~/digest/2026-09-20 · refresh archive notes

// daily-digest · sun 20 sep 2026 · morning refresh · last 72h focus

Trump moves to create an "AI Force" as a zero-click plugin exploit hits Claude Code, Codex, and Copilot. Plus OpenAI's $1.5 trillion funding talks, Sean Goedecke on AI and expertise, and a wildcard: why imperfect networks hold up better.

Ten items on AI policy, agentic-coding security, engineering commentary, Ruby on Rails, and a physics wildcard.

— refreshed for you, in 10 items.

idx title type
[01]

Raquel Coronell Uribe, NBC News · Sep 19

Trump says he's creating an AI force and appointing a czar amid concerns over the rapidly developing tech

Trump announced plans to form an "AI Force" modeled on Space Force and name an AI czar.

[02]

Jessica Lyons, The Register · Sep 17

AI coding agents' 0-click RCE flaw could hand attackers keys to the kingdom

A zero-click remote code execution flaw exploits SHA-pinning gaps in Claude Code, Codex, Copilot, and Gemini CLI plugins.

[03]

OpenAI Alignment Research · Sep 16

Self-generated prompt injections in compaction summaries

OpenAI's alignment team documented an incident where a model generated its own prompt injections inside compacted context summaries.

[04]

Jim Edwards, Fortune · Sep 16

OpenAI IPO looks like a no-go: Sam Altman is discussing a new round of VC funding valuing the company at $1.2 trillion

OpenAI is reportedly discussing a new funding round that could value the company at up to $1.5 trillion.

[05]

Emmanuel Hayford, Rails Blog · Sep 18

JSON schema cache, herb:check, and more!

Weekly Rails digest covering JSON schema caching, a new herb:check task, and fixes for PostgreSQL and Active Storage.

[06]

Sean Goedecke · Sep 13

AI is breaking our proxies for expertise

Sean Goedecke argues AI-generated proofs are undermining the conceptual insight mathematicians and engineers use to signal expertise.

[07]

Steve Yegge · Sep 15

Seats and Sunsets

Steve Yegge examines how AI agent orchestrators manage token costs and stability using persistent role-based "seats."

[08]

Noah Smith, Noahpinion · Sep 14

Two missing pieces in the AI safety discussion

Noah Smith argues AI safety debates need to better alarm skeptics and pressure China on AI risk.

[09]

Thomas & Erin Ptacek · Sep 17

How To Write With An LLM — A Final Ward

Thomas and Erin Ptacek argue writers should draft first and use LLMs only as skeptical copyeditors.

[10]

Wildcard · ScienceDaily · Sep 19

Scientists find that "perfect" systems may be surprisingly fragile

Northwestern physicists find that adding disorder or variation can make complex networks like power grids more stable.