Week of 28 September 2026 (2026-W40)

In the week of 28 September 2026 (2026-W40), the Valdrans Market Pulse tracked 108 corroborated market-moving stories — 100 market stories and 8 qualified breaking-news stories — drawn from 2,046 source items across 14 distinct sources.

Market stories by category

corporate36
macro31
geopolitical14
political9
regulatory8
flight1
shipping1

Market stories by magnitude

High (7–10)18
Medium (4–6)77
Low (1–3)5
Unscored0

Most active sectors

Stories may touch multiple sectors; counts can exceed the story total.

energy41
technology27
artificial intelligence16
financials16
semiconductors16
financial services15
manufacturing12
shipping11

Breaking-news pipeline

Corroborated clusters tracked166
Qualified as market-relevant (HIGH + MEDIUM)8
Tier HIGH / MEDIUM / LOW / unclassified0 / 8 / 77 / 81

Coverage context

Source items processed2,046
Distinct sources14
Instruments touched306
Sectors touched127

Natural-hazard signals (USGS/NOAA)

Governmental hazard alerts ingested — a separate stream with its own severity scales, never mixed with market-story magnitude.

flood1909
earthquake139
fire weather7
severe weather591
tropical cyclone25
Total2671

Corroborated stories per week

Hollow points are backfilled weeks (see methodology).

Archive

Weeks that fail our publication thresholds are not published — see methodology.

Methodology

What counts as a story. Ingested items are grouped into stories by embedding-based clustering, so one real-world event arriving from many outlets is counted once. A cluster is counted as a corroborated story only when at least two distinct sources independently reported it — a single source cannot manufacture a story. Market stories additionally carry a pipeline magnitude score; breaking-news clusters count as stories only when our classification tier marks them market-relevant (HIGH or MEDIUM). Lower tiers remain visible in the “clusters tracked” figure but are never called stories.

Stories vs source items. “Source items processed” counts individual classified inputs (one story typically arrives as several source items). The two quantities are reported separately and never conflated.

Week boundary and snapshot timing. Weeks are ISO weeks, Monday 00:00 UTC to the following Monday 00:00 UTC (exclusive), bucketed by first-seen time. Each snapshot is computed Mondays at approximately 05:45 UTC. Published snapshots are immutable: corroboration that arrives after the snapshot is not retroactively recounted. Retrospectively computed (“backfilled”) weeks use final cluster state and are flagged as such.

Classification provenance. Story grouping, categories, magnitudes and tiers are assigned by our LLM-assisted, human-reviewed editorial pipeline; the Pulse counts themselves are computed by plain database arithmetic — no model output enters the numbers. Category and sector labels reflect the pipeline’s vocabulary as-is; figures may include a negligible number of internal test items.

Quality gates and revisions. A week is published only if internal consistency checks pass and volumes are within an expected band of the trailing median; failing weeks are held for human review, and weeks below a minimum activity floor are skipped. Corrections to a published week, if ever required, are announced on our corrections page. Methodology version: 1.

Reuse. The Pulse is free to quote and reuse with attribution to Valdrans (CC BY 4.0), citing valdrans.com/pulse. It is information about news flow, not investment advice.