vibe coding business studio

Inspiration

VibeCoding Business Studio began with my children.

They had seen me build software with AI and wanted to learn how to use it themselves. One day, they asked me a simple question:

«“Can we use AI to create our own game?”»

At first, I thought the project would only be about giving them a safe place to experiment with Codex. I created a private development server and helped them begin turning their ideas into something they could actually build.

But as we worked together, I noticed something important.

The difficult part was not only writing code. They also needed to learn how to explain an idea clearly, decide what to build first, divide responsibilities, verify that something worked, and avoid adding endless features before finishing the first version.

They wanted to build with friends, so new questions appeared:

• Who owns the project roadmap?
• What should the designer be responsible for?
• What should the tester verify?
• Who can propose new features?
• How should business ideas be discussed?
• Which decisions need help from a parent, teacher, or mentor?

Little by little, the original game experiment became something much broader: an AI-guided learning environment where students can build projects they care about while practicing coding, writing, planning, teamwork, leadership, and business thinking.

That became VibeCoding Business Studio.

What It Does

VibeCoding Business Studio guides a learner or small team through a structured project-building process.

First, students describe what they want to create, who it is for, what problem it solves, and how much time they have.

The Studio then challenges the scope. Instead of simply praising every idea, it helps the learner identify the smallest useful version that can be built and tested. It separates what should be included now from what should be postponed.

Next, the project is divided into milestones with:

• a clear objective;
• a definition of done;
• specific activities;
• a learning concept;
• a product or business question;
• evidence that can prove progress.

Students can also define roles such as project owner, designer, tester, developer, or business lead. Each role has responsibilities and decision boundaries, helping friends collaborate without everyone changing the roadmap independently.

Before asking Codex to implement something, students write what they want to build. The Studio helps them improve the clarity, grammar, specificity, context, and verification criteria of their request.

Codex can then act as both a technical build partner and a learning coach. It can help implement, explain, debug, and document the project, while returning important decisions about scope, priorities, risk, privacy, cost, or business strategy to the learner.

At the end of a milestone, the student completes a reflection and generates a learning report showing:

• what was built;
• what the learner decided;
• what Codex helped with;
• which roles participated;
• how the result was verified;
• what was learned;
• what remains unclear;
• what should happen next;
• where adult or mentor guidance is still needed.

How We Built It

We built VibeCoding Business Studio with TypeScript and Vite as a browser-based application.

The current prototype includes:

• a project brief workflow;
• AI-assisted scope coaching;
• milestone generation;
• prompt-writing feedback;
• team-role documentation;
• business and product questions;
• structured Codex prompts;
• learner reflection;
• local browser persistence;
• downloadable Markdown learning reports.

Codex was used as a development partner throughout the project. It helped us transform the original idea into a structured product, rethink the game-centered direction, design the learning workflow, implement the application, debug issues, validate builds, and prepare the documentation and judging experience.

The application also includes an optional server-side OpenAI coach endpoint. The API key remains on the server and is never exposed in the browser. When that endpoint is unavailable, the Studio uses transparent local coaching rules, allowing judges and learners to continue testing the full workflow.

Challenges We Faced

The biggest challenge was separating the example project from the actual product.

The repository originally focused on building a game. Over time, we realized that the game was only the learning context. The real product was the process around the project: defining the idea, improving the writing, reducing the scope, assigning roles, making decisions, verifying the result, and reflecting on what was learned.

Another challenge was balancing AI assistance with learner ownership.

It would have been easy to create a system where the student enters one sentence and AI makes every decision. However, that would produce software without necessarily producing learning.

We designed the Studio so AI can accelerate execution without silently taking control of the project. Decisions about scope, priorities, business, privacy, publication, money, and risk remain visible and attributable to the learner or mentor.

We also had to keep the prototype realistic. Features such as authentication, real-time collaboration, classroom administration, automatic repository access, payments, and advanced role permissions were intentionally postponed. We focused first on validating the central learning loop.

What We Learned

We learned that AI literacy is not only about knowing how to write a prompt.

A learner also needs to understand:

• what they are asking the AI to do;
• whether the request is clear;
• which decision belongs to them;
• how to inspect the result;
• how to prove that it works;
• what assumptions the AI made;
• what should not be built yet.

We also learned that project-based learning can teach much more than programming.

A student building a game about running a sneaker company can practice coding, but also writing, product design, marketing, customer value, operating costs, teamwork, leadership, and business decision-making.

The most important lesson was that students do not need to write every line of code themselves to own a project. But they should be able to explain what was built, why it was built that way, how it was verified, what they decided, and what should happen next.

What’s Next

Our next step is to test the workflow with more learners, parents, teachers, and mentors.

Future versions may include:

• mentor comments;
• parent and teacher summaries;
• reusable project templates;
• more structured role descriptions;
• role-based collaboration;
• project activity history;
• automatic connection between Codex sessions and learning evidence;
• age-appropriate learning rubrics;
• privacy-preserving classroom storage.

The long-term vision is to help students build projects with AI without surrendering their curiosity, judgment, creativity, or responsibility.

AI can help build the project. VibeCoding Business Studio helps the student own it.

Built With

• vibe-coding