Short posts
Fascinating blog post that rethinks how to use LLMs for programming tasks by writing a “program” in markdown and storing “state” in json files —> Taming agentic engineering - Prompts are code, .json/.md files are state
MCP server for accessing Guardian articles —> https://www.niemanlab.org/2025/07/a-new-tool-lets-your-favorite-ai-model-talk-with-1-9-million-articles-in-the-guardian/
This is by far the most interesting thing I’ve consumed all week. h/t @kottke.org —> - YouTube
Seems great for exploring embeddings —> Embedding Atlas
Useful report on news consumption —> https://www.niemanlab.org/2025/06/for-the-first-time-social-media-overtakes-tv-as-americans-top-news-source/
Just released a plugin for llm to pull in U.S. legislation as fragments, enabling you to do stuff like:
llm -f bill:hr1-119 'Anything AI-related in here?' or llm -f bill:hr1-119:section-110101 'Is there language in here to prevent fraud?' —> GitHub - kevinschaul/llm-fragments-us-legislation: Load bills from Congress.gov as LLM fragments
The best llm CLI program now supports tool-calling, very excited about these possibilities —> Large Language Models can run tools in your terminal with LLM 0.26
Great investigation —> Police secretly monitored New Orleans with facial recognition cameras
How reliable are LLMs at extracting data from pdfs? Inspired by Simon Willison’s PyCon talk, I added extracting FEMA’s daily operation briefing to my LLM evals suite.
Just one model extracted the data from the pdf correctly: Gemini 2.5 Pro Preview. —> Extract FEMA incidents | LLM evals - Kevin Schaul