Top story
TOP STORY: Anthropic raises $65B Series H at a $965B valuation. The company also reported $47B in annualized run-rate revenue, signaling an unprecedented scale of enterprise AI adoption. Anthropic Funding Announcement | Simon Willison's Take
Industry
Claude Opus 4.8. Anthropic releases a frontier-class model upgrade described as "a modest but tangible improvement" over its predecessor. Anthropic | Simon Willison
OpenAI's Frontier Governance Framework. OpenAI publishes a formal framework outlining how it intends to govern development and deployment of its most powerful models. OpenAI
Anthropic Opens Milan Office. Anthropic expands its European footprint to support Italian enterprise customers, researchers, and developers. Anthropic
SF Startup Accused of Secretly Testing Robots in Airbnbs. A lawsuit alleges a San Francisco robotics startup used rented Airbnb properties as undisclosed test environments, causing property damage. SF Standard
Research
Various LLM Smells. A catalogued guide to common anti-patterns and failure modes observed when working with large language models. shvbsle.in
Discovering Cooperative Pipelines: Autoresearch for Sequential Social Dilemmas. Researchers use automated research methods to find multi-agent pipelines that encourage cooperation in sequential social dilemma settings. HuggingFace
Why Larger Models Learn More: Effects of Capacity, Interference, and Rare-Task Retention. A new paper explains mechanistically why scaling model size improves performance, focusing on interference and rare-task memory. HuggingFace
Verifiable Rewards Beyond Math and Code. Proposes lightweight corpus-grounded process supervision to extend verifiable reward signals to factual question answering tasks. HuggingFace
Tools
Ollama. Now supports Kimi-K2.5, GLM-5, MiniMax, DeepSeek, and other leading open models for local inference. GitHub
Ktx, Open-Source Executable Context Layer for Data Agents. A new open-source tool provides a structured, executable context layer designed to improve data agent reliability. GitHub
llm-anthropic 0.25.1. Simon Willison releases an updated plugin adding Claude Opus 4.8 support to his LLM command-line tool. Simon Willison
Dify. Production-ready open-source platform for building and deploying agentic workflows hits 143K GitHub stars. GitHub
Community
Continue? Y/N: A 60-Second Game About AI Agent Permission Fatigue. A browser-based game satirizes the exhausting approval loops modern AI agents impose on users. llmgame.scalex.dev
How Endava Builds an Agentic Organization with Codex. OpenAI profiles how consulting firm Endava is restructuring its engineering workflows around OpenAI's Codex agent. OpenAI
MUFG Aims to Become AI-Native with OpenAI. Japanese banking giant MUFG announces a deep partnership with OpenAI to embed AI across its core operations. OpenAI