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.

NESformer: NES CPU Compiled into Qwen3.5 Weights

Connections

Related to

C64 Keyboard Font Both are retro-computing artifacts centered on 6502-era hardware aesthetics and architecture — NESformer emulates the NES CPU (6502-based) while C64 Keyboard Font references the Commodore 64 (also 6502-based)
Language Models that Play Chess and Explain Their Moves Both encode structured rule-following computation (chess moves vs CPU instructions) into language model weights, pushing the boundary of what LLM inference can execute beyond natural language

Supports

WorldTUI: Training a 360M Model as a Terminal Emulator Both demonstrate training/compiling deterministic computational systems (terminal emulator vs CPU emulator) into neural network weights, exploring models as substrate for running conventional software logic
Animations for System Builders: Metastability and Lost Writes Both engage with low-level CPU/hardware semantics (metastability, memory writes vs NES CPU opcodes), exploring how precise computational behavior can be represented and reasoned about in non-traditional contexts

Challenges

Regenerative Software: The Phoenix Architecture NESformer inverts conventional software architecture by making inference weights the execution substrate, challenging the Phoenix Architecture's assumption that software and compute remain separate regenerable layers
Dust: Pretraining Transformers Without Backpropagation NESformer encodes deterministic CPU computation into weights via compilation rather than gradient-based learning, implicitly challenging assumptions about how knowledge/behavior gets embedded in transformer weights

Type:	link
First published:	Friday, 30 October 2026 at 18:37 UTC
Location:	https://github.com/dbreunig/nesformer
URL:	https://mattwood.fyi/i/d99556ba

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