AI Foundationspredict · compress · act

concepts → Architecture

Architecture

Reasoning model

frontieras of 2026-094 connectionsdraft

a model trained to spend more thinking before it answers

Where it sits

The prism has six jobs across and eight layers down. Its primary cell isinfer × L1. Hatched cells cannot exist — a GPU does not learn, an institution does not infer.

What must come first — and what it unlocks

Left to right is reading order, derived from the prerequisite_of edges. Nothing here is hand-ordered: the diagram is the graph.

Reinforcement learningLanguage modelReasoning modelnothing depends on it yet — a leaf in the reading order

Before it: Reinforcement learning · Language model

What kind of thing it is — and what it is made of

It is made of: Chain of thought.

Where to read it

The chapter that introduces it, and any chapter that uses it again.

41Thinking Before Answeringact 9 · The Frontier

Where it comes from

paperScaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model ParametersCharlie Snell, Jaehoon Lee, Kelvin Xu, Aviral Kumar · 2024

Every connection

All 4 edges touching this node, grouped by relation family — the sections above are highlights from this list. Colours match the relation families inthe atlas.

Structure · 1
has as a partChain of thoughtMechanism
Order · 2
requiresLanguage modelArchitecture
Lineage · 1
improves onLanguage modelArchitecture

This page is a projection of one node in src/data/concepts.ts. It has no prose file of its own — 230 declared edges produce all 236 of these pages. Edit an edge and both endpoints change.