In one three hour session, teams of students design a real app, direct an AI coding agent to build it, test each other's work, fix what breaks, and demo it live to the room. Beginners welcome. The students do the thinking, and they take the app home.
Built and taught by Walt Jordan, who builds AI systems for businesses every day and brings the same tools into the classroom.
$499 per session, not per student. A room of thirty costs the same as a room of twelve.
Real session. Summer 2026, D.B.G.E.M. camp at Chattahoochee Technical College.
3 hours
One session, start to live demo
18
Bots designed by students in one afternoon
29
Arena matches run from those designs
100%
Of teams demoed working software
Students already use AI. What almost nobody has shown them is how to be the one giving the instructions instead of the one asking for answers. That is a different skill, and it is the one that will matter to them.
So we do not lecture. We put them in software teams, hand them a real tool, and give them three hours and a deadline. The room sounds like a startup by minute forty.
Pick by the time you have and the age you are serving. Programs often run the short game first to hook the room, then the full build later in the week.
The flagship. Teams design a real app on paper, then direct an AI coding agent to build it, test it, break it, and fix it. Every team demos working software at the end.
A startup-building game. Crews invent a product for the year 2035, brand it, survive a mid-game twist from the future, film a 30 second commercial, and pitch a panel of investors for a live valuation.
Students describe a battle bot in plain English. The system turns each description into a balanced fighter, then the bots fight in a projected arena with live commentary while the room loses its mind.
Every line on this list is something the AI cannot do for them.
How to decide what to build and say it in one sentence the whole team agrees on
How to write a five line spec by hand before anyone touches a keyboard
How to sketch the screen so four people are building the same thing
Why vague instructions lose, in a game where invincible gets you a slow, overheating machine
How to cut a big idea down to what ships today and what waits
How to direct an AI coding agent that plans, writes files, runs code and fixes its own errors, instead of chatting with a bot
The difference between a vague prompt and a precise one, shown side by side on the same task
How to read back what the agent built and check it against the spec
How to fix what breaks, one change at a time, testing after each one
How to hold a real software team role: Product Lead, Developer, QA Tester, Designer and Presenter
How to say no to feature creep when the clock is running
How to test honestly by trying to break another team's app on purpose
How to write a bug ticket someone else can actually act on
How to triage the bugs filed against your own work and decide which ones matter
How to demo working software live: what we built, what broke, how we fixed it
How to stand behind your choices when the room asks why you did it that way
And most importantly: what to do when you get stuck
This is the real running order of the three hour session, not a sales summary of it.
We build a small game live on the projector in about four minutes, then break it on purpose and have the AI fix it. The room goes quiet, then loud.
The difference between a chatbot and an agent that plans, writes files, runs code, and fixes its own errors. Vague prompts versus specific ones, shown side by side.
Students take real software roles and write a five line spec on paper. Nobody touches a keyboard until the Product Lead can say out loud what they are building.
Build the core of the app. Checkpoints at 15, 30 and 45 minutes. We coach prompts, we never type for them.
Every team tries to break the next team's app for ten minutes and leaves written bug tickets. This is the part teachers tell us they remember.
The Developer seat rotates so a second student drives. Fix the bugs that matter, add one polish feature, then stop and test.
Every team presents on the big screen. What we built, the hardest bug, the best instruction we wrote, and a live play through.
Only the Developer types. That one rule stops four kids from crowding a keyboard and forces the team to think and talk instead. The Developer seat rotates at the halfway point, so at least two students drive the agent.
The quiet kid who is great at finding problems becomes QA and is suddenly the most important person on the team. Teachers notice that one.
Owns the spec. Decides what gets built and in what order, keeps time, and says no to feature creep.
The only one typing. Turns the team's spec into instructions for the AI, then reads back what it built.
Tries to break the app after every change, writes real bug tickets, and verifies the fixes.
Owns the look, the names, and the demo. Sketches the screen before a single prompt is written.
It is the first thing a good parent asks, and the surveys back them up. Losing critical thinking is the number one concern American parents have about AI in school. We designed this session as the direct answer to that concern, not around it.
Nobody touches a keyboard until the team has written a spec by hand and said it out loud. If they cannot describe it in one sentence, the AI cannot build it, and they find that out fast.
Every team spends ten minutes trying to break another team's app and files written bug tickets. Then they have to triage their own. Judgment is the assignment.
In the bot game, every design is normalized to the same point budget. Write invincible and you get a slow, overheating machine. Precise thinking beats shortcuts, by design.
At the end, every student stands up and explains what they built, what broke, and how they fixed it. You cannot fake that part, which is exactly why we make it the finale.
Every image below is from a real session. No stock photography, no staged classroom.
A crew invented a device, branded it, and presented the 30 second commercial they built for it. They pitched it to an investor panel for a live valuation.
Teams described a fighter in plain English and the system built it. Then they defended their design choices in front of the room before the arena decided who was right.
Teams built and kept playable browser games, including an arcade road crosser and two endless runners, each with scoring, difficulty and a restart. Single files that open on any computer at home.
Every team left with the five line spec they wrote by hand, a bug ticket they filed against another team, and a prompting cheat sheet so they can keep building at home.
This is not a slide deck with a timer. The facilitator console, the arena, the scoring and the awards are ours, which is why we can adapt them to your room instead of asking your room to adapt to us.
Eight timed phases that add up to exactly 90 minutes, a live valuation board every crew can see, event cards that reward good decisions, and a mid-game twist from the future that forces every team to pivot. You drive it from a laptop and project it on the big screen.
When a team submits a bot, the system shows its work: which words it read, which stats those words moved, and what the team forgot to specify. Students immediately see that the description they wrote is the reason their bot is fast or fragile. Then they rewrite it and try again.
Bots fight in a shrinking ring while a commentator calls the action out loud. The room is on its feet. Underneath the noise, every match is a fair test of whose description was clearest, because every design is balanced to the same budget before it ever enters the ring.
Jordan & Borden served as the AI STEM workshop presenter for Developing Boys Into Gainfully Employed Men, a metro Atlanta nonprofit that runs a summer camp for young men in grades 7 through 12. The sessions ran across two camp weeks in June 2026 at the Chattahoochee Technical College Paulding Campus, alongside their leadership, trades and career exploration programming.
Students who had never written a line of code shipped working software the same day, then stood up and explained how they built it.
“Because of partners like you, our campers experienced far more than a traditional summer camp. They were immersed in an environment that inspired them to dream bigger, lead with confidence, and begin preparing for successful futures.”
Georgia has made computer science, including artificial intelligence, a high school graduation requirement, phasing in across the state over the next several years. Most districts and the programs that feed them are still deciding what that looks like in practice.
A single three hour session is the least expensive way to find out what your students actually respond to, and to have something real to point at when someone asks what you are doing about AI.
We keep your lift small on purpose. Most programs are running on volunteers and borrowed rooms, and we plan for that.
No laptops? The Bot Royal Rumble and Mission 2035 both run from one facilitator console on your projector, with students working on paper and phones.
$499per session
Per session, not per student. A room of thirty does not cost you more than a room of twelve, so you are never choosing between budget and how many kids get in.
Booking several sessions, or working from a grant? Tell us the budget you actually have and we will work with it. We have run these for nonprofits.
Check a dateThis is usually the first question a director asks, and it should be. Here is the honest answer.
No. This is designed for beginners. Students who already code are given the Developer role and pushed to read and question the code the agent writes, so they are stretched rather than bored.
All AI use is supervised, screens face the room, and an instructor circulates the entire session. Projects are pre-scoped to school-appropriate options, students never enter personal information, and every project has to pass what we call the show your principal test, which we state out loud at the start.
We check the network the day before. The Bot Royal Rumble and Mission 2035 both run without internet, and if the network fails mid session we switch to the paper spec and prototyping exercise until it comes back. We have never lost a session to this.
Ten to fifteen per instructor is the sweet spot for the coding workshop. Mission 2035 and the Bot Royal Rumble scale to a full room of thirty or more, because they run from one facilitator console on the projector.
The opposite. The whole session rewards the things AI cannot do for them: deciding what to build, describing it precisely, testing it honestly, and standing up to explain it. In the bot game we deliberately normalize every design to the same point budget, so a student who writes invincible gets a slow, overheating machine. Clear thinking wins, shortcuts do not.
499 dollars per session, and that is per session rather than per student, so a room of thirty does not cost you more than a room of twelve. That covers the instructor, the lesson plan, every printed handout, the facilitator software and the setup check the day before. If you are booking several sessions or you are a nonprofit working from a grant, tell us the budget you actually have and we will work with it.
Tell us what your program requires and we will have it to you before we set foot on your site. That normally means a certificate of insurance, background screening documentation, and a reference you can actually call. We would rather over-document than have you chasing us the week of the session.
We hand you the parent packet so you are not writing it yourself: a plain-English letter explaining what students will do, a disclosure naming the exact tools we use and what data they touch, and a media release if you want to photograph the session. Most parents are not against AI, they are against not being told. This solves that before it becomes a phone call.
Tell us what you are planning. We will come back with a straight answer on dates, group size and cost, usually the same day. If we are not the right fit we will say so.
