Simon Willison

Simon Willison is a British programmer who co-created the Django web framework in the mid-2000s at the Lawrence Journal-World newspaper, co-founded the conference-directory startup Lanyrd (acquired by Eventbrite), and now builds Datasette, an open-source toolkit for exploring and publishing data. Since late 2022 his blog has become required reading in the AI era: a running, meticulously linked chronicle of what large language models actually do when a working programmer points them at real problems.

The empiricist of the LLM age

Willison occupies an unusual position in the debate: he is bullish on the tools while allergic to the hype around them. He waves off AGI talk, describes LLMs as fancy autocomplete that happens to be extremely useful for stringing code tokens together, and coined widely-cited framings like the “lethal trifecta” for prompt-injection risk. His March 2025 essay “Here’s how I use LLMs to help me write code” distilled two years of daily practice into the field’s most-shared practical guide.

Days later he drew the line that stuck in “Not all AI-assisted programming is vibe coding (but vibe coding rocks)”: Andrej Karpathy’s vibe coding means not reviewing the generated code, which is fine for toys — but professionals who use AI while reading, testing, and understanding everything it produces are doing something else entirely, and something enormously productive.

He ships constantly with these tools — his LLM command-line utility and hundreds of small projects are built with AI assistance — and he documents mistakes, model regressions, and security failures as prominently as breakthroughs, which is precisely why skeptics and boosters alike cite him.