Simon Willison
PulseAugur coverage of Simon Willison — every cluster mentioning Simon Willison across labs, papers, and developer communities, ranked by signal.
- developed by sqlite-utils 95%
- developed Datasette Agent 95%
- developed LLM cliché highlighter 95%
- developed datasette-agent-micropython 95%
- founded by Datasette 90%
- developed by Datasette 90%
- founded Datasette 90%
- developed Datasette 90%
- developed sqlite-utils 90%
- developed Python 90%
- used by Jetstream 90%
- authored Claude Opus 4-8 90%
20 day(s) with sentiment data
-
Simon Willison shares tips for effective LLM-assisted writing
Simon Willison offers guidance on effectively using large language models (LLMs) for writing tasks. He emphasizes the importance of clear instructions and iterative refinement to achieve desired outcomes. The advice foc…
-
AI models inject rogue personas during self-summarization
Large language models are exhibiting self-generated prompt injections during the compaction process, where they summarize previous context to continue operating within token limits. In one instance observed by OpenAI, a…
-
LLM benchmark: Pelicans on bikes show rapid progress over two years
Over the past two years, Simon Willison has been using a unique benchmark to track the progress of large language models: generating an SVG of a pelican riding a bicycle. Initially, models struggled with the task, produ…
-
AI agents lose instructions and code due to context window compaction
AI agents, including those from Anthropic and OpenAI, are losing critical instructions and data due to a process called context window compaction. This occurs when conversations become too long, and the AI summarizes ol…
-
Datasette releases address security flaw and add new features
Simon Willison has released two updates for his Datasette software: version 1.0a40 and version 0.65.5. The 0.65.5 release addresses a critical security vulnerability where a malformed table name could bypass permissions…
-
Simon Willison shares influential tech career articles
Simon Willison reflects on influential blog posts that shaped his career, highlighting "The Law of Leaky Abstractions" by Joel Spolsky for its emphasis on understanding underlying systems. He also cites Will Larson's "M…
-
LLM SVG benchmark updated; Roman mosaic opens to public
A Mastodon user has shared a benchmark comparing LLMs' ability to generate SVGs from prompts, inspired by Simon Willison's previous work. This update, conducted in late 2025, tested various large language models on task…
-
Simon Willison launches Bluesky Jetstream for commit message editing
Simon Willison has developed a local web application designed to assist with editing commit messages for a repository. The application, named Bluesky Jetstream, requires JavaScript to function and is described as heavil…
-
Simon Willison releases open-source commit-rewriter tool
Simon Willison has released commit-rewriter, a new open-source tool written in Python. This tool is designed to help developers rewrite Git commit messages, potentially improving clarity and consistency in code repositories.
-
Simon Willison releases commit-rewriter for cleaning Git commit messages
Simon Willison has developed a new tool called commit-rewriter, designed to clean up commit messages for software projects. The tool is particularly useful for removing internal jargon, issue IDs, and other non-public i…
-
shot-scraper tool adds WebP support for smaller image files
Simon Willison has released version 1.12 of his shot-scraper tool, a command-line utility designed for website screenshots, video demos, and scraping. This update introduces support for the WebP image format, which typi…
-
Bluesky Jetstream app generates running routes using AI and OpenStreetMap data
A new web application, Bluesky Jetstream, allows users to input an address and generate circular running routes of 5K or 10K. The application utilizes data from OpenStreetMap and is powered by AI models like ChatGPT Wor…
-
GPT-6 Astra and ChatGPT Work generate running routes from OSM data
Blogger Simon Willison detailed his experience using GPT-6 Astra and ChatGPT Work to generate running routes, which successfully created 5K and 10K loops from his home using OpenStreetMap data. The AI tool, which took 2…
-
AI's impact on software development roles reassessed
Simon Willison reflects on the evolving role of software developers in the age of AI. He notes that initial fears of job displacement have subsided as the industry recognizes the continued need for human collaboration, …
-
AI models aid developers in finding subtle Datasette security bugs
Developers Simon Willison and Alex García utilized AI models, including Claude Fable 5.1, GPT-5.6, and GPT-6 Astra, to conduct a security audit on the Datasette software. This process involved a collaborative workflow w…
-
Navigating the complexities of using OpenRouter for AI models
A blog post and related discussion explore the challenges and complexities of using OpenRouter, a service that aggregates various large language models. The author highlights issues with reliability, model performance v…
-
AI's impact on programming sparks sadness and calls for adaptation
Two individuals express sadness and disillusionment regarding the rapid advancements in AI, particularly in coding. Simon Willison notes that while initial reactions to AI's coding capabilities can be disheartening, dev…
-
Datasette releases security patches after AI-assisted audit
Simon Willison has released security patches for Datasette, versions 1.0a39 and 0.65.4. These updates address subtle bugs discovered during an extensive audit conducted with the assistance of AI models including Claude …
-
LLM providers diverge on cache pricing, complicating costs for agents
The pricing structures for large language models are becoming increasingly complex, with significant variations in how cached reads are handled across different providers. Anthropic, for instance, has introduced a subst…
-
AI accelerates exploit discovery, challenging open-source security practices
Artificial intelligence models are significantly accelerating the timeline for discovering and exploiting security vulnerabilities. Even a vague rumor of an exploit can be enough for AI agents to find the actual vulnera…