This week, the goal is to build comfort and familiarity with the Anthropic Claude interface and tools specifically before we engage with more complex tasks. Try experimenting with other projects and larger questions in addition to the exercises, and aim to spend enough time with the interface to get frustrated and bring questions to our next live session.
Asynchronous expectations. Read all three Light readings, and complete at least one of the five tasks below. Doing more is welcome — and the cohort Discord is where the exchange lives. Share each task back in Discord as you finish it.
Recommended Viewing
Melanie Mitchell — The past, present, and uncertain future of AI — pairs directly with the Deep reading below, and is a one-sitting talk this series can point to.
Reading Menu
- Required Light Mollick, Ethan. “Assigning AI: Seven Ways of Using AI in Class.” One Useful Thing. If you skipped this last week, it is a useful 10-minute read for new adopters.
- Required Light Willison, Simon. “Not All AI-Assisted Programming is Vibe Coding (But Vibe Coding Rocks).” March 19, 2025. The piece that named the working vocabulary the rest of the series uses.
- Required Light Anthropic. “Use Projects to organize your work.” Skim the help-center docs before W3.
- Standard Bender & Hanna, The AI Con, Chapter 1: An Introduction to AI Hype. The skeptical counterweight to the practitioner blogs.
- Standard Emerson, Lori. “Interfaced.” From Further Reading. Interface is not transparent; reading this changes how you see the chat window.
- Deep Mitchell, Melanie. Artificial Intelligence: A Guide for Thinking Humans, Part I. A clear non-technical explanation of how these systems work.
Five Tasks This Week (Pick At Least One)
The interface is the curriculum this week. Pick at least one task; pick more if you have the time. Each task lands in the cohort Discord — your post is the deliverable.
1. Tour the settings, set your defaults
Walk through every settings panel in Claude. Set the defaults that make sense for your work — writing style, response length, what to call you, capability toggles (Artifacts, web search, code interpreter), conversation-history retention. Try voice mode once if your account has it. Switch the model selector between Opus, Sonnet, and Haiku on the same prompt and notice the trade-offs.
Discord post: what surprised you about what’s adjustable, and what you wish were? Drop one screenshot if useful.
2. Add a skill or connector
In Claude, open Customize → Add a skill or connector. Investigate the menu — Anthropic’s official skills (skill-creator, document and slide builders, spreadsheet skill, the others), and the connectors to outside services (Google Drive, Gmail, Notion, your library catalog if it’s listed). Pick one that looks useful for your work and try it on a real task you’d otherwise do by hand.
Discord post: which skill or connector you tried, what it changed about Claude’s defaults, and what you’d build on top of it.
3. Upload a document and research
Use Upload file to give Claude a real document of your own — your past syllabus for this course, your notes on what you want to teach, an outline of an assignment you’d like to redesign, or a piece of writing you want feedback on. Then ask Claude to research what other instructors have done in similar classes with generative and agentic AI: examples, frameworks, places it would push you to think harder. Iterate on the question; don’t take the first answer.
Discord post: what (if anything) Claude surfaced that was useful, and where its research felt thin or wrong. Don’t reverse-engineer a “win” — the negative finding is also data.
4. Iterate a poem in an Artifact
Pick a poetic form — haiku, sonnet, ghazal, blackout, cento. Iterate ten times in Claude Sonnet on a topic that matters to you. Publish the final as a Claude Artifact and copy the share URL. Source: adapted from HumanitiesAI/weektwo.
Discord post: the Artifact URL and one observation about how the conversation history shaped each draft.
5. Multi-chatbot comparison
Pick a more complex task you’d actually want help with — a research question for your work, a piece of poetry or other creative writing, an assignment you’d give a class, a syllabus paragraph. Run it through two or three of the free / institutional chatbots:
- UCF Copilot (Microsoft, OpenAI under the hood, UCF-data framing on top)
- ChatGPT free
- Gemini free
Notice what each tool lets you control (model selection, custom instructions, file upload, web search, voice) and what it doesn’t (defaults you can’t override, refusals, formatting it imposes, what it does with your data).
Discord post: what you noticed — what’s different in the interfaces and the controls? Where did one tool give you something the others didn’t? Where did one refuse what another offered?
What to Carry Into Workshop 2
If you do nothing else this week, assemble three to ten short texts you might want to analyze during the workshop session — chapters, articles, primary sources, or anything you would assign or study: we’ll discuss copyright and fair use during the workshop. Bring them as plain text or PDFs. We will upload them to a fresh Project on May 27.
Also worth bringing: the questions and frustrations you accumulated while clicking around. The point of this week is to get stuck enough to have something to ask.
Cross-references
- Source materials: HumanitiesAI/weektwo (interface and iteration as method).