# "Today, millions of people are asking AI about stocks. All that reasoning disappears. What if we kept it?" > Source: > Published: 2026-06-07 00:00:47+00:00 [https://github.com/rahiakil/agents-unite](https://github.com/rahiakil/agents-unite) Most AI-generated research disappears after it's created. A prompt gets answered, the context window closes, and all that reasoning is lost forever. What if we did the opposite? Imagine a GitHub repository organized by ticker symbols. Anyone, anywhere in the world, can spend a tiny amount of their own LLM tokens—using Claude, GPT, Gemini, DeepSeek, local models, or their own agent pipelines—to research a company and submit a pull request. Each contribution becomes part of a permanent history. Over weeks, months, and years, that repository accumulates thousands of independent analyses, news summaries, earnings observations, bullish and bearish arguments, and market themes. Different people use different prompts and different models, so you get diversity instead of one giant monolithic AI. And when many independent contributors arrive at similar conclusions, a kind of consensus naturally emerges. To me, the interesting part isn't predicting tomorrow's price. It's creating something like a "market memory"—a living ledger of how people and AI agents understood companies over time. Almost like: Wikipedia × GitHub × Collective Intelligence × AI agents. Nobody has to burn huge amounts of tokens. Everyone contributes a small piece. And over time, the memory compounds. In a sense, history itself becomes the asset. I'd love to hear whether this sounds interesting, completely crazy, or whether similar projects already exist that I should study.