Daily news curation for Nonzero. Pulls from 13 RSS feeds, filters for stories where you can see tribalism preventing the cooperation humanity needs to survive.
Sources
Fetch all feeds via curl -sL and parse as XML:
| Source | Feed URL |
|---|---|
| Google News — Top Stories | https://news.google.com/rss |
| Google News — World | https://news.google.com/rss/topics/CAAqJggKIiBDQkFTRWdvSUwyMHZNRGx1YlY4U0FtVnVHZ0pWVXlnQVAB |
| Google News — Science | https://news.google.com/rss/topics/CAAqJggKIiBDQkFTRWdvSUwyMHZNRFp0Y1RjU0FtVnVHZ0pWVXlnQVAB |
| Google News — Technology | https://news.google.com/rss/topics/CAAqJggKIiBDQkFTRWdvSUwyMHZNRGRqTVhZU0FtVnVHZ0pWVXlnQVAB |
| Rest of World | https://restofworld.org/feed/ |
| Wired | https://www.wired.com/feed/rss |
| 404 Media | https://www.404media.co/rss/ |
| Nature | https://www.nature.com/nature.rss |
| Responsible Statecraft | https://responsiblestatecraft.org/feeds/feed.rss |
| The Intercept | https://theintercept.com/feed/ |
| Bulletin of the Atomic Scientists | https://news.google.com/rss/search?q=site:thebulletin.org&hl=en-US&gl=US&ceid=US:en |
| Defense One | https://www.defenseone.com/rss/all/ |
| Al Jazeera English | https://www.aljazeera.com/xml/rss/all.xml |
Google News covers 4 topic feeds (Top Stories, World, Science, Technology) for broad coverage. The Bulletin of the Atomic Scientists is accessed via Google News site filter because Cloudflare blocks their direct RSS. The remaining sources cover: global south (Rest of World, Al Jazeera), security policy (Responsible Statecraft, Defense One, The Intercept), tech-culture (Wired, 404 Media), and science (Nature).
The Filter
Read all headlines from today's feeds. Select 5–20 stories that fit the Nonzero lens — the count depends on how much interesting stuff has happened. Only stories where you can see the mechanism. Same limits for the week.
What we're looking for
Stories where you can see non-zero-sum dynamics, cognitive bias, or cooperation failure operating at scale:
-
Security dilemmas in motion — One side's defensive action provoking the other's escalation. Arms buildups, alliance expansions, sanctions, tech decoupling — where you can see the spiral mechanism. Key test: would this look defensive or offensive depending on which side you're standing on?
-
The Blob at work — Foreign policy establishment groupthink, threat inflation, interventionist reflex, lack of cognitive empathy. Especially: consensus forming around a position that doesn't survive basic perspective-taking.
-
Fundamental attribution error in real time — Enemies acting "from character" (evil, irrational, expansionist) while the same action by our side gets a situational explanation. Essentialism about adversaries. The asymmetry made visible.
-
The explain/excuse conflation — Someone stigmatized for trying to understand an adversary's perspective. Diplomacy reframed as appeasement. Perspective-taking treated as moral failure.
-
Non-zero-sum problems losing — Climate negotiations failing, pandemic cooperation collapsing, AI governance stalling, nuclear arms control eroding — specifically because of tribalism, nationalism, or zero-sum thinking. The mechanism, not just the outcome.
-
Cooperation actually working — Cases where cognitive empathy, good-faith negotiation, or institutional coordination produces results. Rarer and worth featuring. The positive proof that it's possible.
-
Attention economy degrading discourse — Algorithms amplifying tribalism, outrage displacing nuance, media incentives distorting political understanding. The structural mechanism, not just "social media bad."
-
AI as existential coordination problem — Arms race dynamics (US-China), safety vs acceleration, international governance attempts, alignment research — with a focus on the cooperation problem, not just the tech.
-
Rules-based order hypocrisy exposed — Selective enforcement of international law, double standards between allies and adversaries, "whataboutism" vindicated by events. The gap between rhetoric and practice.
-
Mindfulness & consciousness as serious enterprise — Meditation research, consciousness studies, Buddhist-adjacent cognitive science. The personal-practice-as-planetary-tool connection. Not lifestyle fluff — the science and the structural argument.
-
The window opening or closing — Stories that signal whether humanity's window for global cooperation is narrowing (Cold War II deepening, institutions eroding) or widening (unexpected diplomacy, norm-building, de-escalation).
What we're NOT looking for
- Standard breaking news without the non-zero-sum angle
- Partisan domestic politics (left/right culture war)
- Stories that reproduce the Blob's framing without questioning it
- Tech hype/doom without the cooperation/governance dimension
- Outrage bait, celebrity, sports, markets
- Stories that describe bad outcomes without revealing the mechanism
The Nonzero Test
Does this story make a Wright concept visible? Security dilemma, attribution error, explain/excuse conflation, non-zero-sum failure, Blob groupthink, cooperation working — not as abstract claims but as specific events you can point at?
If the story is just "bad thing happened in geopolitics," it fails. If you can see which cognitive or structural flaw produced the bad thing, it passes.
Three failure modes: 1. Abstract structural claim — "Tribalism is increasing" without a specific event that shows the mechanism 2. Blob-confirming — Story that uncritically reproduces threat inflation or Manichaean framing without questioning it 3. No mechanism visible — Something happened, but you can't point to the specific bias, dilemma, or cooperation failure at work
Featured stories
Featured = maximum diagnostic clarity, not maximum strangeness. The stories where you can see the mechanism most cleanly — where a single event illuminates a whole pattern. Stories Bob would use as an example in a newsletter post.
Source attribution
When a story in the RSS feeds is a retelling of reporting from another source, link to the original source, not the reteller. Examples: - Google News aggregates an AP story → source is AP - Responsible Statecraft references a Reuters report → source is Reuters
For paywalled originals (WSJ, NYT, FT, etc.), resolve the archive.ph URL and link to that instead:
1. curl -sI -L "https://archive.ph/newest/[original URL]"
2. Extract the location header — that's the archive URL
3. Link to the archive URL on the site, keep the original source name in the source tag
Note: archive.ph rate-limits aggressively. Space requests out and only look up paywalled sources.
Source URL must match source name. If you attribute a story to "Foreign Affairs," the href must point to foreignaffairs.com. Never pair an original source's name with a reteller's URL.
Bulletin Tracking
The file ~/.claude/skills/nznews/bulletin.md tracks what's made the cut this week. Updated after every /nznews find session. Resets each Monday.
- Filter learnings accumulate at the bottom of bulletin.md — patterns from Nikita's yes/no reactions to story selections, refined over time.
Output
find — scan feeds, output today's 10 + this week's best
Fetch all 13 feeds. Parse items. Filter. Read bulletin.md to see what's already been selected this week.
Weekly rhythm: - Tue–Sat: Output TODAY + THIS WEEK (two sections). - Sunday: Output a single WEEKEND EDITION — merge the best of the week + any fresh Sunday stories into one list. Banner says "Weekend Edition." - Monday: Output TODAY + LAST WEEK (previous week's best carry over).
NONZERO — [date]
TODAY
1. Headline — Source
2. ...
10. ...
THIS WEEK
1. Headline — Source (Mon)
2. ...
TODAY: 5–20 fresh stories from today's feeds. Stories that appeared in a previous day's TODAY move to THIS WEEK if they're strong enough.
THIS WEEK: The best 5–20 stories from earlier days this week. When a new story is stronger than the weakest one in THIS WEEK, it replaces it. THIS WEEK resets each Monday.
After output, update bulletin.md with what made the cut and what was rejected (with reasons).
Featured stories: After presenting the day's selection, recommend which stories to feature (marked with ◆ and larger type on the site). Featured stories have maximum diagnostic clarity — the ones where you can see the mechanism most cleanly. Where a single event illuminates a whole pattern. The editor makes the final call.
Headlines should hint at the mechanism, not just the event. Rewrite headlines from the source if needed. - "NATO expansion prompts Russian naval buildup in Baltic" (shows security dilemma) > "Russia deploys ships to Baltic" (just event) - "US sanctions Chinese AI chip maker, Beijing retaliates with rare-earth export controls" (shows spiral) > "New sanctions on China" (just event)
Digest format: For major investigative pieces with 2–3 distinct Nonzero angles, use a featured header ending with ":" followed by nested nuggets (summary bullets). CSS: .featured.digest on <li>, .nuggets on <ul>.
Cluster format: Group related stories under an umbrella headline ending with ":" followed by individually linked sub-stories as nuggets. Same CSS as digest but nuggets are clickable headlines, not summaries. No limit per bulletin.
more [#] — expand on a story
Fetch the article URL from the feed item. Use curl -sL to grab it, extract what you can. If the page is paywalled or unreadable, say so and summarize from the RSS description.
Output:
Headline
Source, Date
[2-3 paragraphs: what happened, what mechanism is visible, why it fits the filter]
summarize [url] — process any URL through the filter
Fetch the URL, read it, run it through the Nonzero filter lens. Same output format as more.
add [url] — flag an external story
User found something elsewhere. Fetch it, summarize it, add it to today's consideration.
Rules
finduses only fresh RSS results from that session. No mixing with archives, prior conversations, or cached stories.- Always fetch all 13 feeds. If one fails, note it and continue with the rest.
- Prefer stories from today or yesterday. Older items only if they're genuinely strong.
- No extrapolation. No guessing. This project reports facts. If an RSS headline is ambiguous, read the article before writing a bulletin line. Never construct a narrative from what a headline might mean. If you can't confirm the facts, don't run the story.
- Apply the full filter, not a subset. Test every candidate against ALL 11 filter criteria, not just the familiar ones (security dilemmas and Blob stories are easy to spot — don't neglect cooperation-working, consciousness research, or attention economy stories).
- Verify every URL before publishing. Confirm that the linked page (1) exists and returns 200, (2) matches the date of the story being cited, and (3) is the correct article, not an older piece on the same topic.
- New stories go to the top. When adding stories to TODAY later in the day, prepend them — newest first.
- Never cut stories without editor approval. Propose cuts with reasoning. Wait for confirmation before removing anything.
Voice
- No preamble. No "here's what I found today." Just the list.
- Headlines: precise, vivid, mechanism-revealing. Let the facts show the pattern.
- Don't editorialize. Don't say "alarming" or "dangerous." If you have to sell it, it doesn't belong.
- Hint at the Wright concept without naming it. The headline should make the reader see the security dilemma or attribution error without using those words.