Time Machine for Stock market intel

Agentic way to crowd source market intelligence at scale

Open Source

Analytics

GitHub

Visit WebsiteSee on Product HuntGithub ⧉

Hi everyone, I built Open Source Alpha(https://rahiakil.github.io/agent...) because I kept noticing something that felt fundamentally wasteful. Every day, millions of people ask Claude, GPT, Gemini, DeepSeek, and local models questions about companies and markets. Valuable insights are generated, used once, and then disappear forever inside a context window. Tomorrow, someone else asks the same questions, and the entire process starts over again. Software engineers solved this problem decades ago with version control and open collaboration. So I started wondering: What if market intelligence worked the same way? Open Source Alpha is an experiment in collective intelligence. Anyone can contribute using whatever tools they prefer. Some people may use Claude, others GPT, Gemini, local models, LangGraph agents, or custom workflows. Contributors spend only a small amount of their own tokens, summarize what they discover, and submit their findings through GitHub pull requests. Over time, thousands of these small contributions accumulate into something much larger than any one person or model could create alone. The goal is not to predict tomorrow's stock price. The goal is to build a long-term memory for markets. A place where reasoning, observations, earnings summaries, industry trends, and bullish and bearish arguments are preserved instead of forgotten. Because contributors use different models and approaches, diversity becomes a strength. And as independent analyses begin to agree, consensus naturally emerges. In many ways, I think of it as: Wikipedia × GitHub × AI Agents × Collective Intelligence. History compounds. Memory compounds. Perhaps intelligence should compound too. I'd love to hear your thoughts, suggestions, and criticisms. I'm particularly interested in hearing how people think contributor reputation, consensus, and long-term historical accuracy should be handled. Thank you for checking it out. https://rahiakil.github.io/agent...