How does information cross time zones?
When one market closes, another is still pricing the same world. We study how information is handed from one session to the next.
Quantitative research / US / Hong Kong
Markets areliving systems.
Dnalyaw is an AI-native quantitative trading lab focused on US and Hong Kong markets. We bring signal research, execution, and independent risk controls into one continuous validation process.
Component
Component
Component
Signal intelligence
Adaptive, long- and short-horizon, multi-mechanism
Execution science
Costs and fill conditions in the research loop
Machine governance
Risk veto remains absolute
Evidence memory
Link research, risk, and execution
What we study
We work on questions,not recipes.
A recipe stops working when the market changes. A good question keeps producing research. These are the questions our work returns to.
Who has to trade?
Some trades are chosen; others are forced by mandates, limits, and calendars. Forced trading leaves footprints.
When the market changes state, what should change with it?
A market is less one system than a sequence of them. We study how to recognise a change of state, and what must stay fixed when it happens.
Is consensus a source of risk?
When many models reach the same conclusion, a trade becomes a position someone must eventually unwind. We study crowding as fragility, not confirmation.
Which mechanism left this footprint?
Similar price paths can have different causes. Every explanation is a hypothesis until it survives costs and time.
Can execution learn without taking risks it cannot see?
Placement, timing, and capital efficiency as a learning problem, inside limits that learning cannot move.
Questions describe what we study. They are not claims about strategies in live trading.
The mathematics behind the questionsPrinciples
Markets change.Principles do not.
Models are revised, retrained, and retired. The rules that govern them are not.
Risk holds an absolute veto.
No signal, model, or person can trade around it. A veto is not a parameter to optimize.
The question comes before the data.
Hypothesis, sample, and test are written down before any result is seen.
Costs count from day one.
A signal that cannot pay its way through real frictions is not a signal.
Failures are kept.
Rejected ideas are recorded with the reason they failed. They are evidence too.
Machines do the work. They do not hold the keys.
AI enumerates, verifies, and monitors at a scale no team can. The authority to act stays behind controls.
Research notes
Thinking,written down.
Essays on markets, risk, and learning systems.
How Agents Run the Research PipelineHow research rules become contracts that agents can execute, why every rejection is a first-class output, and which decisions must stay with people.Two Languages of Risk ControlSizing decides how large a bet should be; safety decides when to stop. They are different control systems, and merging them is the most dangerous mistake in risk design.The Geometry of AlphaWhat compounds is not a recipe but the factory that keeps producing, validating, and retiring signals, and the geometry that explains why many weak, independent signals beat one strong one.
Written by Wayland Zhang, founder of Dnalyaw. About the author ↗
All notes