of the organization’s published charter.","uncertainty":"The definition is authoritative for OpenAI but is not a universal standard or proof that a system meets it.","counterevidence":"Google DeepMind proposes a multidimensional levels framework rather than one binary threshold.","affectedRoutes":["/artificial-general-intelligence","/artificial-superintelligence"]},{"id":"deepmind-levels-framework","claim":"Google DeepMind’s Levels of AGI framework separates depth of performance from breadth of generality and discusses autonomy as a deployment consideration.","dimension":"AGI measurement","evidenceClass":"Measured finding","status":"verified","source":{"id":"deepmind-levels","publisher":"Google DeepMind","title":"Levels of AGI for Operationalizing Progress on the Path to AGI","url":"https://deepmind.google/research/publications/66938/","published":"2024-07-21","accessed":"2026-07-26","tier":2,"evidenceClass":"Measured finding"},"firstAdded":"2026-07-26","lastVerified":"2026-07-26","geography":"Global research framework","metric":null,"methodology":"Review of the publication abstract and framework description.","uncertainty":"The ontology is proposed by its authors and is not a binding certification system.","counterevidence":"Other organizations use different threshold definitions focused on economic work or autonomy.","affectedRoutes":["/artificial-general-intelligence"]},{"id":"deepmind-four-asi-paths","claim":"Google DeepMind analyzes four possible AGI-to-ASI pathways: scaling AGI, AI paradigm shifts, recursive improvement, and large-scale multi-agent collectives.","dimension":"AGI-to-ASI pathways","evidenceClass":"Scenario","status":"verified","source":{"id":"deepmind-agi-to-asi","publisher":"Google DeepMind","title":"From AGI to ASI","url":"https://deepmind.google/research/publications/239142/","published":"2026-06-12","accessed":"2026-07-26","tier":2,"evidenceClass":"Scenario"},"firstAdded":"2026-07-26","lastVerified":"2026-07-26","geography":"Global theoretical analysis","metric":{"name":"Proposed pathway count","value":"4","unit":"pathways","denominator":"Pathways analyzed in the report"},"methodology":"Direct extraction from the publication abstract.","uncertainty":"The report presents pathways and open research questions, not probabilities or dates.","counterevidence":"Unknown future paradigms may not fit the four categories, and material frictions may limit every path.","affectedRoutes":["/artificial-superintelligence","/research/agi-to-abundance-bottleneck-map"]},{"id":"deepmind-asi-definition","claim":"Google DeepMind describes artificial general superintelligence intuitively as more intelligent and cognitively capable than large organizations of humans.","dimension":"ASI definition","evidenceClass":"Attributed definition","status":"verified","source":{"id":"deepmind-agi-to-asi","publisher":"Google DeepMind","title":"From AGI to ASI","url":"https://deepmind.google/research/publications/239142/","published":"2026-06-12","accessed":"2026-07-26","tier":2,"evidenceClass":"Attributed definition"},"firstAdded":"2026-07-26","lastVerified":"2026-07-26","geography":"Global theoretical analysis","metric":null,"methodology":"Direct reading of the report abstract.","uncertainty":"No accepted operational test or verified current instance follows from this intuitive definition.","counterevidence":"Domain-specific superhuman performance does not establish general superiority over human organizations.","affectedRoutes":["/artificial-superintelligence"]},{"id":"openai-personal-agi-goal","claim":"OpenAI’s 2026 plan states a goal of giving everyone on Earth a personal AGI.","dimension":"Institutional objective","evidenceClass":"Verified product fact","status":"verified","source":{"id":"openai-benefit-plan","publisher":"OpenAI","title":"Built to benefit everyone: our plan","url":"https://openai.com/index/built-to-benefit-everyone-our-plan/","published":"2026-06-08","accessed":"2026-07-26","tier":1,"evidenceClass":"Verified product fact"},"firstAdded":"2026-07-26","lastVerified":"2026-07-26","geography":"Global stated objective","metric":null,"methodology":"Direct reading of OpenAI’s published plan.","uncertainty":"This records a goal, not delivery, coverage, affordability, or an independently verified AGI capability.","counterevidence":"The same source says transformative technologies can concentrate power and that broad benefit will not happen automatically.","affectedRoutes":["/artificial-general-intelligence","/universal-basic-abundance"]},{"id":"openai-abundant-affordable-intelligence","claim":"OpenAI’s 2026 plan identifies making advanced AI abundant, affordable, safe, useful, and easy to use as a central task.","dimension":"Intelligence access","evidenceClass":"Attributed definition","status":"verified","source":{"id":"openai-benefit-plan","publisher":"OpenAI","title":"Built to benefit everyone: our plan","url":"https://openai.com/index/built-to-benefit-everyone-our-plan/","published":"2026-06-08","accessed":"2026-07-26","tier":1,"evidenceClass":"Attributed definition"},"firstAdded":"2026-07-26","lastVerified":"2026-07-26","geography":"Global stated objective","metric":null,"methodology":"Direct reading of OpenAI’s published plan.","uncertainty":"No universal affordability threshold, coverage result, or completion date is supplied.","counterevidence":"Access to intelligence alone does not establish access to housing, energy, healthcare, food, or productive capital.","affectedRoutes":["/universal-basic-abundance","/research/agi-to-abundance-bottleneck-map"]},{"id":"anthropic-abundance-labor-scenario","claim":"Anthropic’s Economic Policy Framework considers a scenario in which AI generates unprecedented abundance while acting as a general substitute for labor.","dimension":"Labor and distribution","evidenceClass":"Scenario","status":"verified","source":{"id":"anthropic-economic-framework","publisher":"Anthropic","title":"Policy on the AI Exponential: Economic Policy Framework","url":"https://www.anthropic.com/policy-on-the-ai-exponential/epf","published":null,"accessed":"2026-07-26","tier":1,"evidenceClass":"Scenario"},"firstAdded":"2026-07-26","lastVerified":"2026-07-26","geography":"United States policy proposal","metric":null,"methodology":"Direct reading of Anthropic’s policy page and its description of the framework.","uncertainty":"Anthropic explicitly says the pace of capability development and diffusion is uncertain.","counterevidence":"The scenario is not a measured forecast of unemployment, productivity, or abundance.","affectedRoutes":["/universal-basic-abundance","/universal-basic-abundance-vs-ubi"]},{"id":"metr-time-horizon-definition","claim":"METR defines a 50% task-completion time horizon as the human-expert task duration at which an agent is predicted to succeed half the time.","dimension":"Agent capability measurement","evidenceClass":"Measured finding","status":"verified","source":{"id":"metr-time-horizons","publisher":"METR","title":"Task-Completion Time Horizons of Frontier AI Models","url":"https://metr.org/time-horizons/","published":"2026-02-06","accessed":"2026-07-26","tier":3,"evidenceClass":"Measured finding"},"firstAdded":"2026-07-26","lastVerified":"2026-07-26","geography":"Public frontier models; task suite is primarily software engineering, machine learning, and cybersecurity","metric":{"name":"Illustrative reliability threshold","value":"50","unit":"percent predicted success","denominator":"Tasks at the reported human-expert duration in the evaluated suite"},"methodology":"METR fits a logistic curve relating agent success to estimated human-expert task duration.","uncertainty":"METR says measurements above 16 hours are unreliable with its current suite, and performance does not generalize to all jobs.","counterevidence":"Real jobs include context, people, ambiguous goals, and outcomes that are difficult to score automatically.","affectedRoutes":["/artificial-general-intelligence"]},{"id":"arc-rhae-method","claim":"ARC-AGI-3 uses Relative Human Action Efficiency to score level completion and action efficiency against controlled human baselines.","dimension":"Generalization benchmark","evidenceClass":"Measured finding","status":"verified","source":{"id":"arc-methodology","publisher":"ARC Prize","title":"ARC-AGI-3 Scoring Methodology","url":"https://docs.arcprize.org/methodology","published":null,"accessed":"2026-07-26","tier":3,"evidenceClass":"Measured finding"},"firstAdded":"2026-07-26","lastVerified":"2026-07-26","geography":"Global evaluation environment","metric":{"name":"Total score range","value":"0–100","unit":"percent","denominator":"Average of all game scores"},"methodology":"Review of ARC-AGI-3’s documented completion, per-level efficiency, weighting, cap, and aggregation rules.","uncertainty":"The benchmark is one task environment and does not operationalize every dimension of AGI.","counterevidence":"A model can perform strongly in a benchmark environment while remaining unreliable or narrow elsewhere.","affectedRoutes":["/artificial-general-intelligence"]},{"id":"capacity-access-distinction","claim":"Higher productive capacity does not by itself establish delivered affordability or universal access.","dimension":"AI-to-abundance transmission","evidenceClass":"Inference","status":"supported","source":{"id":"openai-benefit-plan","publisher":"OpenAI","title":"Built to benefit everyone: our plan","url":"https://openai.com/index/built-to-benefit-everyone-our-plan/","published":"2026-06-08","accessed":"2026-07-26","tier":1,"evidenceClass":"Attributed definition"},"firstAdded":"2026-07-26","lastVerified":"2026-07-26","geography":"General analytical distinction","metric":null,"methodology":"Synthesis of primary sources that separately discuss capability, affordability, access, concentration, labor substitution, and distribution.","uncertainty":"The size and order of the bottlenecks vary by sector, country, technology, and policy.","counterevidence":"In competitive, elastic markets, higher capacity can lower prices without an explicit redistribution program; that still does not prove universal coverage.","affectedRoutes":["/universal-basic-abundance","/universal-basic-abundance-vs-ubi","/research/agi-to-abundance-bottleneck-map"]}],"changelog":[{"date":"2026-07-26","note":"Version 1.0.0: published ten claim-level records with uncertainty and counterevidence fields."}]}