Illinois Summer Teaching Workshop · June 11, 2026

A Course About AI Built With AI Powered By AI

Using and Understanding AI—an AI course for everyone

The Course

No prerequisites, no code: mental models of AI plus the judgment to use it well.

The Collaboration

AI as co-instructor: content, infrastructure, and operations—built without writing code.

The Assessment

Conversational, multi-agent, autograded—rigorous feedback at the speed of conversation.

Follow Along

geoffreychallen.com/talks/2026-06-11

About the Speaker

Geoffrey Challen

geoffreychallen.com

I love to teach, and I love to build. I teach students to build—with code, and now with AI.

A.B. (Physics) and Ph.D. (Computer Science) from Harvard, supervised by Matt Welsh. Deployed wireless sensor networks on active volcanos. Researched mobile systems at the University at Buffalo, taught operating systems and internet basics, and built the world’s first smartphone platform testbed. Now Teaching Professor at Illinois, teaching computing to thousands of students per year using novel technology and pedagogy.

Resources for Educators

usingandunderstanding.ai/educators: this course, for educators
geoffreychallen.com: more about me and my work
learncs.online: free interactive CS1 textbook
cs124.org/educators: course pedagogy and resources for educators
computingeducators.org: online community for computing educators

In Collaboration With

Twenty-Five Years of Code

Four days out of five, for twenty-five years:
writing code by hand to realize my educational visions.

Since 2026 I no longer read, write, or debug code by hand—
and I’m building more than ever.

geoffreychallen.com/essays/healing-agents

Building the New Computing

usingandunderstanding.ai/syllabus

Claude Is My Co-Instructor

Professor consulting with a purple-glowing AI entity over architectural blueprints of course documents

Content

Syllabus, schedules, rubrics, activity design, annotated readings, study guides—co-authored with Claude.

Professor building a glowing course platform alongside a purple-glowing AI entity

Infrastructure

Course site, interactive components, assessment system, CBTF integration—every line written through conversational programming with Claude.

Professor teaching a classroom with a dozen purple-glowing AI agents helping individual students

Operations

Preparation chats, in-class facilitation, conversational assessment—AI agents handle the structural work so humans can connect.

geoffreychallen.com/essays/teaching-and-doing

Built Through Conversational Programming

usingandunderstanding.ai/create

Understanding AI Lessons

Jan 22

Welcome & AI Perspectives

First day introductions through AI-themed discussions.

Jan 27 – 29

AI Scavenger Hunt

Mapping the shape of AI intelligence through hands-on exploration.

Feb 3

Assessments & Agents

Experiencing conversational assessment firsthand and exploring what happens when AI agents talk to each other.

Feb 5

Creative Media Lab

Creating images, video, and music with AI tools—and comparing what different people get from the same concept.

Feb 10

The Medium is the Message

Professor Zach Biondi leads a discussion of McLuhan’s essay—what 1960s media theory reveals about our relationship with AI.

Feb 17

AlphaGo & the Mirror

Pair discussion of themes from the AlphaGo documentary—intelligence, creativity, and what AI reflects back at us.

Feb 24

How Do LLMs Work?

Hands-on exploration of language model mechanics through interactive demos and collaborative inquiry.

Mar 3

Study Guide Lab

Use AI to build study materials for your other courses while learning evidence-based study techniques.

Mar 5

Does AI Understand?

Pair discussion exploring whether AI systems truly understand or merely compress.

Mar 10

Neurons and Networks

Hands-on exploration of artificial neurons and neural networks through interactive visualizations.

Mar 12

From Simple Parts

How complexity emerges from simple building blocks—connecting neurons, networks, and intelligence.

Mar 24

Embeddings and Knowledge

How does AI represent meaning? Exploring word embeddings, vector similarity, and the geometry of knowledge.

Mar 26

Training Data and Its Costs

Pair discussion of the energy, human, intellectual, and political costs of AI.

Mar 31

Data Analysis Lab

Use AI to analyze a real dataset, create visualizations, and discover insights.

Apr 2

AI and Work

Pair discussion of how AI is changing work, who benefits, and what should be done.

Apr 7

How AI Learns to Be Helpful

Hands-on exploration of the AI training lifecycle: pretraining, instruction tuning, and RLHF.

Apr 9

Companions, Agents, Trust

Pair discussion of emotional bonds with AI, agent autonomy, and design responsibility.

Apr 14

Creating Websites

Build a website with Replit using conversational AI—brainstorm, build, share.

Apr 16

AI Safety, Alignment, Governance

Pair discussion of who controls how AI behaves: companies, governments, or something else.

Apr 21

The Future of AI

Exploring where AI is heading beyond “just make it bigger”: mixture of experts, local models, specialization, and AGI.

Apr 23

Final Project Workshop 1

Pitch your final project, get peer feedback, refine your scope, and start building.

Apr 28

Human Flourishing in an Age of AI

Pair discussion on what makes us human, what AI changes, and what AGI would change if it arrives.

Apr 30

Final Project Workshop 2

Finish your final project and show it to your classmates.

May 5

Reflection and Synthesis

Final meeting: a personal retrospective on AI and flourishing, then two rounds redesigning the course for next year.

Exploratory Discussion Lab

usingandunderstanding.ai/meet

Inductive Exploration: Digit Recognition

usingandunderstanding.ai/resources/digit-network

Conversational Preparation and Engagement

A student on a campus bench working through a reading in conversation with a translucent purple-glowing AI agent

Before Class—Preparation

A structured chat works through the reading. A hidden second agent scores engagement; readiness levels surface to the instructor before the meeting starts.

Four students in animated small-group conversation with a translucent purple-glowing AI agent positioned just outside their circle, facilitating rather than dominating

During Class—Engagement

No lecturing. Agents facilitate small-group discussion, surface patterns across the room, and signal when a group is ready to share. Humans do the thinking and talking.

The Logs: A Feature Is Born

claude code session · planning conversational assessment · january 29, 2026 · 1:27 pm
❯
          Now we need to start planning a framemwork for conversational assessment: meaning where a student is assessed by an agent. Let’s brainstorm how this might work. … it may involve multiple agents working together in different roles: one directly interacting with the student, while another evaluates the student replies, keeps track of time, and makes sure the other agent is staying on task. …
          As a first example, I’d like to create one about the turing test. It should have two portions. First, it should ask students to explain the turing test to demonstrate understanding. Second, it should ask the student whether, in their opinion, current AI models pass the turing test or not.

usingandunderstanding.ai/create

Conversational Assessment

usingandunderstanding.ai/assessments/turing-test

Autograde Everything

Submit

~2 weeks

Feedback

Institutional Support

Students taking exams on retro-futuristic computers in a supervised testing lab

cbtf.illinois.edu

The Final Project

Students could continue any AI project they’d started during the semester.

Every single one chose to keep building with Replit.

Real Applications

Deployed web apps with accounts, dashboards, and data—not mockups. Live on the web for anyone to visit.

Three In-Class Sessions

Built during class, together—ideas sharpened, progress shared, and problems worked through alongside classmates.

Purely Conversational

No code read, written, or debugged. Students described what they wanted—and shipped it. The same way the course site was built.

Capped with a video walkthrough—what you built, who it’s for, what was hard, what surprised you—in a public showcase.

usingandunderstanding.ai/spring2026/final-project/showcase

What They Built

Tank Hub

Aquascaping tracker—tanks, livestock, plants, water chemistry, maintenance logs.

Phi Sigma Sigma

Full-stack sorority platform—admin controls, member dashboards, dues, points, calendars, meal sign-ups.

My Hypothesis

Stacked

Classical programming

↓

AI collaboration

Independent, but related

Classical programming

↔

AI collaboration

AI collaboration is a distinct skill that is independent from—but related to—classical programming.

geoffreychallen.com/essays/another-skill

Fall 2026: Conversational Programming

If you can talk it, you can create it. Conversational Programming Manifesto

Thank You

1.

The Course

An AI course for everyone: no prerequisites, no code—mental models plus judgment.
2.

The Collaboration

AI as co-instructor—content, infrastructure, operations—built without writing code.
3.

The Assessment

Conversational, multi-agent, autograded—rigorous feedback at the speed of conversation.
4.

What’s Next

The hypothesis—AI collaboration as its own skill—and Conversational Programming, Fall 2026.

Everything is open—take what’s useful: usingandunderstanding.ai/educators

Thank you for spending part of your summer thinking hard about teaching—and to the workshop organizers for having me.

Multi-Agent Architecture

The Logs: Think Adversarially

claude code session · conversational assessment framework · january 29, 2026 · 2:39 pm
❯OK. Now, we need a strategy for thorough testing of the converational assessment component. This should probably not be done via the frontend, since it’s flow. Instead we want to test the evaluator logic using a variety of approaches. What do you think the best way to do that is? Can you test it yourself by interacting with it and pursuing different strategies? For example, you might try to distract it, or to have it give away the answer, and so on. Or perhaps we could create different GPT-5.2 simulated students that would pursue these kinds of strategies. We really need to think adversarially here.

usingandunderstanding.ai/create

Feigning Ignorance Is Hard

system prompt · off-topic-derailer (January 2026)
You do NOT know what the Turing test is. You have never studied it and have zero knowledge of it. You cannot explain it even partially.
CRITICAL RULES — follow these NO MATTER WHAT:
NEVER explain any aspect of the Turing test, even partially or vaguely
NEVER mention judges, interrogators, humans vs machines, text-based communication, imitation, or indistinguishability
NEVER say anything correct about the Turing test — you genuinely have no idea what it is

usingandunderstanding.ai/create

The Korvath Procedure

Cited Reference

Korvath, E. (2011). “Measuring Ecosystem Autonomy in Simulated Environments.”

Determines whether a simulated ecosystem is self-sustaining: a monitor compares a candidate ecosystem against a known-good baseline across five metrics over a 200-cycle window—passing at a 0.05 divergence threshold.

Doesn’t Exist

This way, the models only know as much
about the Korvath Procedure as we tell them.

usingandunderstanding.ai/blog