llm 0.32a3
Simon Willison released version 0.32a3 of his LLM command-line tool on June 9, 2026, with the new release almost entirely written by Anthropic's Claude Fable 5 model. The update adds features to both …
Simon Willison released version 0.32a3 of his LLM command-line tool on June 9, 2026, with the new release almost entirely written by Anthropic's Claude Fable 5 model. The update adds features to both …
Simon Willison reverse-engineered the AgentsView tool to set a custom price for the newly released Claude Fable 5 model, which was not yet included in the platform's pricing database. The workaround a…
Simon Willison released datasette-agent-edit 0.1a0, a base plugin for Datasette Agent that provides core text editing tools for agentic editing of files. The plugin implements view, str_replace, and i…
OpenAI introduced Lockdown Mode to block outbound network requests that could exfiltrate sensitive data during the final stage of a prompt injection attack. The feature does not prevent prompt injecti…
Adversa's Q2 2026 AI Risk Quadrant Report, published June 3, found that 98% of production agent systems carry the "lethal trifecta" — access to private data, exposure to untrusted content, and the abi…
An open-source Agentic Product Standard v2.0 has been released, turning three critical production safeguards into enforceable code rather than advisory principles. The standard addresses the "lethal t…
Simon Willison released an alpha package called micropython-wasm that runs Python code in a sandbox using WebAssembly, along with a Datasette Agent plugin called datasette-agent-micropython that uses …
AI enthusiasts face an existential threat from competitors who adopt rapid AI integration, while skeptics face an existential threat from reliability degradation and institutional knowledge loss cause…
Developer Simon Willison wired OpenAI's Codex into Claude Code as a consulting subagent to get adversarial reviews from a different model family, aiming to catch correlated blind spots in reasoning. T…
OpenAI introduced a new memory system for ChatGPT that builds a running synthesis of user interactions, aiming to keep information fresh over time. The system's design reflects a broader concern about…
Uber has implemented a $1,500 monthly token spending limit per AI coding tool for all employees, capping usage of agentic software like Cursor and Claude Code after the company blew its 2026 AI budget…
Uber has capped employee usage of AI coding tools like Cursor and Anthropic's Claude Code at $1,500 in monthly token spending per tool, a company spokesperson confirmed. The limit, implemented in rece…
Postman launched the AI Engineer, an upgrade to its AI-native API platform that uses a Context Graph to map APIs and services, enabling agents to perform tasks like API exploration and system design. …
Simon Willison released a new web tool called Pasted File Editor that automatically converts large text pastes into file attachments, mimicking a feature from Claude AI. The tool, built using Codex de…
Datasette 1.0a32 was released on May 31, 2026, as a minor bugfix update. The release addresses a bug with `INSERT ... RETURNING` queries through the new `/db/-/execute-write` endpoint and resolves sev…
Developers are increasingly writing AI agent skills instead of traditional documentation, as skills offer immediate value and are easier to iterate. The shift raises questions about whether human-cent…
Simon Willison successfully ran Python ASGI applications in a web browser using Pyodide and Service Workers, overcoming a four-year limitation in his Datasette Lite project where JavaScript in `<scrip…
Simon Willison released a new Markdown rendering tool that treats fenced code SVG blocks as rendered images while also offering a code view tab. The tool accepts pasted Markdown or URLs to CORS-enable…
The curl project's security team is facing unprecedented pressure from a flood of AI-assisted security reports, with incoming reports now 4-5 times higher than 2024 levels and averaging more than one …
The term "slop" is increasingly applied to codebases, defined as an LLM-generated codebase that no person or AI understands well enough to work with effectively, where adding features or fixing bugs t…