Tencent's Hunyuan Research Agent version 2.0, Hyra 2.0, made its debut Monday, claiming to build reusable tools—trained protein models, optimized quantum codes, and verified mathematical proofs—rather than discarding methods after solving problems. Multiple editorial outlets independently confirmed the announcement within hours. Coverage accounts state Hyra 2.0 is now participating in training Tencent's next-generation Hunyuan foundation model, the same model that powers Hyra itself—a recursive self-improvement loop that more than 20 leading researchers, including Geoffrey Hinton and Yoshua Bengio, warned of in September 2026. The article notes four facts matter more than benchmark numbers: the benchmarks are Tencent's own claims with no independent replication published; the agent code and methodology have not been released; Tencent operates under China's National Intelligence Law, whose Article 7 legally compels cooperation with Chinese intelligence authorities regardless of where data is stored; and autonomous research agents have a documented hallucination problem that Hyra's engineering does not structurally solve. Two protein models—HyInv and HyPLM-Flash—are on GitHub with weights, inference code, and technical reports. A quantum error-correction code constructed by the agent has been entered in the Unitary Foundation qLDPC Challenge leaderboard. A mathematical proof was checked by mathematicians whose names Tencent has not released. Hyra 1.0 launched July 21, 2026.