Auto mode for Claude Code
Auto mode for Claude Code (https://claude.com/blog/auto-mode) Really interesting new development in Claude Code today as an alternative to --dangerously-skip-permissions: Today, we're introducing
Simon Willison's Weblog
Package Managers Need to Cool Down
Package Managers Need to Cool Down (https://nesbitt.io/2026/03/04/package-managers-need-to-cool-down.html) Today's LiteLLM supply chain attack
Quoting Christopher Mims
I really think "give AI total control of my computer and therefore my entire life" is going to look so foolish in retrospect that everyone who went for this is going to look as dumb as Jimmy Fallon
Malicious litellm_init.pth in litellm 1.82.8 — credential stealer
Malicious litellm_init.pth in litellm 1.82.8 — credential stealer (https://github.com/BerriAI/litellm/issues/24512) The LiteLLM v1.82.8 package published to PyPI was compromised with a
Streaming experts
I wrote about Dan Woods' experiments with streaming experts the other day (https://simonwillison.net/2026/Mar/18/llm-in-a-flash/), the trick where you run larger Mixture-of-Experts models on hardware
datasette-files 0.1a2
Release: datasette-files 0.1a2 (https://github.com/datasette/datasette-files/releases/tag/0.1a2) The most interesting alpha of datasette-files (https://github.com/datasette/datasette-files) yet,
Quoting Neurotica
slop is something that takes more human effort to consume than it took to produce. When my coworker sends me raw Gemini output he’s not expressing his freedom to create, he’s disrespecting the
Quoting David Abram
I have been doing this for years, and the hardest parts of the job were never about typing out code. I have always struggled most with understanding systems, debugging things that made no sense,
Beats now have notes
Last month I added a feature I call beats (https://simonwillison.net/2026/Feb/20/beats/) to this blog, pulling in some of my other content from external sources and including it on the homepage,
Starlette 1.0 skill
Research: Starlette 1.0 skill (https://github.com/simonw/research/tree/main/starlette-1-skill#readme) See Experimenting with Starlette 1.0 with Claude skills
Experimenting with Starlette 1.0 with Claude skills
Starlette 1.0 is out (https://marcelotryle.com/blog/2026/03/22/starlette-10-is-here/)! This is a really big deal. I think Starlette may be the Python framework with the most usage compared to its
Profiling Hacker News users based on their comments
Here's a mildly dystopian prompt I've been experimenting with recently: "Profile this user", accompanied by a copy of their last 1,000 comments on Hacker News. Obtaining those comments is easy. The
Using Git with coding agents
Agentic Engineering Patterns (https://simonwillison.net/guides/agentic-engineering-patterns/) > Git is a key tool for working with coding agents. Keeping code in version control lets us record
Turbo Pascal 3.02A, deconstructed
Turbo Pascal 3.02A, deconstructed (https://tools.simonwillison.net/turbo-pascal-deconstructed) In Things That Turbo Pascal is Smaller Than (https://prog21.dadgum.com/116.html) James Hague lists
Quoting Kimi.ai @Kimi_Moonshot
Congrats to the @cursor_ai (https://x.com/cursor_ai) team on the launch of Composer 2! We are proud to see Kimi-k2.5 provide the foundation. Seeing our model integrated effectively through Cursor's
Thoughts on OpenAI acquiring Astral and uv/ruff/ty
The big news this morning: Astral to join OpenAI (https://astral.sh/blog/openai) (on the Astral blog) and OpenAI to acquire Astral (https://openai.com/index/openai-to-acquire-astral/) (the OpenAI
Autoresearching Apple's "LLM in a Flash" to run Qwen 397B locally
Autoresearching Apple's "LLM in a Flash" to run Qwen 397B locally (https://twitter.com/danveloper/status/2034353876753592372) Here's a fascinating piece of research by Dan Woods, who managed to get
Snowflake Cortex AI Escapes Sandbox and Executes Malware
Snowflake Cortex AI Escapes Sandbox and Executes Malware (https://www.promptarmor.com/resources/snowflake-ai-escapes-sandbox-and-executes-malware) PromptArmor report on a prompt injection attack
Quoting Ken Jin
Great news—we’ve hit our (very modest) performance goals for the CPython JIT over a year early for macOS AArch64, and a few months early for x86_64 Linux. The 3.15 alpha JIT is about 11-12%
GPT-5.4 mini and GPT-5.4 nano, which can describe 76,000 photos for $52
OpenAI today: Introducing GPT‑5.4 mini and nano (https://openai.com/index/introducing-gpt-5-4-mini-and-nano/). These models join GPT-5.4 which was released two weeks ago