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.

Compression is prediction

Connections

Supports

Ten Advances In Mathematics Compression-prediction duality has deep mathematical roots (Kolmogorov complexity, Solomonoff induction, MDL principle) — this connects to advances in mathematics where information theory intersects with learning theory
Ways to think about token pricing Understanding compression as prediction directly informs token pricing economics — if tokens represent compressed predictions, the cost structure of inference has deeper theoretical grounding than simple throughput metrics
Neuronpedia Jacobian Lens: Interactive J-Space Explorer for Qwen3.6-27B Mechanistic interpretability work like Jacobian Lens analysis implicitly relies on the compression-prediction equivalence — understanding how LLMs compress training data into weights explains what features like those in J-Space actually represent

Related to

11–16× Faster LLM Inference with llama.cpp Faster LLM inference via llama.cpp optimizations relates to compression-prediction: understanding that LLMs are fundamentally predictive compressors informs where computational bottlenecks lie in token generation
LLMs can't jump Both probe fundamental limitations of LLMs — 'LLMs can't jump' examines what models fail to generalize, while compression-as-prediction reveals what LLMs are actually doing when they model sequences, making them complementary perspectives on LLM capabilities

Challenges

Intelligence is not the main bottleneck 'Intelligence is not the main bottleneck' argues against pure capability scaling, but compression-as-prediction suggests intelligence itself may be reducible to compression efficiency — directly challenging claims that intelligence is separable from prediction/compression

Supported by

Understanding Embeddings in Language Models Embeddings fundamentally encode compression and prediction — the piece on 'compression is prediction' directly relates to how dense numerical representations capture semantic meaning through learned statistical patterns
Dynamic Programming: Unifying Principle Across Algorithms Compression-as-prediction is deeply connected to DP: Bellman's principle of optimality underlies both sequence prediction and data compression algorithms, sharing the same recursive substructure logic

Type:	link
First published:	Tuesday, 11 August 2026 at 22:57 UTC
Location:	https://ngrok.com/blog/compression-is-prediction
URL:	https://mattwood.fyi/i/0f0fb60b

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