Building an Alternative to Examine сom Using AI

I'm on a mission to build an alternative to Examine.com, and let me tell you—it’s no walk in the park!

Many of you might already be familiar with Examine. It’s a massive resource for supplement research, focusing on various health areas, scouting research papers that mention related health outcomes, and linking those to different supplements. This process is incredibly labor-intensive, so it’s understandable why they charge a fee for access. However, with the power of AI, many parts of this process can be automated, and that's exactly what I'm doing.

Here’s a glimpse into the research linked to Reduced Body Weight:

https://pillser.com/health-outcomes/reduced-body-weight-158

I’ve indexed thousands of research papers, extracting insights that connect these studies to various supplements available on the market. My long-term vision is to create a one-of-a-kind supplement store where every product is linked to scientific research. Unlike Examine, I plan to make all research summaries publicly accessible and focus on generating affiliate revenue from related supplement sales.

The biggest challenge I face is ensuring data accuracy. Given the complexity of these topics, I’m currently sticking to demonstrating a link between studies, health outcomes, and substances. Users can then explore the actual research papers to build confidence in their decisions. However, as AI models advance, I aim to expand this into a comprehensive insight engine.

AI and LLMs are crucial in this process. Finding research papers is straightforward thanks to sites like PubMed. I use a combination of different API services (varying in cost and speed) to first scout for potential mentions in the research (fast and cheap model), then validate the relevance (top-tier models), and finally verify the accuracy of the summary versus the excerpt using another model. The idea is that the first model may make mistakes, the second filters out false positives, and the third acts as a final safeguard.

I’m passionate about this problem domain, particularly data normalization and the applications of LLMs to solve these issues. This project has become a hobby that I foresee being a money-losing venture for quite some time, but I believe there’s a good chance that Pillser could become the go-to site for people looking to buy supplements, thanks to its unique blend of science and inventory.

I’m probably a few months away from having the complete database, but I wanted to share this for early feedback. The site is live at https://pillser.com/, and you can already search for different health goals and associated research right from the landing page.

Would love to hear your thoughts!