Talking Postgres with Claire Giordano AI for data engineers with Simon Willison
14 snips
Aug 8, 2025 
Simon Willison, an open source developer and creator of Datasette, dives into how AI is reshaping data engineering. He shares the importance of structured data extraction and accessibility, underscoring Postgres's fit for AI projects. Their discussion spans the transformative power of AI in journalism, the evolving role of tools like large language models in data tasks, and the humorous benchmark of pelicans in AI-generated art. Willison also highlights essential AI tools, pushing for a playful and adaptive use of technology.
AI Snips
Chapters
Transcript
Episode notes
Alt Text Extraction With Claude
- Simon asked Claude to write a 150-line Postgres SQL query that extracted alt text across HTML and Markdown stored in multiple tables.
- The model produced working regex-based SQL immediately, enabling a task Simon wouldn't have done manually.
Draft Alt Text, Then Edit For Context
- Use LLMs to draft alt text but always review and edit the output for context and editorial relevance.
- Prompt models to focus on the image elements that support the article's message, not every visual detail.
Structured Extraction Unlocks Hard Data
- Structured data extraction (text→JSON) is hugely valuable for automating tedious data entry across varied formats like PDFs and images.
- Long-context multimodal models (e.g., Gemini 2.5) make large-document extraction practical where manual entry was infeasible.