Research infrastructure · Quantitative finance first
AI for research.
Second Order turns research ideas into experiments, and experiments into persistent knowledge. We build AI systems that formulate hypotheses, work with data, run experiments, validate results, and remember what they learn.
- Hypothesis
- Data
- Experiment
- Validation
- Knowledge
01The research loop
One continuous process, from idea to knowledge.
Every capability we build serves a single loop. Each step produces something the next step can use, and the last step feeds the first.
- 01
Hypothesis
A precise, falsifiable statement about a market, a dataset or a method.
- 02
Data
Datasets located, cleaned, aligned and inspected before anything is tested.
- 03
Experiment
Research code written and executed: signals, statistics, backtests.
- 04
Validation
Results stress-tested for leakage, overfitting and regime dependence.
- 05
Knowledge
Findings, failures and artefacts recorded, so the next question starts further ahead.
02What we build
Systems that do research, not only describe it.
Language models can already write about research. Second Order builds systems that carry it out: they reason over a problem, reach for tools and data, write and run code, check the result, and keep what they learn.
Reason over research problems
Decompose a question into testable parts and decide what evidence would settle it.
Use tools and datasets
Query, transform and inspect data through real tools rather than descriptions of them.
Write and execute code
Produce research code, run it in an isolated environment, and read the output back.
Run experiments
Backtests, statistical tests and ablations, executed reproducibly with tracked parameters.
Validate hypotheses
Check results for leakage, multiple testing and fragility before they count as findings.
Remember previous research
Keep hypotheses, results and failures as structured knowledge that later work builds on.
03First domain
Quantitative finance, first.
Systematic finance is a demanding place to start: noisy data, adaptive markets, and a long record of results that fail to replicate. It rewards exactly what the research loop enforces: careful data handling, honest validation, and memory of what has already been tried.
- Market data
- Signal research
- Statistical analysis
- Backtesting
- Portfolio research
- Systematic strategies
Second Order is research infrastructure. It is not an investment product, does not provide investment advice, and makes no claim about financial performance.
- Hypothesis
- Dataset
- Experiment
- Result
- Failed approach
04Research memory
Research that accumulates.
A useful discovery should not disappear when a conversation ends. Second Order keeps a persistent, structured record of the research itself, so every new question starts from everything already learned.
- Hypotheses
- what was asked, and why it seemed worth asking
- Experiments
- code, parameters, data versions, execution traces
- Results
- what held, under which conditions, with what confidence
- Failed approaches
- what did not work, so it is not tried twice
- Useful datasets
- where the data lives and what its quirks are
- Relationships
- how each idea connects to the ones before it
Private preview
Explore the research environment
The demo shows how Second Order runs the research loop end to end. Access is by password during the private preview.