How We Built WebLens: Creating an AI-Powered Hypothesis Generator

The authors, from GrowthBook, developed an AI-powered hypothesis generator to analyze websites and suggest actionable changes for desired outcomes. They partnered with Firecrawl for web scraping, used Google Gemini Pro for large language model prompting, and implemented multi-shot prompting techniques to improve hypothesis quality. The tool generates DOM mutations in JSON format, allowing users to create new experiments in GrowthBook. Despite challenges such as context window limits, the authors have successfully developed a novel approach to online controlled experiments, providing a rich roadmap for further innovation with generative AI. By analyzing websites through a lens and enhancing focus, WebLens aims to improve product usability and enhance productivity, making it an exciting example of how companies are leveraging generative AI in various industries. The tool is now available at https://weblens.ai, inviting users to test it and provide feedback via the community slack.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.