Chat with Fundamentals – Video Demonstration
Version 1.0 · Released 7 May 2025 · Includes first-month development summary
This short video introduces Chat with Fundamentals, now available as a standalone desktop application. It has been separated from the original QuantCoderFS prototype, which is not yet released but will be made available as a distinct tool in the coming days.
First-Month Output
During the first month of activity on Substack, five executive summaries and two quantitative strategies were published, excluding retrospective material previously shared on Medium. Both Chat with Fundamentals and internal development builds of QuantCoderFS were used to support these outputs.
Ongoing Development and Final Application
A third and final application is in development. It will feature a lightweight chat interface built on a SmolAgents architecture, designed to coordinate deep research flows. This interface is already functional and will support tasks such as article retrieval, investment lead exploration, and potentially, stock screening. Once the three applications are in place, development will focus on their continuous improvement.
Planned Improvements
QuantCoderFS
A backtesting module will be added on top of the QuantConnect LEAN engine. Initial tests involving direct integration of LLM-driven code generation significantly increased latency, so the AI flow will remain decoupled from the execution environment.
Chat with Fundamentals
The application will transition from end-of-day (EOD) data toward delayed and intraday data. The main focus is to expand the available analytics. Currently, four modules are implemented; at least ten in total are planned. Each module is designed to answer a specific question relevant to quantitative or fundamental analysis. The planned modules include, list is non - exhaustive:
Cumulative return vs. SPY benchmark - CODED
Have I been compensated for taking idiosyncratic risk?
Identifies alpha drift and relative performance timing.Rolling 20-day annualized volatility (plus benchmark) - CODED
Is the stock currently volatile or stable?
Useful for adjusting position sizing and risk limits.Rolling 20-day Sharpe ratio
Is recent performance attractive on a risk-adjusted basis?
Supports rank-ordering across a universe of securities.Maximum drawdown (underwater plot)
How deep and persistent are losses?
Emphasizes path-dependent risk over standard deviation.Volatility cone / fan chart
Is current realized volatility high or low vs. history?
Relevant for options pricing and hedging strategies.Histogram of log returns with skew and kurtosis - BASELINE CODED
Are returns normally distributed or do they exhibit asymmetry?
Adds higher-moment context to simple volatility measures.Autocorrelation function (ACF)
Is there mean-reversion or momentum at short lags?
A quick diagnostic prior to signal development.Calendar heatmap of daily returns
When did significant moves cluster?
Helps identify seasonality or event-driven patterns.Rolling beta vs. rolling volatility
What is the balance between systematic and idiosyncratic risk?
Useful for identifying regime changes or diversification roles.Expected Shortfall (ES) vs. Value at Risk (VaR)
What is the shape of the left tail beyond the VaR threshold?
Aligns with modern risk mandates emphasizing tail sensitivity.
Longer-term, this framework may serve as the foundation for a machine learning pipeline for intraday nowcasting.
Closing Note
I would like to thank all subscribers for their interest and support — especially paid subscribers, and in particular those who joined as founding members. Your early engagement is instrumental in shaping the direction of this work.
I hope these tools can support your own research workflows, or at the very least, that some of the strategies and analysis shared here provide useful insights for your trading decisions.
Further updates will follow as development progresses.
Feel free to subscribe — or unsubscribe — depending on whether this work aligns with your trading style or current interests. Time is limited, and in an overwhelming data environment, relevance matters.