About Simon Willison

Biography and Background

Simon Willison is a software engineer based in the San Francisco Bay Area, best known as co-creator of the Django web framework.[1] He created Datasette, a tool for exploring and publishing data, and the LLM CLI, a Python library and command-line tool for interacting with large language models (LLMs).[9][38] Willison maintains an active blog at simonwillison.net, a TIL (Today I Learned) site at til.simonwillison.net, and publishes a weekly newsletter featuring AI experiments, blog posts, and links.[70] He is a frequent conference speaker, including organizing PyCon US 2026 tracks on AI and security.[96]

LLM Experimentation and AI Tools

Willison is a prolific experimenter with LLMs, frequently benchmarking models using whimsical prompts like 'Generate an SVG of a pelican riding a bicycle'.[24][44][62][71][90][197] He tests quantized models, compares pricing (e.g., DeepSeek V4 as cheapest in class[71]), and evaluates local inference on Apple silicon using tools like MLX and Ollama.[47][48][188] Key projects include LLM CLI enhancements for shebangs[9] and reasoning models,[38] plus vibe-coded UIs like LiteParse for PDF extraction.[78] He explores agentic engineering, blurring 'vibe coding' with structured agents,[22] and critiques AI reply bots on Bluesky.[25][105]

Pelican Benchmark Controversy: Willison's SVG pelican tests[72] highlight model creativity but face implicit challenges: do they measure meaningful capabilities or just prompt engineering quirks? Critics might argue they prioritize fun over rigor.[91]

Datasette and Data Tools

Datasette remains a core project, with recent TILs on Google Sheets integration via importdata() and custom functions.[101] Willison pushes code updates frequently[127][128][156] and stars related repos like sqlglot and syntaqlite.[143][169] He envisions RSS-based search engines reviving Feedster concepts.[32]

Developer Productivity and AI Coding

Willison champions AI agents for tedious tasks like refactors[190] and TDD-driven features when codebases provide style precedents.[185] He discusses Claude Code pricing shifts,[81][93] Anthropic system prompt diffs,[98], and Zig's AI contribution ban rationale—prioritizing contributor growth over code quality.[33][34] GitHub activity shows stars in AI orchestration (Flyte[124]), workflows (Prefect[132]), and uv migration tools.[117][163]

AI in Open Source Challenge: Zig's ban[33] aligns with Willison's respect for maintainer time,[29] but he notes it sidesteps LLM code quality debates—countered by evidence of AI finding real vulnerabilities.[196]

AI Industry Analysis

Willison dissects pricing (GitHub Copilot pauses[81], Claude token hikes[100]), deals (xAI/Anthropic Colossus[16]), and announcements (OpenAI-Microsoft revenue share[51]). He highlights Gen Z's heavy AI use despite negative feelings,[58][75] critiques 'zombie internet' bots,[10] and notes AGI glimpses in emergent behaviors.[62]

Personal Interests and Life

Beyond tech, Willison tap dances in community college recitals,[12] birdwatches (white-crowned sparrows[30], iNaturalist integration[27]), visits factories (Heath Ceramics[1]), and enjoys shows like Widow's Bay[7] and Crazy Ex-Girlfriend.[14] He praises San Jose's immigrant food scene countering 'ghost town' narratives.[186][187]

Structured Claims and Challenges

From analyzed content, Willison claims like 'AI agents excel at replicating familiar code patterns'[185] face counters: agents struggle with novel tasks without precedents (medium strength). His ChatGPT voice wishes[183] are challenged as casual vs. formal proposals (moderate).[counter-claims]