To delve into the real-world application of AI-powered drug discovery, an investment research institute formed a five-member team, none of whom had a professional background in life sciences. With a budget of $200,000 and a strict three-month deadline, they harnessed the power of Claude to orchestrate over 300 agents, utilized more than 10,000 hours of H100 computing resources, and developed a research workflow guided by 35 pages of detailed prompts. In the end, they successfully designed a groundbreaking protein that exhibits resistance to scorpion venom. The team generated over 310,000 protein design proposals, rigorously screened 211 candidate molecules for testing, and pinpointed 17 that successfully passed binding tests, covering four high-risk scorpion venoms. Some candidates specifically targeting Mexican scorpion venom displayed low nanomolar affinity and hinted at potential interactions with key functional regions of the toxin. However, the team did not achieve the goal of binding to five toxins simultaneously. Although the project has not yet resulted in a marketable drug, it demonstrates that AI is significantly lowering the barriers to scientific discovery. It enables small, non-specialist teams to undertake previously highly specialized early-stage protein design research using AI tools and outsourced experimental resources. This marks the dawn of a new era in AI-driven drug discovery, where a wide range of participants can collectively propel the field forward, and the cost of exploring niche early-stage problems is substantially reduced.