Research¶
Notes from the lab. Short, dated, authored pieces on market microstructure, point-process modelling, deterministic replay, and the engineering that makes the Aetelier SDK reproducible.
The full articles live on the IteraLabs research blog — this page is the reference index: what exists, and why it matters to the platform documented here. These are not marketing posts and not API docs — they are the methodology behind the numbers: the testable claims we make about market data, how we validate them, and the cases where the honest answer is "the simple model is enough."
Reproducibility
Each post names the SDK crate, example, and version it draws from. Where a result depends on data, the dataset and the command that generated it are stated inline — run it yourself.
How we validate the Aetelier platform¶
June 16, 2026 · Methodology · 3 min read
A research platform makes two kinds of claims: claims about data ("these arrivals are self-exciting") and claims about itself ("the system does what the spec says"). We hold both to the same standard — a claim survives only if it is verified against a source of truth by an adversary trying to break it. This note is about how that standard is applied to the platform itself.
Continue reading on aetelier.xyz/research
When are crypto order arrivals self-exciting?¶
June 16, 2026 · Microstructure, Methodology · 5 min read
A Hawkes process makes a sharp, falsifiable claim about market data: that one event raises the probability of the next. It is a seductive model — order flow looks clustered, and a self-exciting process reproduces that clustering almost by construction. Which is exactly why fitting one and declaring victory proves nothing. This note is about the discipline that separates a real finding from a plausible-looking fit.