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.

Build custom IoT devices using ESP32 or Raspberry Pi with Muse SDKs to connect displays, buttons, sensors, and actuators for DIY projects.

Via Deepak

Raven is an open-source framework that automatically constructs and evolves modular harnesses for AI agents, enabling them to decompose complex goals, coordinate specialized agents across domains, and improve performance through shared experience. The system demonstrates superior performance on long-horizon, cross-domain tasks compared to existing agent systems.

Strands releases Decider 2B, a compact open source decision model designed for AI agents to make autonomous decisions efficiently.

Cloudflare introduces Clef, an open-source platform featuring decision models and reinforcement learning fine-tuning capabilities for AI applications.

Language models that manage their own context by treating it as an editable file, enabling dynamic context updates and multi-agent systems while outperforming existing context management strategies across various tasks. The approach supports both in-context and parametric learning through natural language steering and reinforcement learning, with optimized serving via suffix cache reuse.

OpenAI and Synopsys announced a partnership to develop GPT-Synopsys, an AI system designed to revolutionize semiconductor chip design using frontier intelligence technology.

Guide to using Claude Code, Codex, or Kiro to connect with an AI agent that helps plan AWS re:Invent 2026 sessions, find relevant talks, and build an optimized conference schedule.

Explores physical modular dashboard systems found in German and Polish museums—used for air traffic control and transit monitoring—as historical design and interaction systems that preceded modern digital interfaces.

Fun.

Ledge is a markdown notebook tool for developers that executes shell commands, code, SQL queries, and AI prompts directly within notes, with persistent shell state and real-time output streaming across macOS, Windows, Linux, and mobile.

This paper introduces agentic meta-reasoning, a control framework that helps AI agents make structured decisions about task execution by balancing exploration, assessment, and resource allocation during long-horizon problem-solving. The approach shows significant improvements on benchmarks testing program reconstruction, abstract reasoning, and proof generation, outperforming direct control baselines as computational budgets increase.

Explores principles for designing evaluations and improving performance against them without overfitting, demonstrating how Claude's build-eval and hillclimb commands automate these practices to help assess and optimize AI applications.

Earendil Engineering explains their reversal on MCP support for Pi, detailing how MCP evolved over time and why integrating it into the core—rather than as an extension—enabled better sandbox functionality and tool composition for modern language models.

Perishable assumptions.

The article argues that MCP servers for browser automation and web scraping are often unnecessarily complex, and that agents can accomplish the same tasks more efficiently using simple Bash commands and code generation instead.

A comprehensive guide tracing the history of text classification techniques from traditional bag-of-words and logistic regression through RNNs, CNNs, and Transformers, examining how modern models like Jev balance generality, speed, and cost for classification tasks.

Explores how staff engineers on platform teams identify and prioritize work by reading signals from systems, users, organizations, and the industry, since platform teams lack traditional product roadmaps and must self-direct their priorities.

Specifies how to transmit stateless MCP (Model Context Protocol) requests over ACP (Agent Client Protocol) infrastructure. Details the protocol extension for integrating MCP functionality within ACP sessions.

A framework that enhances LLM-based recommendation systems with explicit reasoning capabilities through item-textual alignment, reasoning scaffolding, and recommendation-specific reward functions, validated on public benchmarks and deployed at scale on Kuaishou.

This paper describes GLIDE, a production-scale generative recommender system that uses Semantic IDs and large language models to balance user preferences with exploration in podcast discovery. The system addresses challenges in catalog grounding, personalization, and serving latency while achieving significant improvements in non-habitual listening through large-scale A/B testing.

PLUM is a framework that adapts pre-trained large language models for recommendation systems at scale, using semantic IDs for item tokenization, continued pre-training on domain-specific data, and generative retrieval fine-tuning. The approach demonstrates significant improvements over production models on video recommendation tasks and has been deployed to billions of YouTube users.

Netflix's GenRec system leverages large language models to power native recommendations, representing a shift toward LLM-based approaches in content discovery and personalization.

Marc Brooker describes building Hobson, a 2-billion parameter classifier optimized for accuracy and calibration at low latency, by fine-tuning a pre-trained LLM with a pointer head instead of a language modeling head to score multiple choice answers.

AI systems have seen dramatic advancement in recent years, bringing many applications that pervade our everyday life. However, we are still mostly seeing instances of narrow AI: many of these recent developments are typically focused on a very limited set of competencies and goals, e.g., image interpretation, natural language processing, classification, prediction, and many others. Moreover, while these successes can be accredited to improved algorithms and techniques, they are also tightly linked to the availability of huge datasets and computational power. State-of-the-art AI still lacks many capabilities that would naturally be included in a notion of (human) intelligence.

Fireworks Research releases Ember-1, a specialized model that delivers the same quality as Kimi K3 while using 40% fewer tokens, making it more cost-effective for coding and reasoning tasks at scale.

Amit Sahai discusses how AI systems are rapidly producing new mathematical ideas, arguing that the mathematics community must prioritize understanding and human engagement with AI-generated discoveries rather than being displaced by them.

AI is fundamentally changing computing by dissolving boundaries between programmers and users, enabling power users to create custom applications using natural language instead of traditional programming, which raises profound questions about software distribution and the future of operating systems in a world of single-user or dual-user applications.

"Word processor? You might as well have asked me to build a diesel locomotive."