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 →

A Staff Engineer's Guide to Inventing Work

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.

permalink10 · sujithjay.com →

MCP-over-ACP: Stateless MCP Requests

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.

permalink10 · agentclientprotocol.com →
OneRec-Think: In-Text Reasoning for Generative Recommendation

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.

permalink8 · arxiv.org →
Semantic ID-based Generative Retrieval for Podcast Discovery at Spotify

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.

permalink11 · arxiv.org →
PLUM: Adapting Pre-trained Language Models for Industrial-scale Generative Recommendations

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.

permalink8 · arxiv.org →
GenRec: LLM-Native Recommendation System at Netflix

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

permalink11 · netflixtechblog.com →

Engineering a Calibrated Classifier Model

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.

permalink13 · brooker.co.za →
Thinking Fast and Slow in AI: the Role of Metacognition

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.

permalink14 · arxiv.org →
Introducing Ember-1

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.

permalink7 · fireworks.ai →

We're Gonna Need a Lot More Mathematicians

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.

permalink8 · terrytao.wordpress.com →
What Even Is An OS Now?

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."

permalink20 · sockpuppet.org →
Contrastive Language Model (CLM)

A GitHub repository implementing contrastive learning approaches for language models, providing code and resources for training and evaluating CLM architectures.

permalink16 · github.com →
AI-pilling our company: lessons learned

Sierra's engineering team deployed AI agents across the company and discovered that collapsing role-specific agents into a single unified agent was more effective, while also emphasizing the importance of proactive rather than reactive AI assistance.

permalink14 · sierra.ai →

Real-Time Video Generation on Trainium

Explores kernel optimization techniques for achieving real-time video generation on AWS Trainium hardware accelerators, focusing on performance improvements through kernel-centric approaches.

permalink11 · www.amazon.science →

Code Execution with MCP: Building Efficient AI Agents

The Model Context Protocol (MCP) enables AI agents to interact with external tools more efficiently through code execution instead of direct tool calls, reducing token consumption and context window overhead when managing hundreds or thousands of tools.

permalink23 · www.anthropic.com →
Can LLMs Play NetHack? A 2026 Agent Benchmark

Explores whether modern large language models can play NetHack, comparing LLM-based agents against symbolic and neural bot approaches, and presents a custom agent harness with experimental results.

permalink9 · kenforthewin.github.io →
An LLM Beat NetHack

An LLM agent named Astra achieved the first recorded ascension in NetHack, a complex roguelike game requiring long-term planning and execution. The post documents how the agent built its own harness and succeeded where previous attempts failed, exploring the gap between game knowledge and reliable execution in complex environments.

permalink7 · kenforthewin.github.io →

SSRN Research Paper #7496559

An academic research paper indexed on SSRN's paper repository. The specific content and subject matter could not be determined without accessing the full paper.

permalink5 · papers.ssrn.com →
Autonomous AI agents discover reverse transcriptases with tandem repeat arrays

Claude identified the ART family by directly examining its DNA sequences and recognizing the atypical repeats, a behavior attributable to specific Mythos 5 internal signals that respond to repeated DNA.

permalink12 · www-cdn.anthropic.com →
The Plunging Price of Thought

Explores trends in the declining costs of AI computation and cognitive processing, examining how price reductions are reshaping the economics of artificial intelligence capabilities.

permalink26 · epoch.ai →

Tokens Too Cheap to Meter

Explores how rapidly decreasing AI token costs across orders of magnitude will transform machine learning from a premium product to ubiquitous computing infrastructure within 1-3 years, examining improvements in GPUs, models, and the implications of supply and demand-side effects.

permalink13 · jyn.dev →
Unreal Agent: Cost-Efficient AI Agent Harness

Unreal Agent is an agent harness that achieves up to 40% cost savings compared to Codex by managing tool calls asynchronously, allowing agents to respond quickly to users while reducing model overhead without sacrificing performance.

permalink27 · unreallabs.ai →
Claude Opus 5.5 Intelligence & Performance Analysis

Comparison of Claude Opus 5.5 (Adaptive Reasoning) across intelligence metrics, performance benchmarks, and pricing, ranking it among top models with detailed technical specifications and cost analysis.

permalink8 · artificialanalysis.ai →

I Don't Want to Read What You Didn't Write

AI-generated writing has become unreadable and inhumane, lacking context, perspective, and human voice. While AI can help people write better, using it to generate design documents, pull requests, summaries, and personal messages creates exhausting, impersonal content that undermines genuine communication and relationship-building.

permalink9 · blog.colinbreck.com →
MiMo-V2.6 | Xiaomi

MiMo-V2.6 is a Xiaomi product or software version, though specific details about its functionality are not provided in the available content.

permalink6 · mimo.xiaomi.com →