CLAUDE.md - Hacker News Story Selection Guidelines

Overview

This document specifies how to select Hacker News stories for the OHGOD Philosophy website’s problem cards. Each problem card has a flip animation that reveals 4 HN stories on the back.

Requirements for HN Stories

Minimum Engagement

• Each story MUST have at least 100 comments on Hacker News
• Stories with 200+ comments are preferred
• Major announcement posts (GPT-4, Mamba, etc.) typically have 300+ comments

Story Selection Criteria

1. First Story Must Support the Stat
   • The statistic shown on the front of the card must be directly mentioned or supported by the first story
   • Example: If stat is “$100M+”, the first story should discuss GPT-4 training costs of $100M
2. Variety of Sources
   • Avoid multiple stories about the same company incident (e.g., don’t use 3 Samsung stories)
   • Mix different angles: research papers, news, lawsuits, Ask HN discussions
3. Landmark/Viral Stories Preferred
   • Major product announcements (GPT-4, GPT-4o, Mamba)
   • Viral incidents (Samsung data leak)
   • Major lawsuits (NY Times vs OpenAI)
   • Research papers that went viral
4. Recency
   • Prefer stories from 2023-2025
   • Older stories acceptable if they’re landmark discussions

How to Find High-Engagement Stories

Search Strategy

site:news.ycombinator.com [topic] [year]

Verify Comment Count

• Visit the HN link directly
• Check the comment count at the top
• Stories with 100+ comments usually have significant discussion

Popular Story Indicators

• Major tech announcements
• Controversial topics
• Research paper releases
• Corporate scandals/leaks
• Lawsuits involving Big Tech

Current Problem Cards and Their Stories (Updated Dec 2024)

Problem 1: Power Concentrated ($10B)

• Elon Musk Spent $10B on AI Training Hardware in 2024 (Oct 2024)
• AI Models That Cost $1B to Train Are Underway (Jul 2024)
• U.S. Antitrust Inquiries (Jun 2024)
• AI Training Costs Increasing 3x Annually (Aug 2025)

Problem 2: Planet Can’t Afford (900k tons CO₂)

• AI Adoption in US Adds ~900k Tons CO₂ Annually (2025 study)
• Big Tech AI Data Centers Driving Up Bills (Aug 2025)
• AI, but at What Cost? Carbon Footprint Breakdown (Feb 2025)
• Google’s Carbon Emissions Surge 50% (Jul 2024)

Problem 3: Data Isn’t Yours (20M+ conversations)

• Fighting NYT’s Invasion of User Privacy (2025) - 20M+ stored chats
• Why is OpenAI Lying About Data Collection? (2025)
• OpenAI Slams Order to Save All ChatGPT Logs (Jun 2025)
• NY Times Suing OpenAI for Copyright (Jan 2024)

Problem 4: Nobody Knows Inside (0%)

• GPT-4 (Mar 2023) - 0% architecture disclosed
• GPT-4 Details Leaked? (Jul 2023)
• GPT-4o (May 2024)
• GPT-4.5 (Feb 2025)

Problem 5: Illusion of Intelligence (covers sycophancy + hallucination + memorization)

• Sycophancy Is the First LLM ‘Dark Pattern’ (2025)
• Claude Says ‘You’re Absolutely Right!’ About Everything (Aug 2025)
• Hallucination Is Inevitable: Innate LLM Limitation (Feb 2024)
• LLMs Regurgitating Training Data (May 2024) - memorization

Problem 6: Architecture Barely Changes (8 yrs)

• ‘Attention Is All You Need’ Coauthor ‘Sick’ of Transformers (Oct 2025)
• Ask HN: Is Anybody Building Alternative Transformer? (Feb 2025)
• How Has DeepSeek Improved Transformer Architecture? (Jan 2025)
• Mamba: Linear-Time Sequence Modeling (Dec 2023)

Story Format in HTML

story1_url="https://news.ycombinator.com/item?id=XXXXXXXX"
story1_title="Title (max ~50 chars)"
story1_summary="One-line summary (max ~50 chars)"

Updating Stories

When updating stories:

1. Search for high-engagement HN discussions on the topic
2. Verify 100+ comments
3. Ensure first story supports the card’s statistic
4. Maintain variety across the 4 stories
5. Keep titles and summaries concise for card display

Git Commit Rules (MANDATORY)

1. NEVER add “Co-Authored-By:” lines to commits
2. NEVER add “🤖 Generated with [Claude Code]” to commits
3. Keep commit messages simple and descriptive
4. Just describe what changed - no AI attribution footer