axDyneIndependent research laboratory
A long-lived institution for producing reliable knowledge through rigorous, reproducible experiments in machine learning and adjacent fields. Every claim traces to an experiment, and every experiment is reproducible from its record. Negative results are included too, reported with equal weight.
Principles
01
Questions before answers
The research question is the real asset, sharpened before any code is written.
02
Hypotheses, not beliefs
Every hypothesis is committed to the record before the experiment runs, along with its limits and failure conditions.
03
Intellectual honesty
Negative results are published, not buried. They report mean and variance, never a flattering single number.
04
Long-term thinking
Built to outlast any tool or funding cycle. Plain Markdown in Git, open formats, no lock-in.
05
Institution over personality
The record is kept so anyone could pick up the work cold. Nothing essential lives in one head.
06
Reproducibility is not optional
Every result ships with the seeds, versions, and compute to reproduce it. What can't be re-run is a note, not a finding.