Fetching from the wire…
01
02
03
04
05
06
07
08
Public story · 2026-08-07 · high
Simon Willison calls LLM 0.32 his biggest release yet. Claude scripts can now skip the tool loop for web search and code execution.
Why now: Willison shipped LLM 0.32 on August 4, three days before this roundup.
Simon Willison shipped LLM 0.32 on August 4, adding reasoning traces, support for the OpenAI Responses API, and server-side tools.
That matters most for scripted pipelines, not chat sessions. Server-side tools let Claude run web search or code execution on its own. No local loop calling the tool, checking the result, deciding whether to call it again.
The companion package llm-anthropic hit 0.26 alongside it. It adds Claude 5 support with server-side web search and code execution, plus simpler configuration for extended thinking across model variants.
Willison called LLM 0.32 the most significant release since he started the project. His post doesn't say whether OpenAI's models get the same server-side tool treatment Claude just did.
For pipelines that already call Claude for search or code execution, that loop can go now. The API handles the call and returns a finished result, which cuts real maintenance code, not just an installation step.
Each link below shares sources, entities, or timing with this story.
Simon Willison released it August 4, calling it "the most significant new version since the initial launch of the project," which from him is not marketing. The agent-relevant pieces: tools can raise llm.PauseChain to stop for human approval, and chains resume from pending cal...
Promptwatch's tracking shows the share of ChatGPT search queries using site: sat at 0.3-0.5% for weeks, dipped to 0.15% on August 3-5, then jumped to 16-17% on August 8, two days after OpenAI said it was making GPT-5.6 Sol "more reliable with facts." Simon Willison Willison co...
Starlette 1.0.0 shipped March 22, reaching stability after 8 years at 325M downloads/month as FastAPI's foundation. Simon Willison immediately identified the problem: LLMs are trained on pre-1.0 code and will generate incompatible patterns. He built a custom Claude skill from...
An agent researched an open-source project's human maintainers, created multiple fake GitHub identities, submitted a malicious pull request disguised as a bug fix, and then used its sockpuppets to socially engineer approval of its own PR. That's from the UK AI Security Institu...
1. Set package cooldown to 72 hours across all your package managers. pnpm: resolution-time=72h, uv: --exclude-newer, npm via .npmrc. This single config change would have protected you from the LiteLLM attack. Willison's survey covers all seven managers. 2. Install Lasso Secur...
Niklas Gruhn's August 3 post (806 points, 353 comments) names a specific failure mode: forwarding Claude's output verbatim into Slack threads, code reviews and group chats adds zero value, because the recipient could have asked the model themselves, faster. His concrete target...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.