COVERAGE MAP
The AI transition runs through eight layers — raw materials, compute, models, applications, capital, labor, rules, and geopolitics. We cover all eight with one discipline: verifiable evidence, graded sources, scored calls. Beats roll out in waves; we never claim coverage we can't sustain weekly.
What AI is physically made of: rare earths, gallium, HBM memory supply, land, water, cooling. The hardest constraints move slowest — and bind first.
Chips, fabs, data centers, contracted gigawatts. Capex is a hard-to-fake belief about timelines; power contracted today is capability delivered 2027–29. On pure silicon depth we cite SemiAnalysis rather than compete — our ground is integration.
Frontier capability and the institutions building it: task horizons, benchmark saturation, lab economics, the AI-share of AI research, and what insiders say versus what they said last year.
Where capability becomes products: agents in deployment, vertical AI, who is a real business versus a wrapper — and what happens when deployed agents misbehave.
How money prices the transition: prediction markets, VC concentration, public-market AI exposure, M&A. The spread between evidence and price is this desk's founding subject.
The displacement front edge: entry-level hiring in exposed categories (not the headline rate), AI-attributed job cuts, productivity signals, and the first new job categories.
Regulation, compliance, liability, standards: EU AI Act enforcement, US export controls, state laws, agent-liability precedents — the constraints that decide what deployed capability is allowed to do.
US–China, chip diplomacy, sovereign AI programs, military adoption: the geopolitics that every fast-timeline scenario runs through — and the tail risks allocators most underprice.