Visualizing Claude System Prompts as a Git Timeline
A new research project transforms Anthropic's publicly available Claude system prompts into a browsable Git timeline. This innovative approach, facili...
Simon Willison
Analyzing System Prompt Changes in Claude Opus 4.6 and 4.7
Simon Willison provides an analysis of the evolving system prompts between Anthropic's Claude Opus 4.6 and 4.7 models. This research highlights Anthro...
Simon Willison
Ecom-RLVE: Adaptive Verifiable Environments for E-Commerce Conversational Agents
Ecom-RLVE introduces a new framework for creating adaptive and verifiable environments specifically designed for e-commerce conversational agents. Thi...
Hugging Face
Building a Fast Multilingual OCR Model with Synthetic Data
NVIDIA and Hugging Face have collaborated to develop a fast multilingual Optical Character Recognition (OCR) model, leveraging synthetic data generate...
Hugging Face
Anthropic Introduces Claude Opus 4.7 with Enhanced Performance
Anthropic has launched Claude Opus 4.7, its latest model designed to deliver stronger performance across various applications, including coding, agent...
Anthropic
Qwen3.6-35B-A3B Outperforms Claude Opus 4.7 in Pelican Image Test
In a comparative test, Simon Willison observed that Alibaba's Qwen3.6-35B-A3B model running locally generated a superior pelican image compared to Ant...
Simon Willison
Hugging Face Guide: Training and Finetuning Multimodal Embedding & Reranker Models
Details in article.
Hugging Face
OpenAI Unveils GPT-Rosalind for Life Sciences Research
OpenAI has introduced GPT-Rosalind, a new frontier reasoning model specifically engineered to boost research in the life sciences. This model is poise...
OpenAI
Hugging Face Explores VAKRA Benchmark for AI Agents
Details in article.
Hugging Face
Exploring the new `servo` crate
Simon Willison is exploring the recently released `servo` crate, which packages the Servo browser engine as an embeddable library. The author utilized...
Simon Willison
Quoting Bryan Cantrill on LLM ‚laziness‘
Simon Willison highlights Bryan Cantrill's critique of Large Language Models, noting their inherent lack of 'laziness' as a virtue. Cantrill argues th...
Simon Willison
Gemma 4 audio with MLX on macOS
A practical guide demonstrates how to transcribe audio files on macOS using the 10.28 GB Gemma 4 E2B model, integrating it with the MLX framework and ...
Simon Willison