MoleculeMind, an AI protein design platform, has completed a Series A funding round totaling over $100M from investors including Bluebridge Capital, Pudong Venture Capital, COFCO Emerging Industry Fund, and others, with existing investors like Cathay Biotech also participating. The company unveiled MMDesign, an AI-driven de novo antibody design platform that achieved over 90% target hit rate on 12 real clinical targets and delivered picomolar-level affinity against the challenging TNFα target after testing only 14–50 AI-designed candidates. Its proprietary MMFold structure prediction engine outperforms Google AlphaFold 3 and all open-source models on the FoldBench antibody-antigen interface benchmark. MoleculeMind has built MoleculeOS, an AI-native protein engineering infrastructure based on its NewOrigin multimodal foundational model, which has been validated in industrial collaborations across drug discovery, green biomanufacturing, and novel materials. The company has established multi-layer commercial partnerships with top global pharma, synthetic biology firms, and chemical groups, demonstrating strong product-market fit and creating a closed-loop data flywheel from wet-lab-verified projects.
The article, prompted by rumors about OpenAI's potential IPO termination, discusses the engineering reality behind recursive self-improvement in AI. It argues that true recursive self-improvement is not a model modifying itself but a system feedback loop involving validation, tool chains, and data pipelines, where real-world experience is accumulated to make AI more reliable over time. This shift would transform a company's core assets from single models to feedback systems, evaluation frameworks, and task trajectories, potentially reducing the relevance of traditional funding and IPO. The analysis highlights that the key challenge is not generating outputs but verifying their correctness cheaply and reliably, and warns against confusing self-confirmation with genuine improvement. Ultimately, the competitive moat will be the learning slope from real-world feedback.
DeepSeek has closed its first funding round, raising over 50 billion RMB ($7.4 billion) at a valuation exceeding $50 billion, marking the largest single round in China's AI industry. The deal employs a unique transaction structure where all external investments, except from the national AI industry fund, are injected into a limited partnership managed by founder Liang Wenfeng rather than directly into DeepSeek. Investors face a five-year lock-up period with no voting rights, a condition aimed at attracting long-term capital aligned with Liang's open-source philosophy. Liang personally invested 20 billion RMB, while Tencent is considering a 10 billion RMB investment and CATL plans 5 billion RMB as the largest external backers. Additionally, the team required verification of all limited partners' identities behind investment funds to prevent unknown capital inflows.