Eight desks · one evidence layer
Follow the change, not the category noise.
AI moves through models, agents, infrastructure, institutions and the world around them. These desks organise discovery; the Atlas remains the evidence layer.
Models & Capabilities
What models can and cannot reliably do: modalities, reasoning, efficiency, evaluation and human–AI interaction.
Open desk →AI-AGT · 6 facetsAgents & Automation
How AI systems use tools, coordinate work and act with bounded autonomy—from workflow agents and protocols to robotics and reliability.
Open desk →AI-INF · 6 facetsCompute, Chips & Cloud
The physical and platform foundations of AI: accelerators, compute, cloud, edge infrastructure, energy and sovereign capacity.
Open desk →AI-DAT · 6 facetsData & Research
The datasets, knowledge systems, search methods, studies, benchmarks and measurement choices that shape what can be known.
Open desk →AI-SGR · 7 facetsSafety, Governance & Regulation
The risks and controls around AI: safety, security, privacy, organisational governance, standards, law, enforcement and documented incidents.
Open desk →AI-ECW · 6 facetsMarkets, Capital & Work
How AI changes investment, competition, organisations, labour, business models and the economics of building and using technology.
Open desk →AI-SEC · 10 facetsSectors & Applications
What changes when AI meets a field of practice—from commerce, finance and health to education, industry, public services, science and infrastructure.
Open desk →AI-GEO · 7 facetsGeopolitics & Society
How AI redistributes power across countries, institutions and communities, including sovereignty, industrial policy, culture, inequality and civic life.
Open desk →Classification is a route, not a verdict.
A desk cannot upgrade evidence class, source health, causal strength, priority or confidence. Default domain mappings are visible migration fallbacks, not confirmed claim assignments. Eligible secondary desks remain review options.