Coding agents as data analysis tools — practical workshop patterns for journalists and analysts¶
Insight: Simon Willison's NICAR 2026 workshop demonstrates that coding agents (Claude Code, OpenAI Codex) are practical tools for data journalists and analysts with limited programming experience. Agents successfully handled data cleaning, web scraping, geospatial visualization, and database querying across a three-hour session, processing approximately 200,000 records for heat map generation at minimal cost.
Detail: The workshop used GitHub Codespaces with budget-restricted Codex API keys, spending only $23 in total Codex tokens across all participants. Specific tasks included decoding neighborhood codes in a San Francisco tree database, creating interactive Leaflet heat maps, and running SQLite queries through agent delegation. The workshop materials are published as a handout at simonw.github.io/nicar-2026-coding-agents/. This demonstrates a concrete non-engineering use case for coding agents — data analysis — validating the pattern that Claude Code's value extends beyond software development into analytical workflows.
Sources
- Simon Willison — "Coding agents for data analysis" (2026-03-16)