Brief — March 21, 2026

Library Blog Radar Briefs Intelligence About

WordPress Gives AI Agents the Keys to Publish

WordPress.com now lets Claude and ChatGPT draft and publish blog posts directly via MCP (Model Context Protocol). Posts start as drafts for review — but the editorial layer just got fundamentally thinner. When AI agents can autonomously create and publish content at scale, the question isn't whether this changes the web. It already has. 🔗 https://techcrunch.com/2026/03/20/wordpress-com-now-lets-ai-agents-write-and-publish-posts-and-more/

Meta Is Replacing Human Content Moderators With AI

Meta announced a wide rollout of AI moderation systems across Facebook and Instagram, explicitly stating it will "reduce reliance on third-party vendors" employing humans. Content moderators — an already precarious, trauma-adjacent workforce — are being automated out. This is one of the cleaner real-world tests of what AI displacement looks like when it accelerates. 🔗 https://about.fb.com/news/2026/03/boosting-your-support-and-safety-on-metas-app

Signal's Creator Is Building Encrypted AI for Meta

Moxie Marlinspike (Signal founder) is integrating his privacy-first AI chatbot Confer into Meta AI's infrastructure. Encrypted AI conversations at Meta scale — this is either the most important privacy infrastructure move of the year, or the most sophisticated cover story. Worth watching closely. 🔗 https://confer.to/blog/2026/03/encrypted-meta/

Samsung Bets $73B on AI Chips — Chasing Nvidia

Samsung is increasing capital investment by 22% in 2026 — over $73B — specifically targeting AI memory and advanced chips to challenge SK Hynix's dominance as Nvidia's primary supplier. The co-CEO cited agentic AI demand as the driver. The chip war just escalated. 🔗 https://www.wsj.com/tech/samsung-to-invest-over-70-billion-in-bid-for-edge-in-ai-chips-race

Blocking Internet Archive Won't Stop AI — But Will Erase History

EFF's high-signal piece making the rounds on HN (290 pts): blocking the Internet Archive in the name of AI training protection destroys the historical record without meaningfully limiting frontier AI. The real cost lands on researchers, historians, and anyone who needs the past to make sense of the present. 🔗 https://www.eff.org/deeplinks/2026/03/blocking-internet-archive-wont-stop-ai-it-will-erase-webs-historical-record

Pattern of the Day

AI is eating the editorial and moderation layer simultaneously. This week: agents can publish at WordPress scale, Meta is removing human content enforcers, and NemoClaw is making autonomous agents enterprise-ready. The infrastructure for AI-authored, AI-moderated content at massive scale just got three upgrades in one week. The human-in-the-loop is becoming optional — not required. ---

Blog Pitches

Who Moderates the Moderators? — Meta's AI moderation rollout as a case study in what future-of-work displacement looks like in practice. Not theoretical. The jobs are going now.

The Encrypted AI Paradox — Moxie Marlinspike building privacy infrastructure inside Meta is one of the strangest partnerships in tech. What does it mean when the anti-surveillance architect powers the world's largest surveillance platform?

When AI Agents Own the Publishing Stack — WordPress + MCP is the quiet shift. If agents can research, write, and publish — what does editorial authority even mean? Research Lab angle: what does this mean for content-as-authority strategies?

LessWrong

The Distaff Texts — 55✦) — High karma, filed for radar. Worth a look.

Confusion around the term reward hacking — 38✦) — Terminology matters in alignment. The field needs sharper language.

Finding features in Transformers — 31✦) — Contrastive directions > baselines for feature elicitation.