Simon Willison — Independent developer and writer documenting practical experiments, tools, and deep analysis around large language models, generative AI, web development, security, and emerging programming workflows through detailed posts and daily TILs.
117 articles from this blog
A quote from John Carmack on the pitfalls of over-architecting software for future needs, shared on a developer's blog.
Argues that AI coding agents should be used to improve code quality and reduce technical debt, not just speed up development.
New PostgreSQL 18 and SQLite features allow developers to copy database statistics to replicate production query plans in dev environments.
Explores how modern LLM coding agents adapt to new or private tools, challenging the 'Choose Boring Technology' assumption.
A quote from Joseph Weizenbaum on how simple AI programs can cause delusional thinking in people, shared on a tech blog.
OpenAI offers 6 months of ChatGPT Pro with Codex to core maintainers of significant open source projects, following a similar move by Anthropic.
Key questions for developers, managers, and stakeholders to audit a Rails codebase and identify technical debt and process issues.
Analysis of Anthropic's Pentagon contract and its branding as a 'moral' AI provider in a commodified AI market.
Explores the importance of manual testing by AI coding agents to complement automated tests, with practical patterns for Python, APIs, and web UIs.
A detailed analysis of a prompt injection attack against Cline's GitHub repo, leading to cache poisoning and a compromised NPM release.
OpenAI announces GPT-5.4 and GPT-5.4-Pro models with enhanced coding, document creation, and a 1M token context window.
Explores the legal and ethical implications of AI coding agents creating "clean room" rewrites to relicense open-source software, using the chardet Python library case.
Reports on high-profile resignations and internal reorganization within Alibaba's Qwen AI team, raising questions about the project's future.
Donald Knuth shares his reaction to an AI solving an open problem he was working on, highlighting advances in AI reasoning.
A skeptic's detailed journey using AI coding agents to port scikit-learn to Rust, showcasing the surprising capabilities of modern LLMs.
Anthropic offers free Claude Max AI access to eligible open source maintainers for six months.
A developer builds a Unicode character lookup tool using binary search over HTTP range requests, with AI assistance.
Explores how AI-generated code creates 'cognitive debt'—a loss of system understanding—which can paralyze developers more than technical debt.
Announcing Interop 2026, a cross-browser initiative by major vendors to improve web platform feature compatibility and parity.
Anthropic's Claude Code creator discusses the evolving role of engineers and why developers are still crucial in the AI era.
