Two axes, not seven rungs: the diagram
The seven-layer AI mountain redrawn on two axes, capability and permission, with technique as marker fill, the seven rungs as a diagonal, and the promotion gate on the right-hand column.
A one-page diagram that redraws the familiar “seven layers of AI” mountain on the grid the literature keeps returning to. What a system is allowed to do on one axis, with columns for human in the loop, on the loop and out of the loop. What it can do on the other, using the capability bands from DeepMind’s Levels of AGI. How it was built becomes the fill of a marker rather than a rung.
The mountain’s seven rungs appear as one dashed diagonal through the grid. Real systems, from AlphaFold to the Knight Capital order router, sit elsewhere. The bottom right corner, acting alone with little or no intelligence, is where two of the decade’s worst software failures lived, and the mountain has no cell for it. A mint strip on the right-hand edge marks where the promotion gate belongs.
A lineage strip credits the frameworks the grid is built from: Parasuraman, Sheridan and Wickens (2000), Shneiderman (2020), ISO/IEC 22989 (2022), DeepMind’s Levels of AGI (2024) and Anthropic’s working definition of an agent (2025). The full argument and sources are in the companion note.
New material lands week by week
Research, the kernel devlog, news and the papers are all published here first. Get them by email, or follow the feeds.
The newsletter is moving in-house and relaunching soon. Until then, email us to be notified when it opens, or follow the RSS feed.
Or follow the work directly: Research · Kernel devlog · News · CDI assessment