Observed arrival · 2026-10-24
Judgment Machines maps the last undo before an AI agent acts
An independent research and publishing project about what systems must establish before taking consequential actions, with cases and practical instruments.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- People designing AI-agent action controls
- Worth noticing
- Its four-stage model moves from proposal through establishment and authorization to a scoped, recorded commitment.
Field notes
The action model separates the agent’s proposal from checks on required properties, authorization, and commitment; the example focuses on the boundary between drafting a message and sending it. The site also frames scope, retries, and evidence as distinct failure patterns, and says the same inquiry applies across aviation, medicine, trade, and software. Book publication details remain pending, while two later volumes are labeled planned.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue