I looked at some current AI systems, and the P123 AI process was closest to what I might use:
1. Create universe. Select target. Add features. ...
4. Evaluate hundreds of models.
5. Create predictors.

This is not a LLM, but is multiple regression using AI methods. The community could contribute: universes, features, models, and criteria for benchmarks. Backtesting could be done with either P123 or GTR1.

An alternate path would be to collect academic papers and train a LLM.

=== current AI systems ===

Deepvue AI: Uses natural language prompts to combine fundamental metrics (like EPS or sales growth) and technical criteria into custom screens. About $500/year, but no backtesting. What themes are hot right now? deepvue.com - Stock market theme tracker

YCharts (Y Agent): Features an assistant that builds interactive financial screens, visual tables, and charts based on company SEC filings and fundamentals. Screen for S&P 500 stocks with a P/E under 20 and revenue growth over 10%.

Horizon.Trade: A purpose-built "agentic trading" platform where you describe a macro or fundamental strategy in plain English. The agent translates your text into a strategy, backtests it against historical data, and gives you a clear performance profile without requiring you to code. About $1000/year for 10-year backtests.

TrendSpider (Sidekick AI): TrendSpider's Sidekick AI agent acts as an on-demand quant analyst. It can simultaneously read corporate filings, build scanners, and process fundamental-to-technical backtests via natural language.

Trade Ideas (Holly AI): Holly runs millions of backtests every night across the entire US equity universe — scanning thousands of stocks, identifying high-probability setups, and generating entry and exit signals with specific stop-loss and profit targets by the time markets open.

Machine learning workflows define which phases are implemented during a machine learning project. The typical phases include data collection, data pre-processing, model training and refinement, evaluation, and deployment to production. Check out the AI Factor User Guide for how Portfolio123 implements these steps.
portfolio123.customerly.help - AI workflow

=== links ===
AI Hedge Fund Stock Screening Agent
v7labs.com - AI hedge fund stock screening agent

16 Best AI for Stock Market Analysis and Trading in 2026 (Ranked by Use Case)
assistents.ai - Best AI for stock market analysis and trading

Community intelligence? The other focus is to create RAG data of examples and Q/A. It's very time consuming for us, and not diverse enough. It might be worth setting up an easy way for anyone to contribute Q/A, and examples.
community.portfolio123.com - Say hello to alphanaut our new AI agent