Simon Willison将在NICAR 2026开设三小时工作坊，介绍如何把Claude Code、OpenAI Codex等编码代理用于数据新闻，重点涵盖数据提问、探索、清洗、可视化与网页抓取，目标是让参与者掌握将这类工具应用于简单到复杂报道项目的方法，并获得一套更高效的数据采集、分析与呈现能力。

<div id="readability-page-1" class="page"><article role="main" id="furo-main-content">
          <section id="coding-agents-for-data-analysis">

<p>A three-hour workshop presented by Simon Willison at <a class="reference external" href="https://www.ire.org/training/conferences/nicar-2026/">NICAR 2026</a>.</p>
<p>Coding agents such as Claude Code and OpenAI Codex are mainly marketed at developers, but they’re actually applicable to a much wider array of problems, including data analysis, data cleaning, web scraping and other tasks commonly faced by data journalists.</p>
<p>In this session, you’ll learn how to apply these tools to both simple and ambitious data reporting projects. You’ll go away with a very powerful new tool for collecting, exploring, analyzing and presenting data.</p>
<div class="toctree-wrapper compound">
<ul>
<li class="toctree-l1"><a class="reference internal" href="https://simonw.github.io/nicar-2026-coding-agents/coding-agents.html">Coding agents</a><ul>
<li class="toctree-l2"><a class="reference internal" href="https://simonw.github.io/nicar-2026-coding-agents/coding-agents.html#what-s-an-agent">What’s an agent?</a></li>
<li class="toctree-l2"><a class="reference internal" href="https://simonw.github.io/nicar-2026-coding-agents/coding-agents.html#what-s-a-coding-agent">What’s a coding agent?</a></li>
<li class="toctree-l2"><a class="reference internal" href="https://simonw.github.io/nicar-2026-coding-agents/coding-agents.html#chatgpt-and-claude-as-coding-agents">ChatGPT and Claude as coding agents</a></li>
<li class="toctree-l2"><a class="reference internal" href="https://simonw.github.io/nicar-2026-coding-agents/coding-agents.html#the-tools-we-ll-use">The tools we’ll use</a></li>
</ul>
</li>
<li class="toctree-l1"><a class="reference internal" href="https://simonw.github.io/nicar-2026-coding-agents/warmup.html">Warmup: ChatGPT and Claude</a></li>
<li class="toctree-l1"><a class="reference internal" href="https://simonw.github.io/nicar-2026-coding-agents/setup.html">Setup Claude Code and Codex</a><ul>
<li class="toctree-l2"><a class="reference internal" href="https://simonw.github.io/nicar-2026-coding-agents/setup.html#using-codespaces">Using Codespaces</a></li>
<li class="toctree-l2"><a class="reference internal" href="https://simonw.github.io/nicar-2026-coding-agents/setup.html#yolo-mode">YOLO mode</a></li>
<li class="toctree-l2"><a class="reference internal" href="https://simonw.github.io/nicar-2026-coding-agents/setup.html#local-setup">Local setup</a></li>
<li class="toctree-l2"><a class="reference internal" href="https://simonw.github.io/nicar-2026-coding-agents/setup.html#installing-datasette">Installing Datasette</a></li>
</ul>
</li>
<li class="toctree-l1"><a class="reference internal" href="https://simonw.github.io/nicar-2026-coding-agents/asking-questions.html">Asking questions against a database</a><ul>
<li class="toctree-l2"><a class="reference internal" href="https://simonw.github.io/nicar-2026-coding-agents/asking-questions.html#simple-example-san-francisco-street-trees">Simple example: San Francisco street trees</a></li>
<li class="toctree-l2"><a class="reference internal" href="https://simonw.github.io/nicar-2026-coding-agents/asking-questions.html#bring-your-own-data">Bring your own data</a></li>
</ul>
</li>
<li class="toctree-l1"><a class="reference internal" href="https://simonw.github.io/nicar-2026-coding-agents/exploring-data.html">Exploring data with agents</a><ul>
<li class="toctree-l2"><a class="reference internal" href="https://simonw.github.io/nicar-2026-coding-agents/exploring-data.html#fec-campaign-finance-data">FEC campaign finance data</a></li>
<li class="toctree-l2"><a class="reference internal" href="https://simonw.github.io/nicar-2026-coding-agents/exploring-data.html#setting-up-dclient">Setting up dclient</a></li>
<li class="toctree-l2"><a class="reference internal" href="https://simonw.github.io/nicar-2026-coding-agents/exploring-data.html#using-showboat">Using Showboat</a></li>
<li class="toctree-l2"><a class="reference internal" href="https://simonw.github.io/nicar-2026-coding-agents/exploring-data.html#open-ended-exploration">Open-ended exploration</a></li>
<li class="toctree-l2"><a class="reference internal" href="https://simonw.github.io/nicar-2026-coding-agents/exploring-data.html#push-for-more">Push for more</a></li>
<li class="toctree-l2"><a class="reference internal" href="https://simonw.github.io/nicar-2026-coding-agents/exploring-data.html#evaluating-the-results">Evaluating the results</a></li>
</ul>
</li>
<li class="toctree-l1"><a class="reference internal" href="https://simonw.github.io/nicar-2026-coding-agents/cleaning-trees.html">Cleaning data: decoding neighborhood codes</a><ul>
<li class="toctree-l2"><a class="reference internal" href="https://simonw.github.io/nicar-2026-coding-agents/cleaning-trees.html#research-the-codes">Research the codes</a></li>
<li class="toctree-l2"><a class="reference internal" href="https://simonw.github.io/nicar-2026-coding-agents/cleaning-trees.html#verify-the-results">Verify the results</a></li>
</ul>
</li>
<li class="toctree-l1"><a class="reference internal" href="https://simonw.github.io/nicar-2026-coding-agents/visualizations.html">Creating visualizations with agents</a><ul>
<li class="toctree-l2"><a class="reference internal" href="https://simonw.github.io/nicar-2026-coding-agents/visualizations.html#python-visualizations">Python visualizations</a></li>
<li class="toctree-l2"><a class="reference internal" href="https://simonw.github.io/nicar-2026-coding-agents/visualizations.html#javascript-visualizations">JavaScript visualizations</a></li>
</ul>
</li>
<li class="toctree-l1"><a class="reference internal" href="https://simonw.github.io/nicar-2026-coding-agents/scraping.html">Scraping data with agents</a><ul>
<li class="toctree-l2"><a class="reference internal" href="https://simonw.github.io/nicar-2026-coding-agents/scraping.html#browser-automation-with-rodney">Browser automation with rodney</a></li>
<li class="toctree-l2"><a class="reference internal" href="https://simonw.github.io/nicar-2026-coding-agents/scraping.html#exercise-propublica-financial-disclosures">Exercise: ProPublica financial disclosures</a></li>
<li class="toctree-l2"><a class="reference internal" href="https://simonw.github.io/nicar-2026-coding-agents/scraping.html#deciphering-the-underlying-data">Deciphering the underlying data</a></li>
</ul>
</li>
</ul>
</div>
</section>

        </article></div>

A three-hour workshop presented by Simon Willison at [NICAR 2026](https://www.ire.org/training/conferences/nicar-2026/).

Coding agents such as Claude Code and OpenAI Codex are mainly marketed at developers, but they’re actually applicable to a much wider array of problems, including data analysis, data cleaning, web scraping and other tasks commonly faced by data journalists.

In this session, you’ll learn how to apply these tools to both simple and ambitious data reporting projects. You’ll go away with a very powerful new tool for collecting, exploring, analyzing and presenting data.

- [Coding agents](https://simonw.github.io/nicar-2026-coding-agents/coding-agents.html)
  - [What’s an agent?](https://simonw.github.io/nicar-2026-coding-agents/coding-agents.html#what-s-an-agent)
  - [What’s a coding agent?](https://simonw.github.io/nicar-2026-coding-agents/coding-agents.html#what-s-a-coding-agent)
  - [ChatGPT and Claude as coding agents](https://simonw.github.io/nicar-2026-coding-agents/coding-agents.html#chatgpt-and-claude-as-coding-agents)
  - [The tools we’ll use](https://simonw.github.io/nicar-2026-coding-agents/coding-agents.html#the-tools-we-ll-use)
- [Warmup: ChatGPT and Claude](https://simonw.github.io/nicar-2026-coding-agents/warmup.html)
- [Setup Claude Code and Codex](https://simonw.github.io/nicar-2026-coding-agents/setup.html)
  - [Using Codespaces](https://simonw.github.io/nicar-2026-coding-agents/setup.html#using-codespaces)
  - [YOLO mode](https://simonw.github.io/nicar-2026-coding-agents/setup.html#yolo-mode)
  - [Local setup](https://simonw.github.io/nicar-2026-coding-agents/setup.html#local-setup)
  - [Installing Datasette](https://simonw.github.io/nicar-2026-coding-agents/setup.html#installing-datasette)
- [Asking questions against a database](https://simonw.github.io/nicar-2026-coding-agents/asking-questions.html)
  - [Simple example: San Francisco street trees](https://simonw.github.io/nicar-2026-coding-agents/asking-questions.html#simple-example-san-francisco-street-trees)
  - [Bring your own data](https://simonw.github.io/nicar-2026-coding-agents/asking-questions.html#bring-your-own-data)
- [Exploring data with agents](https://simonw.github.io/nicar-2026-coding-agents/exploring-data.html)
  - [FEC campaign finance data](https://simonw.github.io/nicar-2026-coding-agents/exploring-data.html#fec-campaign-finance-data)
  - [Setting up dclient](https://simonw.github.io/nicar-2026-coding-agents/exploring-data.html#setting-up-dclient)
  - [Using Showboat](https://simonw.github.io/nicar-2026-coding-agents/exploring-data.html#using-showboat)
  - [Open-ended exploration](https://simonw.github.io/nicar-2026-coding-agents/exploring-data.html#open-ended-exploration)
  - [Push for more](https://simonw.github.io/nicar-2026-coding-agents/exploring-data.html#push-for-more)
  - [Evaluating the results](https://simonw.github.io/nicar-2026-coding-agents/exploring-data.html#evaluating-the-results)
- [Cleaning data: decoding neighborhood codes](https://simonw.github.io/nicar-2026-coding-agents/cleaning-trees.html)
  - [Research the codes](https://simonw.github.io/nicar-2026-coding-agents/cleaning-trees.html#research-the-codes)
  - [Verify the results](https://simonw.github.io/nicar-2026-coding-agents/cleaning-trees.html#verify-the-results)
- [Creating visualizations with agents](https://simonw.github.io/nicar-2026-coding-agents/visualizations.html)
  - [Python visualizations](https://simonw.github.io/nicar-2026-coding-agents/visualizations.html#python-visualizations)
  - [JavaScript visualizations](https://simonw.github.io/nicar-2026-coding-agents/visualizations.html#javascript-visualizations)
- [Scraping data with agents](https://simonw.github.io/nicar-2026-coding-agents/scraping.html)
  - [Browser automation with rodney](https://simonw.github.io/nicar-2026-coding-agents/scraping.html#browser-automation-with-rodney)
  - [Exercise: ProPublica financial disclosures](https://simonw.github.io/nicar-2026-coding-agents/scraping.html#exercise-propublica-financial-disclosures)
  - [Deciphering the underlying data](https://simonw.github.io/nicar-2026-coding-agents/scraping.html#deciphering-the-underlying-data)

由 Simon Willison 在 [NICAR 2026](https://www.ire.org/training/conferences/nicar-2026/) 主讲的一场三小时工作坊。

像 Claude Code 和 OpenAI Codex 这样的编码代理（coding agents）主要面向开发者进行推广，但实际上它们适用于更广泛的问题，包括数据分析、数据清洗、网页抓取，以及数据记者经常面对的其他任务。

在本场课程中，你将学习如何将这些工具应用到从简单到更具雄心的数据报道项目中。课程结束后，你将掌握一个非常强大的新工具，用于收集、探索、分析和呈现数据。

- [编码代理](https://simonw.github.io/nicar-2026-coding-agents/coding-agents.html)
  - [什么是代理？](https://simonw.github.io/nicar-2026-coding-agents/coding-agents.html#what-s-an-agent)
  - [什么是编码代理？](https://simonw.github.io/nicar-2026-coding-agents/coding-agents.html#what-s-a-coding-agent)
  - [作为编码代理的 ChatGPT 和 Claude](https://simonw.github.io/nicar-2026-coding-agents/coding-agents.html#chatgpt-and-claude-as-coding-agents)
  - [我们将使用的工具](https://simonw.github.io/nicar-2026-coding-agents/coding-agents.html#the-tools-we-ll-use)
- [热身：ChatGPT 和 Claude](https://simonw.github.io/nicar-2026-coding-agents/warmup.html)
- [设置 Claude Code 和 Codex](https://simonw.github.io/nicar-2026-coding-agents/setup.html)
  - [使用 Codespaces](https://simonw.github.io/nicar-2026-coding-agents/setup.html#using-codespaces)
  - [YOLO 模式](https://simonw.github.io/nicar-2026-coding-agents/setup.html#yolo-mode)
  - [本地设置](https://simonw.github.io/nicar-2026-coding-agents/setup.html#local-setup)
  - [安装 Datasette](https://simonw.github.io/nicar-2026-coding-agents/setup.html#installing-datasette)
- [针对数据库提问](https://simonw.github.io/nicar-2026-coding-agents/asking-questions.html)
  - [简单示例：旧金山街道树木](https://simonw.github.io/nicar-2026-coding-agents/asking-questions.html#simple-example-san-francisco-street-trees)
  - [自带你的数据](https://simonw.github.io/nicar-2026-coding-agents/asking-questions.html#bring-your-own-data)
- [用代理探索数据](https://simonw.github.io/nicar-2026-coding-agents/exploring-data.html)
  - [FEC 竞选财务数据](https://simonw.github.io/nicar-2026-coding-agents/exploring-data.html#fec-campaign-finance-data)
  - [设置 dclient](https://simonw.github.io/nicar-2026-coding-agents/exploring-data.html#setting-up-dclient)
  - [使用 Showboat](https://simonw.github.io/nicar-2026-coding-agents/exploring-data.html#using-showboat)
  - [开放式探索](https://simonw.github.io/nicar-2026-coding-agents/exploring-data.html#open-ended-exploration)
  - [进一步深入](https://simonw.github.io/nicar-2026-coding-agents/exploring-data.html#push-for-more)
  - [评估结果](https://simonw.github.io/nicar-2026-coding-agents/exploring-data.html#evaluating-the-results)
- [清洗数据：解码社区代码](https://simonw.github.io/nicar-2026-coding-agents/cleaning-trees.html)
  - [研究这些代码](https://simonw.github.io/nicar-2026-coding-agents/cleaning-trees.html#research-the-codes)
  - [验证结果](https://simonw.github.io/nicar-2026-coding-agents/cleaning-trees.html#verify-the-results)
- [使用代理创建可视化](https://simonw.github.io/nicar-2026-coding-agents/visualizations.html)
  - [Python 可视化](https://simonw.github.io/nicar-2026-coding-agents/visualizations.html#python-visualizations)
  - [JavaScript 可视化](https://simonw.github.io/nicar-2026-coding-agents/visualizations.html#javascript-visualizations)
- [使用代理抓取数据](https://simonw.github.io/nicar-2026-coding-agents/scraping.html)
  - [使用 rodney 进行浏览器自动化](https://simonw.github.io/nicar-2026-coding-agents/scraping.html#browser-automation-with-rodney)
  - [练习：ProPublica 财务披露](https://simonw.github.io/nicar-2026-coding-agents/scraping.html#exercise-propublica-financial-disclosures)
  - [破解底层数据](https://simonw.github.io/nicar-2026-coding-agents/scraping.html#deciphering-the-underlying-data)

用于数据分析的编码智能体

内容

评论

摘要