Direction
Intent, imagination, taste, values, and accountability for the outcome.
Don’t stop at prompts or answers. Walk with an agent until an idea becomes a real thing.
The belief
Agentic AI is more than a tool for finishing tasks. It can pursue a goal across many steps—but its highest purpose is not autonomous output. It is a partner for expanding human imagination and turning it into products.
Human intent. Agentic momentum. A shared walk from possibility to product.
The agentic turn
A chatbot waits for an answer-sized request. An agent can hold a goal, plan, use tools, inspect its work, and try again. WalkBuild makes that capacity creative: the agent carries momentum while the human carries meaning.
Intent, imagination, taste, values, and accountability for the outcome.
Conversation, critique, feedback, and course correction while the work is alive.
Planning, tools, context, execution, and the ability to revise across many steps.
Bring the agent into the unfinished thought. Let it investigate, propose, and challenge—then let the exchange change what you see.
Do not disappear after the prompt. Begin with direction, not certainty. Stay in the loop as the agent plans and acts; let the problem become clearer as you move.
Agentic capability is becoming common. Imagination is not. The decisive skill is no longer knowing how to operate AI; it is imagining what is worth asking an agent to make possible.
A good conversation is not the finish line. Let the agent research, organize, prototype, and revise until the thought has a form another person can use.
Agents lower the distance between intention and execution by carrying work across tools and steps. Expertise still matters, but technical gatekeeping no longer gets the final word. The default posture is builder.
The agent may decide how to pursue a bounded goal. The human decides why it matters, what good looks like, what is safe, and when the work is ready. Accountability stays human.
Biggest limitation isn’t the tech, it’s imagination and creativity.
— Peter Steinberger (@steipete), September 25, 2026
The practice
Agentic, not automatic: give the agent room to act inside each phase without handing over the destination. You can circle back at any point.
Set the goal, stakes, boundaries, audience, and what the agent may do.
Let the agent investigate and propose. Follow surprises, argue, reframe, and redirect.
Give the agent tools to turn the strongest idea into a concrete, usable artifact.
Review, test, and approve the work. Put it in someone’s hands, learn, and begin again.
For the classroom
Students should learn to direct, challenge, and audit an agent—and leave with evidence of thought, creative risk, and a useful thing.
Foundations of LLMs and GenAI in Business · Fall 2026 · instructor: Prof. Shane Saunderson
Student showcase
Use the WalkBuild strategy to turn an idea into a working product. Student work created through this process will be featured on this site—so others can see not only what you made, but how your thinking moved along the way.
Start with something useful, meaningful, or genuinely worth exploring.
Show where you challenged the agent, changed direction, and made the key decisions.
Bring your finished work to class. Every completed WalkBuild product will be showcased here.
The declaration
We will not measure our future by how much work we hand to agents. We will measure it by the boldness of what we imagine and the usefulness of what we build together.