mattwood.fyi

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.

Links indicate relevance, not agreement. How to use this site →

Your Game Is A Harness

Explores how to design games to be more legible to AI agents and accelerate agentic iteration in game development, covering practical approaches to integrating LLMs like Claude into real game development workflows beyond one-off demos.

permalink10 · kylekukshtel.com →
Gemini AI Models

Google DeepMind's Gemini is a next-generation AI system for building intelligent agents and creating multimodal content, with specialized variants for images, audio, video, and robotics applications.

permalink7 · deepmind.google →

Developing Provably Correct Rust Code with Verus

Explores using Verus, a verification tool, to write Rust code with formal correctness guarantees, enabling developers to prove program properties and prevent bugs at compile time.

permalink9 · www.amazon.science →
Coding Agent Continual Learning: Analysis of ARC-AGI-3 Problem Solving

A study analyzing how a coding agent learns across abstract reasoning tasks by examining traces of its thoughts preserved in Python scripts and text notes, finding that agents selectively rewrite and abandon knowledge at task boundaries rather than reuse code across tasks.

permalink21 · arxiv.org →
Pine Computer: Cloud AI Computer Platform

Pine Computer is a cloud-based computer platform designed specifically for AI agents to autonomously complete complex tasks across multiple applications and websites, offering 2-5x faster performance and 1/25 the model cost compared to traditional AI agent setups.

permalink8 · pinecomputer.io →
Kevin Kelly's 2016 AI Predictions, 10 Years Later

Examines predictions from Kevin Kelly's 'The Inevitable' about AI and technology trends through 2026, analyzing how his forecast that Google's main product would become AI rather than search has evolved a decade later.

permalink6 · taylor.town →
Thinking Will Become a Hobby

AI systems are surpassing humans in cognitive tasks due to superior performance, lower costs, and 24/7 availability, leading to a future where knowledge work diminishes and human thought becomes a leisure pursuit rather than necessity.

permalink8 · www.spinellis.gr →
permalink6 · z.ai →
Bag of Decisions Reranker

An experimental search reranking technique that replaces single relevance queries with LLM-generated question rubrics, scoring documents by summing probability scores across multiple yes/no questions tailored to each query.

p(yes) is a great mental model for decision models.

permalink10 · softwaredoug.com →
Muse Wonder: Meta's Frontier AI Model for Personal Agents

Meta introduces Muse Wonder and Muse Wonder Full, frontier AI models designed for personal and professional agentic tasks with state-of-the-art performance on complex work and everyday life scenarios.

permalink11 · research.meta.ai →

Research Article S2542435126003624

A scientific research article from ScienceDirect. Without access to the page content, the specific topic and findings cannot be determined.

permalink5 · www.sciencedirect.com →

The Micro-machine Universe

Explores how successful manufacturing systems build one type of machine at massive scale, using semiconductor fabrication as a model, and redefines what constitutes a machine across seven physical domains—mechanical, fluidic, thermal, electromagnetic, optical, charged-particle, chemical, and electrical—arguing that computers are machines that transduce, shape, and direct flows of energy or matter.

permalink6 · www.atomicmachines.com →
Codex Cloud Gains Tailscale Support

Codex Cloud now supports Tailscale integration, enabling secure network connectivity through Tailscale's VPN platform. The announcement highlights how Tailscale's infrastructure can be leveraged for enterprise-grade secure access to cloud services.

permalink8 · tailscale.com →

Claude Code v2.1.295 OSC 7501 Protocol Support

Mitchell Hashimoto announces that Claude Code v2.1.295 now supports the OSC 7501 protocol for program status, enabling downstream applications to benefit from Anthropic's rapid implementation.

permalink7 · x.com →
TerrainSR: Terrain Super-Resolution Model

A diffusion-based super-resolution model for geospatial elevation data, designed to enhance terrain and elevation maps with support for PyTorch and ONNX formats.

permalink10 · huggingface.co →
The Dawn of the Age of the Exoskeleton

Explores the emerging technology of exoskeletons and their potential to transform human capabilities, work, and society as they transition from science fiction to practical applications.

permalink6 · theconversation.com →
Language Models that Play Chess and Explain Their Moves

Describes Queen, a 4B-parameter chess-language model that combines a silent expert chess encoder with an instruction-tuned language model to play at Grandmaster level while generating coherent explanations for its moves, trained via iterative distillation and reaching 2697 Elo rating.

permalink11 · arxiv.org →
Yes, and... Programming's Future with AI

Carson Gross discusses whether computer programming remains a viable career given AI tools, arguing that fundamental problem-solving skills remain valuable while warning junior programmers against over-relying on AI code generation without developing core understanding.

permalink8 · htmx.org →
Virtual Biology Initiative: $500M for AI-powered biology

A major funding initiative allocating $500 million to advance AI-powered approaches in biological research and discovery.

permalink9 · biohub.org →
Global $1.8B Commitment to AI-Ready Biological Data

An announcement of a significant global financial commitment to make biological data accessible and optimized for artificial intelligence applications through expanded virtual biology initiatives.

permalink9 · biohub.org →
OSC 7501: A Terminal Protocol for Program Status

Introduces a new terminal escape sequence specification that allows programs to communicate their status (idle, working, waiting, finished, or failed) to terminals and other tools, addressing limitations in existing approaches especially for long-running processes and coding agents.

permalink9 · mitchellh.com →
DuckLake: Lakehouse Format Built on SQL

DuckLake is a lakehouse format that combines SQL databases as catalogs with Parquet file storage, offering ACID transactions, snapshots, and time travel queries without vendor lock-in or complex infrastructure.

permalink7 · ducklake.select →
UNREAL: Unifying Retrieval and Long-Context with a Single Model

UNREAL is a model-native evidence selection framework that unifies corpus retrieval and long-context inference by adding minimal trainable parameters to frozen LLMs. It significantly improves recall on retrieval benchmarks and accuracy on long-context tasks while reducing computational costs.

permalink9 · arxiv.org →
An App Can Be a Home-Cooked Meal

Robin Sloan reflects on building BoopSnoop, a simple messaging app for his family, and explores why software can be valuable even with a tiny user base, drawing parallels to the defunct Tapstack app that prioritized ambient presence over practical communication.

permalink10 · www.robinsloan.com →
Asking for What You Want: A Key 21st Century Skill

AI agents enable a new approach to personal software development where users simply describe what they want rather than navigating complex APIs, fundamentally reframing how computing and programming work. Robin Sloan explores how AI-assisted app building represents a natural evolution of personal software, using his experience creating a custom notes app.

permalink7 · www.robinsloan.com →