Skip to content
Theory to Things

The process

Your idea. Twelve flexible milestones.

Follow a shared map of capabilities while building your own project. A module may take several meetings, and familiar parts may move faster. Optional agent/API work and public release are never required to finish a strong local project.

Twelve milestones from idea to first version

  1. Milestone 1: Idea & scope

    Explore interests, problems worth solving, who the project is for, and whether it is feasible.

  2. Milestone 2: Requirements & UX

    Define a first version: the user experience, requirements, and what is deliberately out of scope.

  3. Milestone 3: Web foundations

    Build meaningful HTML and readable CSS, learning the foundations your project needs.

  4. Milestone 4: AI collaboration

    Ask for bounded help, review the proposed changes, and credit AI and human assistance.

  5. Milestone 5: First vertical slice

    Create one useful interaction that works from beginning to end.

  6. Milestone 6: Logic, state & validation

    Follow inputs through the code and make behavior predictable for valid and invalid inputs.

  7. Milestone 7: Data & safe APIs

    Model fictional data and understand the boundaries of optional service connections.

  8. Milestone 8: Agents & MCP

    Explore optional depth with simulated tools, permissions and the Model Context Protocol. A local mock is enough.

  9. Milestone 9: Debugging & testing

    Reproduce failures, check normal and edge cases, repair bugs and record meaningful evidence.

  10. Milestone 10: UX, accessibility, privacy & security

    Review who can use the project, what data it exposes, and how to improve the experience.

  11. Milestone 11: Publication review

    Choose to keep the project local or review consent, content rights, costs and hosting before sharing.

  12. Milestone 12: Portfolio & next steps

    Write a clear explanation, README and demo — with an honest account of AI assistance. Then finish, or keep going.

AI agents & MCP, explained plainly

Where a project benefits, explore programs that take bounded steps toward a goal and the Model Context Protocol, a way to connect AI tools to data and apps. Local simulations can teach these ideas without a live service connection.

Understanding over output

Learn to read, test and question AI-generated code. Record help and limitations, make an independent change, and explain how your project works. Tutoring is never a way to outsource graded assignments.

Age-appropriate tools and safe examples

Core learning uses local files and prepared examples with no student accounts required. Optional live tools must meet current vendor age rules and guardian permission requirements. For learners under 13, live AI interaction is instructor-operated.

Finish, pause or keep building

Keep your project files, testing evidence and a clear handoff. You can finish with a private demo, pause, or agree a new scope. Continuing is optional and nothing renews automatically. Grades, admissions, employment and income are not guaranteed.

Have an idea — or just a hunch? Let's shape it.

Let's talk about your idea