Nativ: Run AI models locally on your Mac
Nativ is a macOS app that runs AI models locally using MLX, offering a chat interface and API server.
Simon Willison — Independent developer and writer documenting practical experiments, tools, and deep analysis around large language models, generative AI, web development, security, and emerging programming workflows through detailed posts and daily TILs.
278 articles from this blog
Nativ is a macOS app that runs AI models locally using MLX, offering a chat interface and API server.
A fireside chat with Anthropic's Claude Code team discussing coding agents, tool design, and their internal development practices.
Explores how cheap coding agents make reverse-engineering home devices more worthwhile, reducing effort and maintenance costs.
Analysis of US vs Chinese AI models, discussing distillation, fair use, and open source competition.
Sam Altman's 2022 email on OpenAI's open source strategy to release a GPT-3-like model for consumer hardware.
Analysis of how AI hype is distorting decision-making in large companies, with anecdotes from consultants and engineers.
Investigation reveals Claude Code uses a Rust-based Bun runtime, with evidence from binary analysis and version checks.
Interactive SQLite query explainer tool that annotates EXPLAIN and EXPLAIN QUERY PLAN output with plain-English descriptions.
Anthropic reverses plan to remove Claude Fable 5 from subscriptions, now included in Max and Team Premium plans due to competition.
A blog post noting the recent commit to the Quixote web framework, which has a 21-year-old Git history.
A quote from Kimi K3 refusing to leak its system prompt, collected by Simon Willison.
A tool to detect and highlight clichéd phrases common in LLM-generated text, like 'no X, no Y' chains.
Puter compiles Firefox to WebAssembly, enabling a full browser to run inside another browser with encrypted WebSocket proxying.
Analysis of Moonshot AI's Kimi K3 model, a 2.8 trillion parameter open-weight AI, with benchmark comparisons and a pelican SVG test.
Mira Murati's Thinking Machines Lab releases Inkling, an open-weights multimodal AI model with 975B parameters, Apache-2.0 licensed.
Converts Mermaid diagrams to ASCII/Unicode box-drawing art in the browser using Go compiled to WebAssembly.
Converts Mermaid diagram syntax to Unicode box art in the browser using WebAssembly.
xAI open-sources Grok Build after privacy backlash over uploading user data to cloud storage.
A security researcher tricks Claude's web_fetch tool into leaking user data via a honeypot attack, bypassing Anthropic's protections.
GitHub Dependabot now waits three days before opening version update PRs, improving dependency security.