</> BuildLab Scripted demo · no real AI · no data collectedScripted demo

A concept from G-Force Creative Group

Learn to direct AI. Build things that help people.

BuildLab is an idea for a hands-on space where tweens, teens and adults learn to plan, ask, check and improve with AI by building small apps for real people around them. Nothing here is a product yet. It's a demo of the experience we want to test.

Try the 3-minute demo Open the full walkthrough
coach-session.demoscripted
Student

I want to make an app so my grandma's neighbors can ask for help with groceries.

Coach

Great idea. Before we build, who will use it, and what's the first thing they need to do?

Student

Neighbors post a request. Volunteers see the list and pick one.

Coach

Good. That's two users and two actions. Let's write that down as a plan.

Scripted · this is what a coaching moment could feel like

Kids already use AI.

Asking it to do everything isn't the same as learning how things get made.

BuildLab teaches the loop.

Plan → Ask → Inspect → Test → Fix → Explain. The student does the thinking; AI does the typing.

Every project helps someone.

Missions are small apps for a neighbor, a classroom, a club or a family, so the "why" is real.

Try it

Five steps. One real little app.

Pick a mission, write requirements, watch a simulated build, give feedback and explain what changed. Every coach message is scripted. The Neighbor Help Board in the middle is real and works.

Neighbor Help Board mission Scripted demo · no real AI · no data collected
  1. Mission
  2. Plan
  3. Build
  4. Improve
  5. Explain

Step 1 · Mission

Pick a mission

Every BuildLab project starts with a person who needs something. Only the first mission is playable in this demo.

Everything above runs in your browser. Nothing is sent anywhere, and nothing is saved.

Still undecided

Where could AI fit?

We haven't decided how much AI BuildLab should use, or where. Here are the four roles we're weighing. Each one changes what the student does, what the AI does, and what it costs to run safely.

Our current lean: start with Coach + Builder on rails, and only add Open studio once we see students actually learning.

Who it's for

The same loop, three starting points

Tweens · 10 to 12

Playful and supervised

Sample mission
A Food Drive Tracker for the school.
They practice
Planning in small steps, asking clearly, spotting what changed.
Supervision
An adult sits close. Under-13 privacy work is extra, so this group comes after the pilot.

Teens · 13 to 17

The first group to test

Sample mission
The Neighbor Help Board, as you just built it.
They practice
Precise instructions, testing old and new behavior, explaining a fix.
Supervision
Laptop or Chromebook, at home or in a club. Ages 12 to 15 are the first segment we'd test.

Adults

Parents, teachers, career-changers

Sample mission
A volunteer shift planner for a club.
They practice
The same loop with bigger missions, so they understand what kids or jobs ask of AI.
Supervision
They manage accounts for kids and see projects and skills.

How it could work

A loop for the student, guardrails underneath

The loop

  1. PlanDecide who uses the app and what it must do.
  2. AskWrite an instruction the AI can follow.
  3. InspectRead what changed before accepting it.
  4. TestTry the new feature and the old ones.
  5. FixNarrow down the problem and ask for one small repair.
  6. ExplainSay what changed and why, in your own words.

Then back to Plan with the next idea.

A simplified architecture

Student browserLearning screens and project preview. Never holds AI keys.
Application APIAdult controls, limits, curriculum and AI orchestration. Provider keys stay on the server.
Isolated preview sandboxRuns the student's app. No secrets, no network, strict time and memory limits.

Simplified. A real build needs a threat model and stronger isolation than a browser sandbox alone.

  1. Stage 1 · you are hereClickable conceptThis page. Scripted, no AI, almost no hosting cost.
  2. Stage 2Guided pilot3 to 5 missions, a limited AI tutor, adults supervise.
  3. Stage 3Student appAccounts, parent controls and progress tracking.
  4. Stage 4Open builderLater, and only if learning is proven.

What we'd measure

We don't reward long prompts or heavy AI use. We reward:

  • Identifying requirements before generating anything.
  • Breaking a change into a small task.
  • Writing an instruction with a testable result.
  • Checking new and existing behavior.
  • Spotting a planted bug and narrowing down the cause.
  • Explaining a fix in their own words.
  • Fewer unnecessary retries.

Safety, up front

Built for young people from the first sketch

  • Adult-managed accessGuardian or school consent where required. A typed birth date proves nothing.
  • Data minimizationSkip names, photos, location and tracking. Limit retention and offer deletion.
  • Hint-first, age-appropriate AIFiltered requests and responses, with hints before answers.
  • Sandboxed code executionStudent apps run apart from accounts and secrets, with network and resource limits.
  • Private by defaultNo child-to-child chat and no open public publishing at launch.
  • Labeled AI and simulated contentKids are encouraged to check results, not treat answers as facts.

This checklist is a plan, not legal advice.

Get involved

Want to help shape this?

The best feedback is what a kid actually does, so watch before you ask.

Try it with a kid or a classroom and tell us what they did, not just what they said.

info@gforcecg.com