General Catalyst Institute was created not out of necessity, but as a strategic choice to serve as a bridge between founders and policymakers. This helps accelerate startup growth and positions the VC as a partner to governments in transforming key industries like healthcare.
The healthcare AI startup uses a unique architecture where 30 LLMs supervise one patient-facing LLM. This is combined with extensive "output testing" by thousands of clinicians to ensure the AI's responses are safe before deployment, a more rigorous method than just evaluating training data.
Hippocratic AI's founder identified a strategic sweet spot for AI applications. By focusing on clinical but non-diagnostic tasks (within a nurse's scope), the company avoids the highest safety risks and regulatory hurdles of being an "AI doctor" while providing more value than simple administrative automation.
The most successful AI applications are not those doing existing tasks cheaper. Instead, they enable entirely new, high-volume interventions that were previously cost-prohibitive, such as proactively calling 50,000 at-risk people during a heatwave.
Specialized, vertical AI models allow for comprehensive "output testing" for a specific use case, making safety verifiable. This is nearly impossible for horizontal, general-purpose models like ChatGPT. This suggests AI will be adopted safely one vertical at a time.
Hippocratic AI's CEO explains that open-weight models are crucial for startups, not just for control, but for economic viability. Using open models makes their product cost $9/hour, versus a projected $105/hour with proprietary frontier models, enabling new business models and fostering competition.
General Catalyst's Institute President states that for VCs, the founder's vision and drive are the number one consideration, even in heavily regulated sectors like healthcare. Regulatory complexity is treated as a secondary detail that a strong founder can overcome, not a primary deal-breaker.
Companies won't use AI's productivity gains to fire most of their staff and maintain output. Instead, they will keep their teams and aim to 10x their business goals. This shifts the labor narrative from replacement to expansion and new job creation.
A one-size-fits-all AI regulation is flawed. Like employment laws exempting small businesses, AI rules should be tiered. Smaller models present less risk than frontier models, and startups need exemptions from burdensome compliance to foster innovation and prevent regulatory capture by large incumbents.