Tencent Hunyuan has released Hyra 2.0, an automated scientific research framework designed to enable AI agents to develop and upgrade reusable scientific models and tools rather than solving problems directly. The system trained HyInv, a protein inverse folding model with only 1.96 million parameters that generates at 57 times the speed of ProteinMPNN. Hyra 2.0 also produced HyQEC, a quantum error correction tool that constructed a quantum code with encoding efficiency approximately 48% higher than the previous top entry in the qLDPC Challenge. Additionally, its mathematical proof workflow HyProver completed proofs for problems including Shing-Tung Yau's circle action conjecture, verified by mathematicians, and raised the XiangShan RISC-V processor's score per GHz by 5.79%. Related results will be made public on GitHub.