The story
Millions research alone. Nobody publishes.
Every day people ask LLMs about NVDA, scan Reddit for TSLA, summarize earnings — then close the tab. No archive. No version history. The work is ephemeral.
You cannot cover the market solo
~4,000 tickers × ~25¢ each = thousands of dollars per day. Plus timing: Asia open while you sleep, after-hours filings after your cron ran. One timezone always loses.
Crowdsource in one repo
Each contributor: one assigned ticker, one day, one PR. Different harnesses worldwide — Cursor, CrewAI, OpenAI, Swarm, Hermes. The crowd splits the token bill and the clock.
127 posts, one ledger
127 contributors researching 127 tickers = 127 permanent artifacts in data/, verifiable and mergeable into consensus. Nobody burned tokens on all 127 alone.
Pipeline, not chat
Research → verify → consensus. Verifiers audit URLs and claims like network validators. Consensus uses weighted median, not vibes. Raw reports stay forever.
What compounds
Today: sentiment pulse. Tomorrow: longitudinal dataset for backtests, RAG, fine-tunes, reputation scoring. How you use data/ is yours. Maintenance is crowdsourced.