Simon Willison news
The latest AI and developer news from Simon Willison, with short summaries.
Visit website →ChatGPT Image Model Inserts Text
ChatGPT Images 2.0 added text to an image that wasn't in the prompt.
Quoting Romain Huet
Simon Willison posted a short piece quoting a remark by Romain Huet.
GPT-5.5 Prompting Guide
OpenAI has released prompt-writing and upgrade tips for GPT-5.5.
People don't crave automation
Nilay Patel explains why the public has turned away from AI through the lens of the software brain.
Millisecond Converter for LLMs
Simon Willison introduced a tool that converts milliseconds from LLMs into seconds and minutes.
Simon Willison's Weekly Newsletter
This week's Simon Willison newsletter was a large issue featuring a new chapter of Agentic Engineering Patterns and several links.
Queue/Stream Extension for SQLite
honker is a Rust-based SQLite extension that adds Postgres-style NOTIFY/LISTEN and durable streams.
Claude Code Quality Issues
Anthropic revealed that the Claude Code quality degradation was caused by a harness bug, not the model.
Running Bluesky's For You Feed
Bluesky's For You feed runs on a single Go server and SQLite sitting in a living room.
Extracting PDFs in the Browser
Ran LiteParse in the browser to build a PDF text extraction web app that even supports OCR.
Calling GPT-5.5 via Codex
Simon Willison released an LLM plugin that uses GPT-5.5 through a Codex subscription.
Calling OpenAI models via Codex
llm-openai-via-codex 0.1a0, which calls OpenAI models using an existing Codex subscription, has been released.
Quote from Maggie Appleton
Learning and documenting in public can make you appear more capable than you actually are.
Defense wins
Firefox 150 fixed 271 vulnerabilities found through early AI evaluation.
Copilot Pricing Overhaul
GitHub Copilot has locked individual plans and moved Claude Opus 4.7 up to Pro+.
The $100/Month Controversy
Anthropic appeared to move Claude Code from the $20 plan to the $100 plan, but quickly reversed it.
Agents Are Too Human
AI agents are already too human, repeatedly showing distraction and inconsistency.
Pelican Data Poisoning
Steve Cosman's GitHub repository intentionally poisons 'pelican riding a bicycle' training data.