CASE STUDY / Learning
Steps
Helping learning build on itself.
An experiment in turning worthwhile content into practice, feedback, and recall—and discovering where AI needs human judgment.
Open Steps beta ↗- Initial beta build
- ~15 hours
- Learn, respond, apply, recall
- 4 activities
- First learning curriculum
- Investing
Observe
Listening did not tell me what I had learned.
I spent my commute listening to founders and studying businesses. The stories were useful, but I wanted to explain the ideas, apply them to decisions, and remember them later.
My first experiment, Founder OS inside ChatGPT, paired company stories with questions and feedback. A WeWork discussion challenged my cost assumptions; a Shopify lesson made me test the reasoning behind my own e-commerce experience.
Understand
Learning needs a sequence—and a reason to return.
Those exchanges suggested a loop: selected teaching, a response, application, and recall. A learner’s goal and previous answers needed to influence the next session.
I chose investing for the first course because friends were interested and I could contribute subject knowledge, identify gaps, and compare approaches. That expertise was part of the product, even when AI was producing the screens.
Build
A working teaching-and-checking beta.
What I built
- Learn: selected material tied to a goal.
- Respond: explain an idea so gaps become visible.
- Apply: use it in a new situation.
- Recall: revisit important concepts over time.
How I built it
I defined the concept, curriculum structure, and lesson flow, then used ChatGPT for AI-assisted implementation. Repeated reviews shaped what the app remembered about goals, answers, and progress.
The initial beta took roughly 15 hours over several days, as described in my October 2026 essay.
Learnings
Expertise and motivation are product requirements.
- Quality: extending Steps to business Japanese exposed my inability to judge the generated curriculum. More material did not guarantee suitable teaching. Expert review belongs in the process.
- Motivation: I stopped completing much of my own written Founder OS work while still enjoying the podcasts. Interesting content did not automatically create a sustainable routine; a smaller useful action mattered.
- Next test: pilot one narrowly defined course with a subject expert. Can learners apply an idea without the app supplying the answer, and do they find enough value to return?
The beta demonstrates the learning flow. Retention, learning gains, and sustained participation still need to be established.