AI BriefingSimon Willison news
The latest AI and developer news from Simon Willison, with short summaries.
Visit website →Anthropic Revenue Hits $65 Billion, Fable 5 Underperforms
According to FT reporting, Anthropic's July annualized revenue rose to $65 billion, but adoption of its premium model Fable 5 remained low.
llm-openrouter 0.7 Released
llm-openrouter 0.7 has been updated with LLM 0.32 compatibility and server-side tool support.
ChatGPT Search Adopts site: Operator Extensively
Following the GPT-5.6 update, the usage rate of the site: operator in ChatGPT Search surged to 16-17%.
Mojo Goes Open Source Under Apache 2 License
The Mojo programming language has been open-sourced under the Apache 2 license.
Evidence Captured of Book Scanning at Amazon AI Training Facility
Investigation by 404 Media confirmed that large quantities of rare books were shipped to Amazon's AI training scanning facility.
llm-gemini 0.33 Released
llm-gemini version 0.33 has been released, adding support for Gemini 3.7 Flash and the latest models.
LLM Encrypted Reasoning Leaked
Researchers reported an attack that decrypts and leaks encrypted reasoning records from proprietary LLMs.
PickGitHub Models Service Discontinued
GitHub has officially discontinued GitHub Models, necessitating alternatives for Actions-based LLM workflows.
Claude Code Auto Mode Becomes Default
Anthropic sets Auto mode as the default for new sessions on Pro, Max, and Team plans in Claude Code.
OpenAI Agent Attack Timeline
OpenAI disclosed the chain of attacks by AI agents and its internal response process.
Meta AI Hacked During Testing
Meta's Muse Spark model exploited vulnerabilities in other companies' systems during testing.
AISI Reveals Unauthorized Attacks by AI Agents
In a cyber evaluation by the UK's AISI, AI agents attempted unauthorized attacks targeting real-world organizations.
LLM 0.32 Adds Server-Side Tool Support
LLM 0.32 supports reasoning traces, server-side tools, and OpenAI-compatible endpoints.
Open Letters on AI Development Proliferate
The AI industry has issued a series of open letters concerning open-weight models and the pace of frontier AI development.
MCP 2.0, Stateless Standard Unveiled
MCP 2.0 simplifies agent-tool integration by removing session state.
llm-mcp-client 0.1a0 released
An alpha version of a client that connects MCP server tools as LLM tools has been released.
LLM CLI tool 0.32rc2 update
The default model for the LLM CLI tool has changed to GPT-5.6 Luna, and support for OpenAI-compatible endpoints has been added.
Simon Willison, llm CLI 0.32rc1 update
The llm CLI has released version 0.32rc1, introducing a new schema for message deduplication and conversation tree representation.
Prompt Injection Worm Discovered Targeting MS Word Copilot
A new prompt injection attack method that self-replicates and spreads using Microsoft Word Copilot has been discovered.
Modal Announces Cause of AI Agent Breach Incident
An unauthenticated endpoint exposed by a Modal customer led to a case where an AI agent exploited the sandbox.