FUURAA AI Knowledge Library
A growing library for the ideas, evidence and systems shaping AI civilization.
A structured gateway to frontier AI research, evidence, systems, products and regional intelligence—curated for fast discovery and deeper reading.
Knowledge collections
From research collections to practical records and check tools.
Every collection remains independently addressable while contributing to a shared map of frontier ideas, evidence and real-world systems.
Frontier Radar
Track consequential changes across models, agents, robotics, compute, science and governance.
Explore this section →Evidence dossiersAI Evidence Atlas
Inspect claims, evidence states, limitations and public source trails before conclusions harden.
Explore this section →Regional intelligenceSoutheast Asia Observatory
Read AI development through the region’s languages, economies, institutions and deployment realities.
Explore this section →Research programmesResearch Domains
Explore long-horizon questions spanning agency, identity, memory, embodiment, causality and society.
Explore this section →Systems architectureAI Civilization Architecture
Connect research questions to identity, memory, orchestration, governance and accountable infrastructure.
Explore this section →Signal indexFrontier Signal Library
Search observed, emerging and forecast signals with dates, sources and editorial boundaries.
Explore this section →External reviewGlobal AI Review
Follow research institutions, standards bodies, public agencies and scientific publishers worldwide.
Explore this section →Intellectual foundationsIdeas & Institutional Principles
Examine the ideas that shape FUURAA’s long-term technical and institutional decisions.
Explore this section →Practical methodAI Evidence Literacy Guide
Evaluate claims through source class, study design, artefact provenance, transfer boundaries and a dated evidence state.
Read full guide →Downloadable record templateAI Claim Evidence Record
Create an on-device, handoff-ready JSON record connecting one AI claim to its source, evaluation, artefacts, boundary and review state.
Use record template →Record check toolAI Claim Evidence Record Check Tool
Check a FUURAA AI-claim evidence record locally for structural completeness, chronology and a bounded review state before human assessment.
Open check tool →Accountable review methodAI Claim Human Review Protocol
Move from a structurally reviewable evidence record to a bounded, dated and reopenable conclusion with separated human roles.
Read full guide →System evaluation methodAI Agent Evaluation Protocol
Evaluate an acting system across identity, tasks, process, variance, recovery, adversarial conditions, transfer, human handoff and scoped release gates.
Read full guide →Downloadable record templateAI Agent Evaluation Run Record
Build an on-device JSON record that freezes the agent system, task distribution, evaluation design, per-run outcomes, transfer boundary and scoped release decision.
Use record template →Record check toolAI Agent Evaluation Run Record Check Tool
Check an Agent evaluation run record locally for complete system identity, task and evaluation structure, results, transfer boundaries, release state and expiry before human review.
Open check tool →Commissioning and handover protocolAI Agent Commissioning & Authority Handover
Move from evaluation evidence to a bounded release by binding exact system identity, least authority, human control, canary evidence, independent authorisation and operating ownership.
Read full guide →Operating evidence protocolAI Agent Post-Deployment Monitoring
Connect deployment identity, operating signals, drift, change control, incident containment, rollback and re-evaluation through eight evidence gates.
Read full guide →Downloadable record templateAI Agent Operating-Period Record
Build an on-device fourteen-field JSON record connecting deployment identity, signals, action evidence, change, incidents, rollback and renewed review.
Use record template →Record check toolAI Agent Operating-Period Record Check Tool
Check an operating-period JSON locally for deployment identity, decision expiry, authority, monitoring, action evidence, incidents, rollback and review continuity.
Open check tool →Incident and recovery protocolAI Agent Incident Response & Re-authorisation
Move from authority containment and evidence preservation through impact analysis, recovery verification and an independent, expiring re-authorisation decision.
Read full guide →Downloadable record templateAI Agent Incident Decision Record Template
Build an on-device fifteen-field JSON record connecting incident identity, containment, evidence, impact, causal challenge, recovery verification and an expiring decision.
Use record template →Record check toolAI Agent Incident Decision Record Check Tool
Check an incident decision JSON locally for incident and deployment identity, containment, evidence custody, impact, causal challenge, recovery tests, independent review and decision expiry.
Open check tool →Lifecycle closure protocolAI Agent Retirement & Evidence Preservation
Retire an acting system through authority revocation, dependency mapping, evidence preservation, bounded data disposition, successor handoff and independently verified closure.
Read full guide →Downloadable record templateAI Agent Retirement Closure Record
Build an on-device fifteen-field JSON record for authority revocation, evidence preservation, data disposition, successor handoff, independent verification and expiring closure.
Use record template →Record check toolAI Agent Retirement Closure Record Check Tool
Check a retirement-closure JSON locally for identity, authority revocation, preservation, disposition, continuity, verification, review and expiry through ten bounded checks.
Open check tool →Change decision protocolAI Agent Material Change & Recommissioning
Decide whether an update remains covered by its commissioning decision or requires bounded verification, partial re-evaluation, full recommissioning, rejection or reversal.
Read full guide →Downloadable record templateAI Agent Material Change Decision Record
Build an on-device sixteen-field JSON record connecting a candidate update to its baseline decision, change surfaces, evidence transfer, verification, rollback, authorisation and transition review.
Use record template →Record check toolAI Agent Material Change Decision Record Check Tool
Check a material-change decision JSON locally for baseline identity, differences, dependencies, affected context, evidence transfer, verification, rollback, decision expiry and transition proof.
Open check tool →Downloadable record templateAI Agent Commissioning Decision Record
Build an on-device fifteen-field JSON record connecting exact release, use, evaluation, authority, human control, canary evidence, independent decision and operating ownership.
Use record template →Record check toolAI Agent Commissioning Decision Record Check Tool
Check a commissioning decision JSON locally for release identity, use, evidence, authority, controls, canary, monitoring, incidents, owners and expiry through ten bounded checks.
Open check tool →Lifecycle operating mapAI Agent Lifecycle Evidence Chain
Follow six decision stages and eighteen public resources across evaluation, commissioning, operation, material change, incidents and retirement—without losing identity, authority, ownership or expiry at handover.
Read full guide →Downloadable evidence indexAI Agent Lifecycle Evidence Index
Build a local fifteen-field JSON index that points to authoritative records across six lifecycle stages while preserving current authority, expiry, reference integrity, unknowns and accepted handover.
Use record template →Index check toolAI Agent Lifecycle Evidence Index Check Tool
Run ten local structural, reference-integrity and temporal checks before independently opening every authoritative record; files are not uploaded by default.
Open check tool →Operating decision mapAI Agent Execution Assurance Path
Move from real failure symptoms through five rejectable decision gates and fourteen engineering dossiers covering identity, authority, current execution, retries, revocation, reconciliation and recovery.
Read full guide →Routing evaluation guideHow to Evaluate Multi-Model AI Routing
Use six rejectable evidence gates to inspect route identity, semantic fidelity, objectives, failure handoff, comparable outcomes and data governance—without confusing connector count with resilience.
Read full guide →Human oversight evaluation guideHow to Evaluate Meaningful Human Oversight
Use six evidence gates to test whether human involvement has real authority, usable information, time, capacity, intervention mechanics and accountable closure.
Read full guide →Benchmark reading guideHow to Read AI Benchmark and Leaderboard Claims
Use six evidence gates to reconstruct the evaluated system, measurement target, data lineage, judging protocol, uncertainty and transfer boundary behind a score.
Read full guide →Web-agent evaluation guideHow to Evaluate AI Agents That Use the Web
Use six evidence gates to reconstruct starting state, observation, authority, independent outcomes, side effects, recovery and transfer behind a web-task success claim.
Read full guide →Citation verification guideHow to Verify AI-Generated Citations
Use six evidence gates to resolve the exact source, verify metadata and status, locate claim support, preserve transformations and set a dated evidence state.
Read full guide →Research synthesis review guideHow to Review AI-Generated Research Syntheses
Use six evidence gates to reconstruct the question, search, selection, extraction, appraisal, synthesis, provenance and expiry behind a fluent research review.
Read full guide →Quantitative-claim verification guideHow to Verify AI-Generated Quantitative Claims
Use six evidence gates to reconstruct the population, denominator, units, data version, transformations, uncertainty and dated interpretation behind a precise number.
Read full guide →AI energy and carbon claim guideHow to Evaluate AI Energy and Carbon Claims
Use six evidence gates to align workload, useful output, system boundary, measured energy, electricity factors, hardware life cycle, uncertainty, total impact and expiry.
Read full guide →AI latency and reliability claim guideHow to Evaluate AI Latency and Reliability Claims
Use six evidence gates to align user-visible outcomes, workload, quality, end-to-end timing, load state, tail distributions, failure denominators and expiry.
Read full guide →Multilingual and cross-cultural AI capability guideHow to Evaluate Multilingual and Cross-Cultural AI Capability Claims
Use six evidence gates to align language populations, varieties, scripts, tasks, data provenance, scoring, native review, disparities, safety, transfer and expiry.
Read full guide →AI long-context and memory claim guideHow to Evaluate AI Long-Context and Memory Claims
Use six evidence gates to align the exact system, counted window, usable length, position, distractors, integration, memory persistence, full denominators and expiry.
Read full guide →AI factuality and hallucination claim guideHow to Evaluate AI Factuality and Hallucination Claims
Use six evidence gates to align the factuality definition, reference facts, atomic claims, independent evidence, complete denominators, abstention, uncertainty and expiry.
Read full guide →AI reasoning capability claim guideHow to Evaluate AI Reasoning Capability Claims
Use six evidence gates to align the reasoning construct, task distribution, contamination, prompts, tools, budgets, answers, steps, faithfulness, variance, transfer and expiry.
Read full guide →AI coding capability claim guideHow to Evaluate AI Coding Capability Claims
Use six evidence gates to align the exact task, repository state, contamination, context, tools, permissions, independent tests, review, security, complete denominators, human work and expiry.
Read full guide →Multimodal AI capability claim guideHow to Evaluate Multimodal AI Capability Claims
Use six evidence gates to align modalities and tasks, input fidelity, preprocessing, provenance, cross-modal dependence, validated judges, complete failures, transfer and expiry.
Read full guide →Embodied AI and robot autonomy claim guideHow to Evaluate Embodied AI and Robot Autonomy Claims
Use six evidence gates to align embodiment, task, hardware, calibration, human contribution, complete physical trials, safety, recovery, simulation transfer, operating controls and expiry.
Read full guide →Speech recognition and spoken-language claim guideHow to Evaluate AI Speech Recognition and Spoken-Language Claims
Use six evidence gates to align task and language, audio acquisition, data provenance, reference transcripts, scoring, complete failures, latency, human correction, real-use transfer and expiry.
Read full guide →AI image generation capability claim guideHow to Evaluate AI Image Generation Capability Claims
Use six evidence gates to align use, prompts, generator version, run configuration, provenance, fidelity, composition, human preference, complete sampling, selection, safety, cost and expiry.
Read full guide →AI video generation capability claim guideHow to Evaluate AI Video Generation Capability Claims
Use six evidence gates to align video task, input, prompts, generator and temporal pipeline, provenance, frame and temporal quality, action, physics, complete clips, selection, safety, disclosure, cost and expiry.
Read full guide →AI document understanding and OCR capability claim guideHow to Evaluate AI Document Understanding and OCR Capability Claims
Use six evidence gates to align task, document population, acquisition and rendering, OCR and model versions, transcription, layout, tables, fields, document VQA, complete pages, corrections, cost and expiry.
Read full guide →AI text summarization capability claim guideHow to Evaluate AI Text Summarization Capability Claims
Use six evidence gates to align task, source population, pipeline, source support, salient coverage, critical omissions, contradictions, compression, presentation, complete runs, human correction, cost and expiry.
Read full guide →Retrieval-augmented generation claim guideHow to Evaluate Retrieval-Augmented Generation Claims
Use six evidence gates to align pipeline identity, corpus provenance, permissions, ingestion, retrieval, reranking, claim-level support, citations, conflict, abstention, complete failures, cost and expiry.
Read full guide →AI machine translation capability claim guideHow to Evaluate AI Machine Translation Capability Claims
Use six evidence gates to align language direction, source provenance, references, pipeline, meaning, terminology, critical errors, complete failures, post-editing, cost and expiry.
Read full guide →AI tool-use and function-calling capability claim guideHow to Evaluate AI Tool-Use and Function-Calling Capability Claims
Use six evidence gates to align task and authority, tool schemas, selection, arguments, execution, final state, complete failures, recovery, side effects, human work, cost and expiry.
Read full guide →AI question-answering capability claim guideHow to Evaluate AI Question-Answering Capability Claims
Use six evidence gates to align task and answer contract, questions, sources, retrieval, context, support, completeness, abstention, complete failures, human work, cost and expiry.
Read full guide →AI information-extraction capability claim guideHow to Evaluate AI Information-Extraction Capability Claims
Use six evidence gates to align task and schema, sources and annotations, ingestion, boundaries and types, links and relations, complete records, all failures, human correction, cost and expiry.
Read full guide →AI data-analysis and tabular-reasoning capability claim guideHow to Evaluate AI Data-Analysis and Tabular-Reasoning Capability Claims
Use six evidence gates to align task and data, schema and business definitions, parsing, queries, calculations, result support, charts, all failures, human review, cost and expiry.
Read full guide →AI forecasting and time-series capability claim guideHow to Evaluate AI Forecasting and Time-Series Capability Claims
Use six evidence gates to align targets and horizons, data availability, rolling backtests, baselines, point and probabilistic accuracy, all failures, cost, regime change and expiry.
Read full guide →AI recommendation and personalization capability claim guideHow to Evaluate AI Recommendation and Personalization Capability Claims
Use six evidence gates to align surfaces and outcomes, exposure and labels, candidate generation, ranking and policy, offline metrics, online effects, all failures, feedback loops and expiry.
Read full guide →AI text-classification and intent-routing capability claim guideHow to Evaluate AI Text-Classification and Intent-Routing Capability Claims
Use six evidence gates to align label ontologies, source populations, annotation, processing and scores, classwise errors, calibration, unsupported input, all failures, drift and expiry.
Read full guide →AI speech-synthesis capability claim guideHow to Evaluate AI Speech-Synthesis Capability Claims
Use six evidence gates to align text and voice conditions, pronunciation, generation pipelines, intelligibility, prosody, naturalness, all failures, delivery and expiry.
Read full guide →Curated reading paths
Do not just browse. Follow a question across disciplines.
For founders and leaders
Start with the Frontier Radar, then move from regional implications to governance and system architecture.
2026 Radar → Southeast Asia → Governance → ArchitectureOpen reading path →02For researchers and engineers
Begin with a research domain, inspect its evidence dossiers, then trace the idea into technical systems.
Research Domain → Evidence Atlas → ArchitectureOpen reading path →03For policy and public-interest readers
Combine regional observation, source methodology, institutional principles and disclosure boundaries.
Regional Observatory → Sources → Principles → ProgressOpen reading path →