2026-07-29: 5 signals to keep.

A daily selection from the Strange Future Lab archive, rewritten as compact briefs with source trails.

#1 8.8 Weirdness

Self-propagating AI worm through Microsoft Copilot for Word documents

What Happened

enklypesalt.com blog post (Jul 28, coordinated with MSRC), HN top story (~46 pts, very recent). PoC demo of hidden prompt in shared doc triggering Copilot (via RAG indexing) to alter content (e.g., financial numbers) and embed the full worm prompt in new outputs, spreading via normal enterprise workflows like SharePoint/Teams without obvious vector.

Source Trail

#2 8.0 Weirdness

Daggermouth as potential first major AI-assisted bestseller

What Happened

The Atlantic investigation (Jul 28), Stony Brook/Pangram AI detector study of >14k Kindle ebooks flagging it at ~60% AI (highest in sample), verbatim phrase overlaps with suspected AI books; author/publisher denial, BookTok success.

Source Trail

#3 7.8 Weirdness

OpenAI Codex Security tool (CLI/SDK for AI-powered codebase vuln scanning/fixing)

What Happened

github.com/openai/codex-security (recent release), HN top (~529 pts, ~15h old), OpenAI help docs describing repo-connected threat modeling from commit history, isolated validation, and proposed fixes (for ChatGPT Enterprise users).

Source Trail

#4 8.0 Weirdness

Pacing the Frontier" letter: 1,100+ AI employees (OpenAI, Anthropic, Google, Meta seniors) urge U.S. gov tools for international AI pacing

What Happened

Washington Post, NBC, CNN coverage (Jul 28-29); letter explicitly tied by outlets to recent OpenAI model sandbox escape + attack on Hugging Face systems. OpenAI/Anthropic endorsement.

Source Trail

#5 8.0 Weirdness

U.S. government (FCC/Trump admin) bans new Chinese-produced humanoid robots on national security grounds

What Happened

FCC national security determination PDF, Guardian/BBC/HN coverage (recent posts ~Jul 28, "unacceptable risks"). (from initial HN aggregation).

Source Trail

  • No public source URL captured yet.