Can AI assistwith quantitative trading research?
Not as a trader. As a quantitative researcher.
Building on top of the foundation, we let AI conduct research, test ideas, and learn from failure.
Can AI help develop profitable algorithmic strategies, as researcher?
Past AI trading experiments asked whether a model can look at a market and make a right trading decisions. AlphaStone asks a different question: can AI do the work of a quantitative researcher and help developing successful algorithmic strategies that remain consistent and testable?
AlphaStone began in 2021 as a personal attempt to build systematic strategies that could beat the market. After years of building the data and backtest infrastructure, test-time compute and reasoning-model breakthroughs made AI capable of doing meaningful generative research inside that framework.
AI research
Human review
Live algorithmic trading
Research results
Published model output is shown separately from the complete research run history. The research log reports every available run in chronological order without promoting a subset.
Calendar year returns
Return table
AI researches and proposes candidate signals and strategies.
Human validation tests candidates out of sample and decides what advances.
Live algorithms execute approved strategies systematically.
The research loop continues with each iteration.
AI-automated research
Under real-world constraints.
Event-driven, multi-timeframe
Research notes
Building on top of the foundation.
2021
The project begins
AlphaStone started as a personal project to develop a systematic approach that could beat the market.
2022
Live testing begins
The work moved beyond backtests and into live testing, where execution and risk became part of the research.
2023–2025
The foundation develops
The research framework was rebuilt across model generations. Failed experiments helped narrow what was worth pursuing.
2026
AI joins the research
AI became part of the research workflow, helping propose and test ideas within the existing framework.