ESCAPE THE
CHAT WINDOW
Move AI off the web and onto your machine. Scrape data, process documents, build automation pipelines, and create tools for your newsroom — all from the command line.
TRANSCRIPT
Every time you open ChatGPT, or Claude, or Gemini, you start from scratch. You paste in your documents, you fiddle with your prompts, explain what you're covering, describe the output format, and then next week, when you come back to it, you have to do it all again from scratch.
I'm Joe Amditis, Associate Director of Operations at the Center for Cooperative Media, and hopefully this course will fix that for you.
This is Advanced Prompt Engineering for Journalists, a four-week online asynchronous course from the Knight Center at UT Austin.
This course is all about escaping the chat window on the web, and moving to the command line, or at least getting off the web and getting these tools into the files and folders on your computer, so you can do real work with them.
You'll learn how to create reusable and repeatable commands for your newsroom pipelines and workflows. Together, we will use plain English to describe automation workflows and have AI build them for you right in front of your eyes.
You'll connect your tools to your data and let the AI search across them and cite its sources.
You don't need to know how to code or program or do software development. The AI handles most of that itself. All you need to do is understand and articulate what you want and how you want it, and you already do a lot of that as your day job.
So if you already have some familiarity with AI tools and you want to go from just chatting to actually delegating, then this course is for you.
Coming soon at JournalismCourses.org, and we'll see you there.
Modules
Four weeks covering the fundamentals of CLI-based AI tools for journalism.
From chat window to command line
Install CLI tools, compare web vs terminal workflows, and write a CLAUDE.md context file that persists across sessions.
Custom skills for Claude Code
Package domain expertise into reusable slash commands. Write skills for data processing, content workflows, and beat-specific tasks.
CLI workflows for newsrooms
Build shell pipelines for data journalism. Automate scraping, cleaning, and summarization into cost-conscious, repeatable workflows.
Agents, tools, and data access
Connect AI to your files and data sources via MCP. Run agents that take multi-step action, and learn where these tools break down.
Git & GitHub for journalists
Understand version control well enough to tell Claude Code what you need. Commits, branches, merges, and GitHub — explained visually.
Bonus interviews
Conversations with journalists and media professionals who are adopting AI coding tools in their daily work.
Quick start
Get up and running in three steps.
Install Node.js
Download from nodejs.org and run the installer. This gives you npm, the package manager you will use to install AI tools.
$ node --versionInstall an AI tool
Pick one to start. Gemini CLI is free and works well for beginners.
$ curl -fsSL https://claude.ai/install.sh | sh
# or Gemini
$ npm install -g @google/gemini-cliFork the starter kit
Fork the starter repo, clone your fork, and launch.
$ cd mooc-starter-kit
$ claudeTools
We cover three CLI tools. You only need one to complete the course.
Claude Code
Terminal-based assistant. Strong at reasoning through complex tasks and working with existing codebases.
Gemini CLI
Command-line tool with free tier. Good for quick scripts and data processing tasks.
Codex CLI
Terminal assistant from OpenAI. Integrates well with existing OpenAI workflows and API keys.
Resources
Reference materials to support your learning.