AI / quantitative systems
Battle of the AI Traders
Complete
Three LLM agents race separate Alpaca paper accounts from $100k toward $1M, with scheduled cycles, EOD memory, and an ML signal pipeline.
Problem
Compare how OpenAI, Groq, and Gemini trade under the same timed paper-market constraints — without conflating research with a consumer trading product.
Objectives
- ▪Three independent agents on three Alpaca paper accounts (stocks + options)
- ▪14-session weekday sprint from $100k toward $1M with premarket, hourly, and EOD cycles
- ▪Daily ML bar-feature training feeding agent prompts and tools
- ▪Dashboard for equity progress, charts, and EOD archive (local or Firebase)
Architecture
Python trading engine drives scheduled market sessions; LLM agents call Alpaca independently; ML signals inject into prompts; Firebase Functions + Firestore host the dashboard and schedulers for personal use.