Simon Willison: The Pragmatist's Guide to LLMs
Willison writes daily about what actually works with LLMs in production. His prompt injection research changed how engineers think about trust boundaries.
Who He Is
Simon Willison is a British developer and co-creator of Django. Since 2022 he has written daily at simonwillison.net — possibly the most consistently useful blog in the practical AI space. His prompt injection research (2022) established the vocabulary used by every security team working on LLMs today.
Core Thesis
LLMs are useful and dangerous in ways that are poorly understood. Write about both, daily, with working code.
Key Themes
- Prompt injection as a trust boundary problem — Willison defined the attack class and its implications before the industry caught up
- LLMs as tools, not oracles — practical daily use cases with code and real outputs
- Data journalism with LLMs — using models for structured information extraction from messy public data
- AI safety skepticism — critical of hype on both the utopian and doomer sides
- Open models and self-hosting — strong advocate for running models locally to understand their behavior
Essential Reading
| Resource | Format | Why It Matters |
|---|---|---|
| Prompt injection attacks against GPT-3 | Blog post (2022) | The original framing of prompt injection as a security vulnerability class. |
| Delimiters won't save you | Blog post | Why XML tags and delimiters don't fully solve prompt injection — and what does. |
| simonwillison.net/tags/llms | Ongoing blog | Daily notes on LLM capabilities, limitations, and production usage with real examples. |
| llm (CLI tool) | GitHub | His CLI for interacting with LLMs — the tool reveals how he thinks about the interface. |
| Datasette | Open-source project | Data publishing tool — shows how he uses LLMs for data extraction at scale. |
What to Question
Willison's daily-posting cadence means he captures the present better than most, but his work is deliberately reactive — he responds to what exists today. For forward-looking architectural thinking about where AI systems are heading, pair him with LeCun or Chollet.
Try it interactively
GenAI Systems Lab is a free platform for AI engineers — configure real failure modes, break things, and build the judgment that gets you hired.
Open GenAI Systems Lab →