Articles from simonwillison.net

The author learned to let Claude Code's Fable model use its own judgement to delegate coding tasks to lower-power models, improving efficiency and token usage.

Simon Willison introduces shot-scraper video, a tool that records video demos from a YAML storyboard using Playwright, enabling coding agents to demonstrate their work.

Claude Fable 5 is a large, slow, expensive frontier model that demonstrates deep knowledge and strong coding ability across complex multi-step tasks.

Anthropic and OpenAI have found product-market fit with coding agents like Claude Code and Codex, driving enterprise spending so high that both companies switched to API-based billing and are approaching profitability.

Pope Leo XIV's encyclical Magnifica Humanitas offers clear ethical guidance on AI, emphasizing human dignity, accountability, and the dangers of unregulated data ownership.

Google released Gemini 3.5 Flash at a 3x-6x price increase, yet deploys it across Search and Gemini app, testing price tolerance.

An interactive playground lets users try experimental Redis array commands via WASM, including ARGREP for server-side regex matching.

Opus 4.7's system prompt adds child safety tags, a tool-search mechanism, anti-verbosity instructions, and drops the Trump presidency clarification from the 4.6 prompt.

Claude Code's new auto mode uses a Claude Sonnet 4.6 classifier to grant permissions based on intent, with configurable default allow and soft-deny filters.

OpenAI acquires Astral, the company behind uv, ruff, and ty, to accelerate Codex development while promising continued open-source support.

OpenAI Codex launched subagents in GA, enabling custom agents via TOML, and catalogues identical features across six other coding platforms.

Shopify CEO Tobias Lütke used the autoresearch pattern with a coding agent to achieve 53% faster parse+render and 61% fewer allocations in the Liquid template engine.

Coding agents can improve code quality by handling refactoring, enabling exploratory prototyping, and compounding small quality improvements over time.

Current LLM coding agents no longer bias toward popular languages; they handle niche or new tools well via long context and documentation.

Alibaba's Qwen team lead Junyang Lin and several other key researchers resigned, potentially losing the team behind the impressive open-weight Qwen 3.5 models.

Simon Willison describes creating a custom macOS presentation app using AI-assisted vibe coding in Swift, which sequences URLs as slides and includes a phone-based remote control via Tailscale.

Simon Willison's guide outlines patterns and principles for effectively using coding agents like Claude Code and OpenAI Codex to improve code quality and developer productivity.

Test-first development (red/green TDD) significantly improves coding agent reliability by ensuring tests fail before implementation, reducing unnecessary or broken code.

Coding agents make writing code nearly free, forcing engineers to reorient habits from cost-saving to quality assurance like testing, documentation, and maintainability.

Simon Willison introduces Showboat and Rodney, CLI tools for coding agents to create Markdown demos and browser automation scripts to verify their work.

OpenAI released a macOS app for its Codex coding agent, adding a UI, skills, and automations, and reporting over a million developers have used it.

Moltbook is a social network for AI agents built on OpenClaw, showcasing both useful automation and severe security risks from prompt injection.

A single Codex CLI agent built a basic HTML+CSS browser in three days with 20,000 lines of Rust, challenging the need for large multi-agent projects.

ChatGPT's container sandbox now supports bash, Node.js, pip and npm installs, and file downloads from URLs, totaling a major undocumented upgrade.

Claude Code's TUI is built like a game engine, rendering each frame via a React scene graph pipeline within a 16ms budget.

Claude Cowork wraps Claude Code in a sandboxed UI for non-developers, offering general agent capabilities but raising prompt injection concerns.

Simon Willison predicts LLMs will write undeniably good code in 2026, sandboxing will be solved, and a coding agent security disaster is imminent.

Tailwind CSS creator Adam Wathan says AI caused 75% of engineering layoffs, an 80% revenue drop, and 40% less doc traffic despite growing popularity.

The 2025 LLM landscape was defined by reasoning models, coding agents, Chinese open weight models, prompt-driven image editing, CLI tools, and OpenAI losing its lead.

Software engineers must prove code works through manual and automated testing, shifting accountability away from AI agents to the human submitting the PR.

OpenAI has quietly adopted Anthropic's skills format, now available in ChatGPT's Code Interpreter and Codex CLI, using markdown-based skill folders for file operations and code generation.

Simon Willison shares patterns for building useful single-file HTML tools with no build step, using LLMs to generate them.

Nano Banana Pro delivers best-in-class image generation with precise instruction following, legible text rendering, and SynthID watermarking.

Google released Antigravity, an agentic IDE and VS Code fork with three surfaces—agent dashboard, editor, and browser extension—for building apps with Gemini models.