Read our Mission first — it defines the project vision, audience, and values. Keep the mission in mind when writing articles, building features, or making design decisions.
Objectives
- Build an accessible publication that treats AI as public infrastructure
- Create content that answers: "What does this mean for me? What should I do?"
- Combine education, critical review, and practical guidance
- Neither hype nor doom — pragmatic, equity-focused, systems-thinking
How to Use This Guide
- Articles speak to the reader directly — clear, concrete, no jargon without explanation
- Enter from any article — each piece should stand alone
- Cross-link generously to build the knowledge graph
Tasks
Writing & Editing Articles
Articles live in packages/content/src/ across 8 section directories:
1-explainers/ 2-pitfalls/ 3-paper-reviews/ 4-doc-reviews/
5-the-wire/ 6-policy/ 7-infrastructure/ 8-community/
Voice & Style
- Tone: Direct, concrete, power-aware. "The Economist meets Delia Derbyshire"
- Register: Analytical precision with unexpected texture. Authoritative without being stiff.
- Certainty: Calibrated. Acknowledge uncertainty honestly without hedging into uselessness.
| Do | Don't |
|---|---|
| Be direct and concrete | Use jargon without explaining it |
| Show the mechanism (how something works) | Settle for vague claims ("AI could...") |
| Name power dynamics plainly | Moralize or lecture |
| Show the tension — both the problem AND the possibility, never just one side | Be one-sided — either predictably critical or naively optimistic |
| Show what good looks like alongside what isn't working | Swap criticism for cheerleading — balance is the point |
| Connect ideas across disciplines — help readers see the bigger picture | Hammer one angle without showing how it relates to the larger landscape |
| Use infrastructure and historical metaphors when they earn their place | Force any single metaphor everywhere |
| Respect the reader's intelligence | Assume they have a CS degree |
| Be occasionally wry, human | Be performatively casual or meme-driven |
Content Types
Educational (evergreen, SEO-heavy):
- Explainers — "How it actually works" for non-engineers. Not "what is a neural network" but "how does your bank's fraud detection system decide you're suspicious?"
- Common Pitfalls — One misconception per piece, dismantled carefully.
Review & Analysis (credibility engine):
- Paper Reviews — What they found, why it matters, what they got wrong. 2-person format: technical + contextual reviewer.
- Documentation Reviews — Critical reads of AI product docs, TOS, model cards.
Current Affairs (traffic driver):
- The Wire — Short takes on AI news (300-500 words), published within 48 hours. Always include a "Where This Fits" element — help readers see how the news connects to the bigger landscape (new labs, alternative architectures, policy shifts, governance experiments).
- Monthly Digest — Curated roundup with editorial framing.
Futures & Alternatives (constructive engine):
- What's being built well — Profile projects, labs, governance experiments, and cooperative models that demonstrate what good redistribution looks like. New architectural approaches (world models, open-weight initiatives), public AI infrastructure, community governance — these deserve coverage as much as critique.
Community:
- Contributor Spotlights — Highlight contributors and editorial process.
- Open Questions — Pose unresolved questions, publish best responses.
Article Frontmatter
---
title: Article Title
created: 2026-03-07
modified: 2026-03-07
tags: [explainer, infrastructure]
status: draft | review | published
author: Contributor Name
---
Sources & Citations
Every article must be properly sourced. See docs/EDITORIAL-PLAN.md for full standards.
Minimums by content type:
| Type | Sources | Inline Citations |
|---|---|---|
| Explainers | 4 | 2+ |
| Paper Reviews | 3 + the paper | 3+ |
| Policy/Wire | 3 | 2+ |
Format: End each article with a ## Sources section with full citations and URLs:
## Sources
- Crawford, Kate. *Atlas of AI*. Yale University Press, 2021. [Publisher](https://yalebooks.yale.edu/...)
- Winner, Langdon. "Do Artifacts Have Politics?" *Daedalus* 109 (1980). [JSTOR](https://www.jstor.org/...)
Inline citations: Use narrative or parenthetical — both work:
- "As Kate Crawford documents in Atlas of AI..."
- "AI systems inherit existing inequalities (Benjamin, 2019)."
For policy/news: Always link to primary sources (legislation, statements, filings), not just coverage of them.
Balance: Critical AND Constructive — Never One-Sided
Every article should show the tension, not pick a side:
- Name the problem honestly — concentration, displacement, extraction, inequality are real. Don't soften them away.
- Show the alternatives honestly — cooperatives, public AI, open models, policy experiments exist and some are working. Don't ignore them.
- Never swap one bias for another — replacing one-sided critique with one-sided optimism isn't nuance, it's just a different blindspot. The reader needs both: what's going wrong AND what's going right AND the genuine uncertainty about what happens next.
- Acknowledge complexity — most actors aren't purely villainous or heroic. Companies can do good work and extractive work simultaneously. Policy can be well-intentioned and poorly designed. Show the mess.
Cross-Linking & Connecting Ideas
Cross-links between articles are valuable — they help readers navigate the landscape and see how ideas connect. But they should serve the reader, not promote the publication.
- Link when it deepens understanding — when a reader would naturally want context or background on a concept you're discussing
- Don't carpet-bomb — 8-12 well-placed links per long article is plenty. 20+ links makes every paragraph feel like a landing page ad
- Connect across disciplines — the Sapiens approach. Help readers see how technology connects to economics, to governance, to history, to human flourishing. The value of this publication is in the connections, not just the individual pieces.
Teasing the Landscape
When covering news — new labs, policy shifts, architectural breakthroughs — always help readers see where it fits in the grand scheme. A new open-weight model isn't just a product launch; it's a data point in the story of who controls AI infrastructure. A new lab challenging LLMs isn't just competition; it's architectural diversity that could redistribute competitive advantage. Show the map, not just the territory.