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.

Language Models that Play Chess and Explain Their Moves

Connections

Related to

An LLM Beat NetHack Both involve LLMs achieving expert-level game play in complex environments, with Queen achieving Grandmaster-level chess and this item describing an LLM beating NetHack
Can LLMs Play NetHack? A 2026 Agent Benchmark Both evaluate LLMs playing complex strategy games (chess vs NetHack), probing the boundary of what language models can achieve in game environments requiring deep planning

Supports

Thinking Fast and Slow in AI: the Role of Metacognition Queen's ability to both play at Grandmaster level AND explain its moves is a concrete implementation of fast/slow thinking — the silent encoder handles fast intuitive move selection while language generation provides slow deliberative explanation
Strands Decider: Why Not an Encoder? Queen's architecture uses a silent expert encoder feeding into a language model, directly relevant to the architectural debate about whether decoders or encoders should handle domain expertise in hybrid AI systems
Thinking Before Thinking: Scaling Agentic Inference Through Meta-Reasoning Queen's dual-component design (expert encoder + LLM) parallels meta-reasoning approaches where a model must decide how to allocate reasoning resources before generating output, with the encoder acting as a pre-reasoning stage

StarSkirmish Bench v0.2: LLM StarCraft Bot Benchmark Both benchmark LLMs on classic strategic games (StarCraft vs Chess), evaluating reasoning and decision-making capabilities through game performance rather than traditional NLP tasks
StarSkirmish Bench v0.2: LLM StarCraft Bot Benchmark Both benchmark LLMs on strategic game tasks requiring reasoning and code generation - chess explanation vs. StarCraft bot writing - probing similar cognitive/planning capabilities
WorldTUI: Training a 360M Model as a Terminal Emulator Both demonstrate LLMs learning domain-specific interactive behavior (chess moves vs terminal commands) through prediction rather than rule execution, showing models can internalize procedural interaction patterns
Dessa: Conversation Model Trained on Speech Both models are trained on domain-specific interaction data (chess moves vs. spoken dialogue) to produce models capable of explaining or generating natural language in specialized contexts

Type:	link
First published:	Thursday, 8 October 2026 at 21:55 UTC
Location:	https://arxiv.org/abs/2610.03695v1
URL:	https://mattwood.fyi/i/98dad31a

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