Simon Willison reported that instructing Claude Code to delegate coding tasks to lower-power subagents using its own judgment significantly improved efficiency and reduced token usage. The approach, i…
Simon Willison released llm-coding-agent 0.1a0, an experimental coding agent built on his LLM library. The agent provides tools for file editing, command execution, and code search, and is available v…
Simon Willison used DSPy to evaluate and improve the system prompts for Datasette Agent, an AI tool that executes read-only SQL queries. The research identified that including column names in schema l…
Simon Willison released shot-scraper 1.10 with a new video command that records browser sessions as WebM videos from YAML storyboards, a feature that had been blocked for two years until Playwright 1.…
Simon Willison released shot-scraper 1.10 with a new `video` command that records Playwright-driven video demos of web applications from a storyboard YAML file, enabling coding agents to produce visua…
Shot-scraper 1.10, a CLI tool for website screenshots and scraping, was released on June 30, 2026, introducing a new video storyboard feature that allows users to record video demos of their work.…
Trust in systems—whether software, finance, or governance—requires transparency and verifiability, not just intelligence. Citing examples like shot-scraper video for coding agents and OpenAI's bug-fix…
Developer Simon Willison warns against accidentally exposing local apps to the internet, advocating for binding to localhost or private networks like Tailscale by default. He demonstrates how to secur…
Qwen 3.6 27B, a dense local language model from Alibaba's Qwen team, impresses developers with its general intelligence and practical coding abilities, running efficiently on consumer hardware via lla…
Developer Simon Willison built agent-historian, a tool that lets AI coding agents query their own session history, after repeatedly encountering agents that forget solutions from previous sessions. Th…
A new AI model, Ornith-1.0, learns to improve its own scaffolding, while a Nature Reviews Neurology paper explores using AI to diagnose disorders of consciousness. The author argues that the urgent qu…
A hypothetical security incident, CVE-2026-LGTM, documented by Andrew Nesbitt, reveals that a malicious package with embedded instructions to automated reviewers bypassed seven AI-powered security gat…
A veteran developer recounts how using AI coding assistants like Claude Code and OpenAI's Codex led him to believe the tools were limited to boilerplate tasks, but the release of Claude Opus 4.5 in No…
A new paper by Simon Willison, highlighted on Bruce Schneier's blog, reveals that large language models suffer from role confusion, enabling prompt injection attacks. The research shows that models le…
Simon Willison created a GitHub repository that uses AI-generated scripts and GitHub Actions to build a 66MB SQLite database from Mozilla's browser compatibility data, hosted with open CORS headers fo…
Anthropic engineer Thariq Shihipar published a post arguing that HTML outputs improve human-agent workflows over Markdown, releasing 20 self-contained HTML examples that drew 4.4 million views in 16 h…
Anthropic's Claude Code engineering lead Thariq Shihipar argues that HTML is increasingly better than Markdown for keeping humans engaged in agentic workflows, citing richer visualizations and interac…
Researchers Charles Ye, Jasmine Cui, and Dylan Hadfield-Menell found that large language models often ignore explicit role tags like <system> or <user> and instead infer roles from text tone, enabling…
Prompt injection remains the top vulnerability for large language models, according to OWASP's Top 10 for LLM Applications, because it is an architectural property of how LLMs process input rather tha…
Vanta's Staff Software Engineer Noam Rubin outlines the company's agent development principles, emphasizing betting on model intelligence, enabling composition for emergent behaviors, and building str…