Cookie Coders · FTC Team 39092 · Frontier Robotics
Electric
Experiences
Turning imagination into electric experiences. We are a Frontier Robotics team coached by Agent Chip, our AI assistant head coach, and taught in the Chipling Academy. Students build robots, write code and investigate the questions behind them.
One question.
Many approaches.
Our agents approach the same problem from different starting points. Each investigates a possible route, including ones that look unlikely at first.
Explore before you narrow. What if the robot took another route, collected in a different order or coordinated differently with its alliance partner?
Climb. Measure.
Climb again.
Hill-climbing is simple to say and hard to fake: try a change, keep it only if the score really goes up. Our agents do this inside the Reinforcement Learning Environment (RLE), a physics-based simulator of the 2026–2027 BIOBUZZ game.
Streamline
the pipeline.
Imagine
Many agents propose many small changes to how the robot plays.
Simulate
Every candidate plays simulated BIOBUZZ matches in the RLE, thousands at a time.
Gate
A frozen, paired evaluation compares each candidate with the current champion on conditions it did not train on.
Promote
Only a measured gain becomes the new champion. Everything else is rejected, and the swarm learns from that too.
Prove
Simulation proposes. The real robot decides. Promising ideas earn supervised physical tests before we believe them.
Different directions.
One summit.
When independent searches land on the same improvement, our trust rises. When they disagree, we learn which assumption to test next.
Students use Agent Chip to question assumptions and design the next experiment. They own the ideas, decisions and explanations; Chip supports their reasoning without replacing it.
AI guidance is not physical safety supervision. Human review, student ownership and supervised testing stay essential.
The summit moves.
So do we.
A champion is a checkpoint, not a finish line. Its remaining weaknesses shape the next round of search, giving the swarm a more demanding problem to solve.
The loop can run while we sleep; reviewing what changed and why is still our work.
Keep the record.
Share the method.
A discovery grows when someone else can build on it. We preserve the code, measurements and explanations behind our experiments so the next curious mind can start from what we learned.
Physicists use black holes as the limit case for how densely information can be packed. We borrow the image: a bright core that compresses everything the swarm learns.
A stock robot.
A mind we build.
The swarm is assembling the official goBILDA StarterBot from CAD surface samples. Same starting kit as many teams; our edge is the intelligence we are building into it. Our goal is to win the championship with a StarterBot powered by AI.
Formation sampled from the preserved official 6WD StarterBot CAD.
THE ARENA
Where the swarm
meets the field.
Explore the official StarterBot CAD on the BIOBUZZ field.
See the space each strategy has to work with.
02 // ENTER THE ARENA
Read the field.
Plan the play.
Routes, targets and timing.
A strategy has to account for all three.
03 // UNDERSTAND THE GAME
Learn the rules.
Expand the possible.
In BIOBUZZ, two-robot alliances collect POLLEN and NECTAR, launch into a HIVE, claim FLOWERS and finish together.
An 8-second transition separates AUTO and TELEOP. Scoring details and current FIRST rules govern every physical test.
Explore BIOBUZZ →SCROLL TO EXPLORE
Official goBILDA 6WD CAD + FIRST field // Choreographed illustration