Class 8: September 10, 2026 · ENTR 5000
Notes from September 10, 2026. These notes record the discussion and instructions given that day. For current requirements and due dates, use the assignment guidelines and semester schedule.
Move from technical progress to business learning
Being an entrepreneur means finding answers to things you do not yet know—and improving how you find those answers. A technical learning goal, such as understanding a tool, is useful, but it should be distinguished from the business question that lets you move forward.
The business-learning exercise started with a short list:
- If you have a business, list questions whose answers would help you take the next step or grow it. Prioritize action-enabling questions over things that would merely be interesting to know.
- If you are still choosing an idea, list blockers: things that prevent action, such as missing information, access, resources, a contact, or a tool.
- Generate roughly six to ten questions or blockers before evaluating them. Quantity comes first; refinement comes next. The initial list was a thinking exercise, not a separate submission.
Choose a useful next move and define success
When there are too many possible next steps, try two extremes: the smallest tangible action you can take, or the furthest-upstream question that would clarify the whole problem. You need not choose either extreme; comparing them can reveal where learning will be most useful. Breaking a known goal into small steps helps, but when the goal itself is unclear, reconsider the problem before adding more tasks.
The skateboard-to-car MVP illustration showed why an early version should deliver some usable value and generate feedback. Building disconnected pieces of an ambitious final product may postpone that learning.
Choose one question or blocker, then describe observable success criteria. Examples discussed included a concise written vision statement, a functional MVP ready for customers by a target date, or completed marketing experiments with recorded results. These were examples to adapt, not identical requirements for everyone. Ask: what could a camera, microphone, or piece of code verify that shows progress?
Use your business project/work folder to give an AI assistant the blocker, desired outcomes, and relevant context. Ask for several possible approaches and refine them; you do not need to accept its first suggestion. The wardrobe example requested research across existing products, app stores, user discussions, and earlier failed approaches before deciding what demand to investigate next.
Find channels that already reach your customer
A channel is a route to customers. Rather than contacting everyone individually, look for existing organizations, events, communities, or referral relationships that already connect with the people you want to serve. A small successful engagement can create evidence and an introduction to the next customer. Paid advertising is one possibility, not the automatic first step.
Delegate outcomes and connect useful context
The wardrobe demonstration used a goal-based request: import clothing photos, fit the app’s upload limits, use its existing categories, and attempt to read brands and sizes. The agent found a duplicate, imported the unique images, and left unreadable details unknown. The broader lesson: specify the outcome and constraints, let the tool investigate implementation details, and inspect the result.
MCP—Model Context Protocol—provides a standard way for AI applications to connect with external data and tools. An API—application programming interface—defines operations software can request. These connections let an assistant retrieve context and perform supported actions rather than relying only on what you paste into chat. MCP introduction.
Examples included retrieving a client’s requirements from email and using Akiflow to discuss task urgency, time estimates, and notes by voice. Time blocking means assigning tasks space on your calendar rather than maintaining only a separate to-do list. Connecting a tool was encouraged for students comfortable doing so; purchasing Akiflow was not requested.
The photo-culling example extended this into a business experiment: test whether a tool can match human selection standards on a small sample, then expand cautiously. A conservative rule—keep uncertain photos—made a useful partial reduction in work possible without demanding perfect automation. The new local workflow was still experimental, not fully validated for high-volume use.
This illustrated the adjacent possible: solving one problem reveals the next opportunity. Name cards made automated tagging possible, suggesting a change to how future shoots might be captured. It also raised a buy-versus-build question: having a coding agent does not mean every problem needs a custom app. OpenVox, an existing transcription app, supplied capabilities that would otherwise have taken time to build.
Mentioned listening: The AI Daily Brief — “How to Build an AI-Native Company Today”. This was recommended context, not a newly announced required assignment.
Understand the release workflow
The demonstration traced a feature through a working branch, a commit, a push to GitHub, review, and a merge into main. A pull request packages proposed changes for discussion and review. A branch can be ahead of main without changing the deployed app. GitHub’s pull-request guide.
With the demonstrated Railway setup, pushing the merged changes to the connected branch triggered a deployment. The MCP feature still required remote hosting and authentication work; seeing code merged is not the same as demonstrating a complete working integration. Supabase Auth was explored as part of that work.
Before Tuesday, September 15
Leave this class with a plan of action for your chosen question or blocker. Put in reasonable effort to achieve your measurable outcome before Tuesday’s class, and be ready to explain what you aimed to verify, what you tried, and what happened. A thoughtful attempt that does not achieve the target is acceptable; failing to try is not. Start with a small, achievable goal.
The first round of presentations is planned for Thursday, September 17. Detailed format and grading expectations were not yet set in this discussion.
Also review the September 3 and September 8 summaries posted in Slack. Spend a few minutes on each and suggest useful changes in detail, organization, or links—or react to a suggestion you agree with.