Simon Willison Shares Coding Agents Workshop Handout for Data Journalists

Simon Willison

Mar 20, 2026 · Updated Jun 5, 2026

Simon Willison published the handout from his NICAR 2026 coding agents workshop for data journalists. It covers querying databases, cleaning data, building visualizations, and scraping — using Claude Code and Codex. Designed for journalists, it applies to anyone exploring data.

Simon Willison, creator of Datasette, published the handout from his three-hour "Coding agents for data analysis" workshop at NICAR 2026, the data journalism conference. The guide covers eight modules: querying a SQLite database with natural language, exploring FEC campaign finance data, cleaning data by decoding neighborhood codes, creating Python and JavaScript visualizations, and scraping with browser automation. The workshop ran on GitHub Codespaces with OpenAI Codex — attendees burned $23 of tokens.

Coding agents like Claude Code and Codex are marketed primarily at developers, but this handout makes the case they're equally applicable to data journalists, researchers, and analysts. Willison designed it for people who weren't at NICAR — the material transfers directly to any data exploration workflow.

Work through the modules on your own dataset — exercises use Python, SQLite, and Datasette, a data publishing tool, for exploration and visualization. The scraping section covers browser automation for data extraction.

View the full update on simonw.github.io

Simon Willison

@simonwMar 16

Here's the handout for a three hour workshop I presented at the NICAR data journalism conference on using coding agents (Codex CLI, Claude Code etc) for data exploration, visualization and analysis https://t.co/zriTHBGiOP

View on X

Every HeadsUpAI update is written based on its original source and reviewed before it's published. Read our editorial standards →

Keep reading

Simon Willison Publishes Guide on How Coding Agents Work Under the Hood

Simon Willison's new Agentic Engineering Patterns chapter breaks down how coding agents work — from LLM token mechanics and tool calling to system prompts and reasoning. A companion chapter covers subagents across seven major coding agents.

Andrew NgJan 28

DeepLearning.AI and Anthropic Launch Free Agent Skills Course

DeepLearning.AI and Anthropic launched a free course on building agent skills, taught by Anthropic's Head of Technical Education. It covers creating skills for code review, data analysis, and research, then deploying across Claude.ai, Claude Code, and the Agent SDK.

MongoDBApr 1

MongoDB Launches Agent Skills to Embed Database Best Practices in Coding Agents

MongoDB introduced official Agent Skills and plugins for major AI coding tools like Claude Code and Cursor to provide agents with domain-specific expertise. These curated instructions ensure AI-generated code follows NoSQL best practices for schema design and query optimization rather than defaulting to relational patterns.

NVIDIAMay 22

NVIDIA Launches AI-Q Agent Skill to Bring Deep Research to Coding Agents

NVIDIA released the AI-Q Agent Skill, a portable version of its research pipeline that integrates with agent harnesses like Claude Code and Codex. Instead of coding agents attempting complex multi-source synthesis themselves, they can now delegate research tasks to a dedicated AI-Q server that returns cited reports.