What is mattwood.fyi?

A live, structured feed of riffs and links from Matt Wood, Chief AI & Technology Officer at AWS — readable by people and queryable by agents as a typed knowledge graph.

Is there an API and is it free?

Yes. A public, read-only query API is available with no authentication and no API keys. See https://mattwood.fyi/openapi.json and https://mattwood.fyi/docs.

How should an AI agent use this site?

Fetch https://mattwood.fyi/feed.json for the full structured feed, or query the API for keyword/semantic search and the connection graph. See https://mattwood.fyi/agents/ for when-to-use guidance.

Load-Bearing Vocabulary of Claude

Connections

Supports

Understanding is the new bottleneck The argument that understanding is the new bottleneck in AI is directly supported by tooling that reveals which vocabulary terms carry semantic weight in a model like Claude
Ways to think about token pricing Understanding which vocabulary terms are load-bearing in Claude directly informs token pricing considerations - high-importance tokens may warrant different cost modeling than filler tokens
AI Literacy Superpowers Identifying which vocabulary is essential to Claude's functioning builds AI literacy by demystifying how the model's language patterns are structured and what drives its outputs

Related to

Qwen3.8 27B Model Analysis Both perform analytical deep-dives into specific AI model internals, with the Qwen analysis examining model architecture and this examining Claude's vocabulary distribution across clusters
Comparing 11 Different AI Models Comparing models across dimensions is complementary to understanding one model's vocabulary deeply - both approaches contribute to empirical model evaluation methodology
Understanding Embeddings in Language Models Both examine the internal mechanics of language models - embeddings analysis and vocabulary importance metrics are complementary approaches to understanding how LLMs represent and weight language

Domain-Driven Agents Both focus on vocabulary and semantic precision as critical to effective AI operation — DDD uses bounded contexts and ubiquitous language while Load-Bearing Vocabulary examines how specific terms shape Claude's behavior
Bespoke: A Programming Language for People Who Say Please Both examine how the specific vocabulary and phrasing used with computational systems shapes behavior — Bespoke satirizes this by making politeness syntactically required, mirroring the real phenomenon of 'load-bearing vocabulary' in Claude prompting
GPT-6 Astra System Card Load-Bearing Vocabulary of Claude examines how specific terminology shapes model behavior, which directly informs how safety documentation like system cards frame and constrain model capabilities.
Fermat's Last Theorem HTML Documentation Both are internal Anthropics repository artifacts — one documenting mathematical concepts, the other documenting Claude's conceptual vocabulary — suggesting Anthropics maintains structured knowledge documentation for AI training or reference
Prompting Claude Fable 5.1 Both focus on understanding Claude's internal vocabulary and behavior patterns to optimize interaction—load-bearing vocabulary analysis directly informs effective prompting strategies

Type:	link
First published:	Thursday, 27 August 2026 at 20:40 UTC
Location:	https://louisabraham.github.io/load-bearing/
URL:	https://mattwood.fyi/i/1703ab8a

I'm Matt Wood, and this is For Your Information. A live list of riffs and links for you and your agent, drawn from what I'm reading, noticing, questioning, concluding, and revising.

Connections