01
A journalist covering housing
Watches: Chris Bishop, “housing affordability,” Stuff, RNZ.
A free account
·
The whole brief, unlocked
Get told when the political conversation shifts on what you care about.
OpenBrief is a public-benefit tracker of New Zealand political conversation. Browsing is open to everyone — create a free account and you also get watchlists for MPs, topics and sources, a daily or weekly digest of what moved, and four weeks of history so you can see how a story built. No card, no catch.
openbrief.co.nz/pro/dashboard · Today’s Brief
BRIEF
Today’s Brief · built Tue, 5 May 09:12 NZST
This week’s signals: what crossed the threshold against the 4-week baseline.
3 narratives accelerating
2 fading
1 framing shift
2 watchlist moved
Cost of Living Crisis
Open ↗
- +124%vol · 7d
- neutral → crisisframing
- press leads 2dlead/lag
Why this matters. Urgency framing supports intervention narratives and is being adopted by opposition actors.
Term drift · Three Strikes
Open ↗
Apr 14three strikes
→
Apr 28repeat-offender bill
→
May 03tough-on-crime law
Why this matters. Two opposition MPs have adopted the new phrasing this week. Coverage is leading discourse.
See it in action
Watch all OpenBrief features in under 80 seconds.
Three example watchlists
Pick what you watch — we tell you when it moves.
02
A comms lead at a policy shop
Watches: cost of living, coalition tension, opposition MPs.
03
An engaged citizen in Botany
Watches: their local MP, “three strikes,” “rates.”
What a free account unlocks. Watchlists, on-page alerts and four weeks of history all live on your dashboard once you sign in. The watchlist digest is opt-in — choose daily or weekly in account settings, with per-item sensitivity tuning so you can mute a noisy MP without muting the topic. One-click unsubscribe in every email. We’d rather ship no alert than the wrong one.
How OpenBrief detects movement
Behind every alert: a measurable threshold, a sourced signal, and a one-click way to tell us we got it wrong.
Anomaly
A topic, MP, or source breaks out against its own trailing baseline. Minimum base counts so a quiet topic going from 0.8 mentions to 12 doesn’t fire as a 1500% spike. You can tune sensitivity per watchlist item.
Framing & alias drift
The phrasing on a tracked topic moves — new alias adopted, stance distribution tilts critical, two MPs use the new term in the same week. We log the term chain so you can see how the language shifted, with sources.
Lead-lag flip
Cross-corpus correlation between commentary and mainstream press flips on a tracked topic. We use Pearson r over a 12-week window with a minimum r threshold so noisy correlations don’t fire.
New record
A tracked MP files a new pecuniary entry, gives a Hansard speech on a tracked topic, or attends a ministerial diary meeting. A tracked source publishes new coverage on a tracked keyword. No threshold — just “something new in the public record on what you watch.”
Built for
Built for people who can’t read every outlet every day.
Journalists
Get notified when a story is breaking in discourse before it hits the mainstream press.
Watchlists on MPs and beats. Daily email tracking the rolling 7-day shifts against a 4-week baseline. Lead-lag between commentary and press so you can see a story building.
Policy analysts
Track narrative risk around live policy. Get alerted when framing shifts on the topics you own.
Alias-drift detection on a topic. First-use stamps when an MP adopts a new phrase. Every claim links back to the source.
Comms teams
Watch the topics, MPs, and outlets that matter to your client. Get told when something moves.
Per-outlet coverage spikes. Stance and framing shifts. A weekly summary you can paste into a client report.
Researchers & newsrooms
Set watchlists on long-running topics. Get a weekly digest of what changed, with every claim sourced.
Four weeks of history on a free account; the full archive, API access, and AI assistant integration on Team. Pearson-r lead-lag, topic co-occurrence, alias chains.
A worked example · alert that fired April 27
This is the kind of shift OpenBrief alerts you to.
Anyone watching “cost of living” got told the day the framing shifted — not the day mainstream press caught up.
Apr 14
cost of living
→
Apr 28
cost of living crisis
→
May 03
economic mismanagement
The framing-drift detector fired on April 27. Anyone with “cost of living” on their watchlist would have seen the shift surfaced on the dashboard that morning — six days before the third term landed in mainstream coverage.
For Team accounts · Research Explorer
One search across everything we ingest — and an evidence pile you can take with you.
The Research Explorer is the analyst’s workbench behind the alerts. Instead of watching for movement, you interrogate the corpus directly: one full-text search across news, commentary, press releases and social posts, with speaker-attributed radio and podcast previews alongside. It does retrieval and provenance — deliberately no AI summary layer — so what you build is an evidence pile, not a paraphrase.
01
Search the whole corpus
Every result is an evidence card: outlet, date, highlighted excerpt, and a link to the original. Filter by date range, item kind, outlet, and the people or organisations mentioned. Search is macron-insensitive, so kereru finds kererū. Every view is a plain URL you can share with your team.
02
Collect as you go
One click adds any result to a collection — a persistent, annotated evidence pile for the question you’re working on. Add a note to each item, then review the pile as a list or as a date-ordered timeline to see how the story built.
03
Take the evidence with you
Export any collection as Markdown, CSV, or JSON for your own tooling. Or skip the export: Team accounts get AI assistant integration, so Claude, OpenAI, or any assistant that supports MCP connectors can search the corpus and read your collections directly — you bring the analysis, we bring the sourced evidence.
Watch the whole workflow
Here are all three steps in one short demo: an analyst searches the corpus, builds a case in a collection, then asks Claude about the evidence over the MCP connector.
A note on radio and podcasts. Transcripts of third-party audio are searchable, but previews are deliberately limited: short, speaker-attributed snippets with a link to the original audio, capped per episode for copyright. We help you find the moment; the publisher keeps the work.
Plans
Free for individuals. Team for organisations.
Recommended
Free account
$0
forever
The whole brief — get told when something moves.
- Everything on the site
- Browse all MPs, topics, and sources
- Full dossiers, briefings, and analysis
- Press × discourse lead-lag, co-occurrence, alias chains
- Radio and podcast clip library with speaker attribution
- Watchlists and alerts
- Watchlists for MPs, topics, and sources
- Daily or weekly email digest of what moved
- Anomaly, framing-shift, and lead-lag alerts
- Four weeks of history
- Rewind any topic, MP, or source four weeks back
Team
From
$400
per month
For newsrooms, agencies, and research teams.
- Everything in a free account, for every seat
- Research Explorer: full-corpus evidence search
- Collections with notes, timeline review, and MD/CSV/JSON export
- Unlimited historical archive (vs four weeks)
- Full API access for your own tooling
- AI Assistant integration: query our whole corpus from Claude, OpenAI, or any vendor that supports MCP connectors
- Shared team management and seats
- Priority data updates
Methodology
Designed to be checked, not trusted.
→
Every insight traces back to a source.
Each claim, sparkline, and shift detection links to the underlying records. We do not paraphrase quietly.
→
No predictive scoring or editorial bias.
We do not score politicians or assign sympathies. The system reports what is happening in the conversation, not what it thinks of it.
→
Derived from public records and verifiable sources.
Hansard, Beehive press releases, RSS feeds from 14 outlets, and public discourse on identifiable platforms. No private data, no scraping behind paywalls.
→
LLM output is constrained.
Generated text is capped at two sentences, anchored to specific data points you can click through, and one-click correctable from every page it appears on.