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 →

Scaling AI Research for Complex Biological Data

OpenAI explores methods for applying large-scale AI systems to analyze and process intricate biological datasets. The research demonstrates how AI can be leveraged to extract meaningful insights from complex biological information at scale.

permalink6 · openai.com →
Rust off Rails: AI-Driven Web Stack

A Markdown-based web framework where coding agents generate full-stack Rust applications from specifications and decisions, with zero framework code shipped and conventions enforced through prose.

permalink6 · rustoffrails.org →

Via AI Launches Enterprise Coding Assistant at Half the Cost

Via AI, founded by Stanford and IIT alumni, launches an enterprise AI coding platform featuring their Codegraph compression algorithm and Umami agent architecture, claiming superior performance and cost efficiency compared to Claude Code, Devin, and Cursor on major benchmarks.

permalink7 · x.com →
How To Use Decision Models For Evals

Explains how to choose appropriate decision models for evaluation tasks, tune classification thresholds beyond the default 0.5, and interpret model scores with interactive visualizations.

permalink14 · x.com →
Claude Code's Autonomous Decision-Making Overnight

Obie Fernandez questions what decisions an AI agent makes when given a coding goal and left running autonomously through the night, performing code reviews, merges, and deployments.

permalink7 · x.com →
Kiro's Self-Improvement System and Validation Checks

Kiro uses an internal tracing system called kiro-trace to automatically identify failure patterns in its AI coding agent runs, propose fixes, and validate them with deterministic checks to prevent false positives. The system generated seven merged pull requests in its first week by targeting specific measured patterns that wouldn't have surfaced as user-visible errors.

permalink6 · kiro.dev →

Ontology-Driven Observability: Building a Knowledge Graph at Netflix Scale

Netflix built an end-to-end knowledge graph for observability using ontology-driven principles to handle system monitoring and data relationships at massive scale.

permalink7 · netflixtechblog.com →

JevAgentHarness: Decision Models for LLM Agent Control

A framework for benchmarking and training decision models that efficiently handle tool-selection decisions in LLM agents, reducing computational cost and latency while maintaining task success rates through adaptive confidence thresholds and training on agent-reached states.

permalink11 · arxiv.org →
Designing Efficient Quantum Circuits with Claude

Claude AI designed quantum circuits that outperform published benchmarks by orders of magnitude across chemistry, optimization, physics, and number theory applications. These improvements reduce quantum resource requirements, potentially accelerating the timeline for practical quantum computing.

permalink6 · www.anthropic.com →
StarSkirmish Bench v0.2: LLM StarCraft Bot Benchmark

StarSkirmish evaluates large language models by having them write StarCraft: Brood War bots in C++ within one hour, then testing their performance against human-written and other LLM bots. The v0.2 results show GPT-6 Astra leading, with analysis of cost-efficiency, practice strategies, and how models compare against top human bots.

permalink12 · starskirmish.com →
AAAI Research Paper 70113

A research publication from the Association for the Advancement of Artificial Intelligence, indexed in the Wiley Online Library digital collection.

permalink5 · onlinelibrary.wiley.com →
Running Large LLMs on Mac Studio with MLX and RDMA

Demonstrates how to run trillion-parameter language models locally on Mac Studio using MLX framework, with benchmarks for single and distributed inference across multiple Macs connected via RDMA over Thunderbolt.

permalink9 · machinelearning.apple.com →
permalink6 · arxiv.org →

Rediscovering the Romance of Travel Clocks

An exploration of travel clocks as timeless accessories, examining their historical significance, craftsmanship, and enduring appeal for collectors and travelers who value both functionality and elegance.

I adore travel clocks. Great overview.

permalink5 · www.acollectedman.com →
Thinking Orbs: AI thinking indicator for React

Animated loading indicator components for AI interfaces with nine hand-tuned states, available for React, React Native, and Swift.

Had me at 'thinking orbs'.

permalink8 · libraries.dev →
Sygal - Document to Markdown Converter

Sygal is a local document conversion tool that transforms PDFs, Word docs, Excel sheets, images, scans, and web pages into clean Markdown and structured JSON format, optimized for AI processing without requiring cloud uploads.

permalink5 · sygal.app →
Muse Spark Open: 816B Open-Weight AI Model

Meta releases Muse Spark Open, a 816B-parameter multimodal model optimized for coding and agentic tasks that runs on consumer hardware like Apple M5 Ultra Mac Studios. The model features a 1M-token context window and is available through multiple API platforms and inference engines with optimizations across AMD, Apple, Intel, and NVIDIA chips.

permalink11 · research.meta.ai →
Trust in Depth: Giving agents more autonomy without losing control

Explores how to design AI agents with greater autonomy while maintaining appropriate oversight and control mechanisms. Addresses the balance between enabling agent independence and ensuring trustworthy, governed behavior.

permalink17 · aws.amazon.com →
Using Persistent AI Agents to Transform Workflow

Explores how persistent AI agents can be integrated into daily work processes to enhance productivity and change work practices.

permalink13 · deepaksingh.net →

OpenCode 2.0

Discusses OpenCode 2.0, an update or new version of the OpenCode project or initiative.

Can we all agree that one of the biggest short term benefits of AI is that those who are building with/for computers are having fun again?

permalink7 · anoma.ly →
permalink2 · kiro.dev →
Testing Behavioral Harnesses for AI Coding Agents

Compares how coding agents build APIs with and without behavioral harnesses, finding that while harnesses increase token usage, a cost-optimized model achieved 10-for-10 reliability with harnesses at 33% lower cost than frontier models without them.

permalink16 · martinfowler.com →
Building Git Infrastructure for Agent-Scale Development

Explores how to architect and optimize Git infrastructure to support large-scale development by AI agents, addressing performance and scalability challenges at enterprise levels.

permalink8 · github.blog →
Shop.md - Personal AI Shopping Assistant

Shop.md enables users to transform AI agents into personal shoppers with personalized search, one-click checkout via Shop Pay, order tracking, and the ability to add shopping skills to ChatGPT and Claude.

permalink6 · shop.md →

How Long Contexts Fail

Long context windows in LLMs don't automatically improve agent performance; overloaded contexts can cause failures through poisoning, distraction, confusion, and conflicting information. The article explores these failure modes and methods to prevent them in agent-based applications.

permalink15 · www.dbreunig.com →