AI Collective Kansas City

CyberTrack finds AI-native students who verify model output under pressure—not the ones who trust it blindly. Missions run in Cursor with local Gemma; scoring is deterministic; the leaderboard ranks judgment, not fluency. The next generation of operators will not be judged by whether they can get an AI answer. They will be judged by whether they can verify it under pressure. CyberTrack is a tactical mission arena where players complete high-pressure technical incidents inside Cursor using local AI models like Gemma via Ollama. The web arena provides the mission briefing, progression map, scoring, after-action reports, XP/ranks, and leaderboard. Cursor is the actual mission workspace: players inspect evidence files, question the local model, write an answer artifact, and submit their run for deterministic scoring. Each mission simulates the conditions operators face when cloud AI is unavailable, untrusted, or inappropriate: incomplete evidence, time pressure, imperfect model guidance, and the need to make a defensible call. Players are scored on the human-AI skills that matter under uncertainty: evidence discipline, model skepticism, recovery from bad guidance, decision quality, and local/offline compliance. After each mission, CyberTrack generates an after-action report showing where the operator’s reasoning held up and where it broke. The prototype includes local Gemma-powered missions via Ollama, deterministic scoring, after-action reports, Season progression, XP/ranks, a mission map, and a public leaderboard. Cursor is the cockpit. Gemma is the local AI inference. CyberTrack measures what matters when the cloud goes dark and the stakes are high.