Topics, framings, and source patterns derived from the discourse corpus — written commentary plus podcast transcripts, joined on the same canonical topics. Click any card or heatmap row for the full topic detail. Methodology →
Coverage through 6 Jun 2026 · topic analysis updated 8h ago
Topics, MPs, or sources whose volume jumped sharply versus their prior 4-week mean. Detected automatically and reviewed for noise before display.
The canonical topics with the most coverage across commentary and podcast transcripts in the trailing 7 days — a sliding window, so the cards keep working through the Monday handover. The trend figure compares the selected period to its prior 4-week mean. The stance bar (where visible) shows the supportive → critical mix on classifier-tagged edges. Distinct from the Anomalies above, which surface step-change spikes rather than absolute volume.
Each cell = items mentioning that topic in that week. Top 25 by recent volume.
A cross-tab of which sources are talking about which topics, week by week. Rows are the 10 topics with the most coverage in the last four weeks; columns are the 8 sources writing about them most. Each tiny line is 12 weekly counts — the path shape tells you whether the source is just-now picking up the topic, sustaining attention, or letting it fade. The trailing dot marks the most recent week, the number under the line is the 12-week total, and the column-header pill is the source’s editorial lean (which tints the line). Cells with no coverage render as a dashed midline so “quiet” reads differently from “missing data”.
A 30-item run in a 500-reader Substack and a 5-item run in Stuff (~2.3M monthly readers) look identical when you count articles. They are not the same thing. The same topic is being published more often in one place but heard far more in the other.
Toggling to reach-weighted multiplies each cell by
log1p(source_reach_score) — a dampened
weight that preserves tail-source visibility (small blogs
don’t collapse to zero) while letting major-outlet
coverage outweigh long-tail volume. This isn’t a
quality judgement; it’s a closer answer to the
question “whose attention has this topic actually
captured?”. Source reach scores are described in full
on the methodology page.
Item count answers “who’s writing about this most?”. Reach-weighted attention answers “whose attention has it actually captured?”. Same data, different question.
| Topic ↓ / Source → |
Waatea News
Government / N-A
|
Newsroom Opinion
Centre
|
The Daily Blog
Left
|
The Spinoff
Centre-left
|
Point of Order
Centre-right
|
mike-hosking-breakfast
Government / N-A
|
hdpa-drive
Government / N-A
|
Democracy Project
Centre
|
|---|---|---|---|---|---|---|---|---|
| Cost Of Living Pressures On Events | 53 | 2 | 5 | 6 | 2 | 9 | 5 | 2 |
| Public Sector Cuts | 15 | 7 | 12 | 10 | 5 | 3 | 11 | 7 |
| Cost Of Living | 27 | 1 | 11 | 7 | 4 | 5 | 5 | 4 |
| Treaty Of Waitangi Reinterpretation | 14 | 5 | 6 | 5 | 10 | 1 | · | 2 |
| Treaty Of Waitangi Implementation | 29 | 1 | 1 | 1 | 4 | 1 | · | · |
Each bubble is one canonical topic; size scales with item count over the last eight weeks. A line connects two topics when they appear together in at least two articles — thicker lines = stronger pairing. Quartile colours flag the busiest topics (Dominant) versus the long tail (Periphery).
