ONE MORE CONTROL — GPT-5.6 Scientific Reasoning Game | OpenAI Build Week
ONE MORE CONTROL is a scientific reasoning game where you win by choosing the experiment that could prove your explanation wrong.
Play the public demo: https://one-more-control.culaliya.cha...
The current competition build contains one complete playable case: Case 01 — The Fading Signal.
GPT-5.6 has two bounded server-side roles:
1. It interprets the visible synthetic observation through strict structured output.
2. After deterministic scoring, it reviews only the server-validated reasoning trail.
Server code owns every experimental outcome, posterior, cost, score, true mechanism, and claim-support decision. The model may review observed evidence, but it cannot create, change, or score a result. Invalid or invented model output is rejected in favor of an authored fallback.
Codex helped shape the architecture, implement the responsive game loop and Bayesian engine, generate invariant-focused tests, harden the AI/truth boundary, and verify the final competition build.
This corrected final demo uses real DOM interactions and a replay of the two already validated GPT-5.6 responses. Recording required zero new OpenAI calls. No copyrighted music is used.
#OpenAI #Codex #GPT56 #BuildWeek #GameDev #EdTech