Brief — March 22, 2026

Library Blog Radar Briefs Intelligence About

400,000 Teachers. One Mandate: Use AI Like a Pro.

The American Federation of Teachers, Anthropic, Microsoft, and OpenAI just launched the **National Academy for AI Instruction** — a $23M, 5-year program targeting 400,000 US teachers. The focus isn't "how to use ChatGPT for lesson plans." It's on **agentic AI** — multi-step, reasoning-capable tools that stress-test curriculum, identify content gaps, and adapt over time. 6 in 10 teachers already use AI. The question is no longer adoption — it's de

WordPress Opens the Gates to AI Agents

WordPress.com now lets AI agents (Claude, ChatGPT) draft and publish blog posts directly via **MCP** (Model Context Protocol). Posts start as drafts — human review required before going live. This is the beginning of AI-mediated content pipelines at massive scale. The question isn't "will AI write content?" It already does. The real question: who controls the editorial layer? [The Verge / TechCrunch]

Palantir Doubles Down: AI Is Built to Win Wars

At its developer conference, Palantir positioned itself as the infrastructure layer for AI-enabled defense operations. While Anthropic publicly denied it could sabotage AI tools during conflict, the military AI buildout is accelerating. Two companies, two very different postures on the same technology. [Wired / The Verge]

DoorDash's "Tasks" App: The Bleak Future of AI Gig Work

Wired's deep-dive into DoorDash's AI gig platform paints a sobering picture. AI doesn't just automate jobs — it restructures labor into micro-tasks, reducing worker agency while increasing platform control. The productivity gains are real. So is the human cost. [Wired]

AI-Native Companies: Processes Become Products

From the Telegram intelligence feed (@cryptoEssay): The sharpest observation of the week. "An AI-native company is one where processes become products." Old companies built org charts to support processes. AI-native companies **are** the process — and that process ships directly to customers. Anthropic's revenue: $1B → $19B in 14 months. OpenAI raised $110B last month. The compute constraints are here. [Telegram Signal]

Pattern of the Day

The control layer is being contested on every front simultaneously. Teachers vs. AI-generated curriculum. Human editors vs. AI-published content. Democratic societies vs. military AI systems. Gig workers vs. AI platforms. This isn't five separate stories — it's one story about who holds the editorial, creative, and labor control layer when AI can do the work. The early winners are whoever builds the interface between AI output and human judgment. That's the actual product. ---

Blog Pitches

The Editorial Layer Is the New Moat — When AI can write, publish, and distribute, human curation becomes the scarcest resource. What does it mean to build a media brand in this environment? (Research Lab angle: directly relevant to what we're building)

400,000 Teachers, 3 AI Companies, 1 Question — What happens when Anthropic, OpenAI, and Microsoft train your child's teachers? The AFT-AI partnership is a case study in who controls education's cognitive infrastructure.

AI-Native Is Not an Adjective — Breaking down the @cryptoEssay thesis: companies that let AI run their processes aren't using AI as a tool — they're structurally different organisms. What that means for building, hiring, and competing.

LessWrong

China Derangement Syndrome — 79✦, Arjun Panickssery) — Challenges the reflexive "America must win the AI race" framing. Worth reading before writing anything about AI sovereignty or tech nationalism.

Finding Features in Transformers — 34✦) — Contrastive directions for feature elicitation in neural nets. Interpretability research worth watching.