AI for sCO2 power cycles: in this video we walk through sCO2RL, an open-source deep reinforcement learning framework for autonomous control of supercritical CO2 Brayton cycles recovering steel plant waste heat. Using an OpenModelica FMU, a 7-phase Gymnasium curriculum, PPO with Lagrangian safety constraints, and GPU-accelerated surrogates, the controller outperforms a Ziegler–Nichols PID baseline while maintaining strict safety limits at the compressor inlet. We also cover deployment to TensorRT for sub-millisecond edge inference and share practical engineering lessons from debugging FMU-based RL pipelines, with all code and artefacts released under the MIT licence at https://github.com/SharathSPhD/RLpower. Paper: https://zenodo.org/records/18524794