Methodology
Each Friday we explain where capital, adoption and founder behaviour are moving in AI and technology, why it matters and what would prove us wrong. We don't publish link dumps or hype.
Narratives (lead section)
Each week we rank 5–8 narratives. Conviction (0–10) = capital 25% + adoption 25% + founder behaviour 20% + corroboration 20% + attention 10%. Each factor is rated 0–5 by the analyst. Direction is shown as ↑↑ ↑ → ↓ ↓↓, and each narrative is classed NEW, ACCELERATING, CONFIRMED, STABLE, FADING or REVERSING against the prior week. The capital-flow ledger lists each transaction's amount, type and status, and gives totals only within comparable categories, never as one blended number. Valuations, undrawn credit and reported purchase orders are not counted as capital deployed. A 12-point quality gate withholds the issue if evidence is materially incomplete. No trading signals, no OSINT.
AI ecosystem signals
Layers: research/models · compute/infra · developer stack · apps/distribution · enterprise adoption · capital/talent · policy/safety/law · geopolitics/supply chain. Types: capability, cost/performance, access, adoption, constraint, competitive move, proof/disproof. radar_score = 0.25·materiality + 0.20·durability + 0.20·spillover + 0.15·evidence + 0.10·novelty + 0.10·utility, each factor 1–5; signals need 3.5 or higher.
Sourcing
Every material claim links to a dated source, with two independent sources where practical. Claims are tagged fact, interpretation or forecast, with high, medium or low confidence; uncertain items are marked [VERIFY]. We fail closed. Each issue ships a public JSON dataset and claim-to-source ledger.
Corrections
Corrections are appended to the issue with a timestamp (Asia/Jerusalem). Original wording is never silently changed.
Market context, not investment advice.