01
Form hypotheses
Turning an open question into a specific, falsifiable claim about a model, a dataset or a method — written down before anything is run.
Autonomous Machine Learning Research
CHOUSORUS
Machines that do science.
Chousorus is building autonomous systems capable of designing, executing, evaluating and iterating machine-learning experiments.
the skilled creator
Concept
A machine should not merely answer.
It should investigate.
Chousorus is being developed as autonomous research infrastructure: a system that holds a question, decides what would answer it, and does the work. Not a tool that produces text about experiments, but an apparatus that moves through the experimental loop and keeps a legible record of every pass.
The loop below is the shape we are building toward. Each stage is a capability under active development, not a shipped feature.
- I
Observe
read the field
- II
Hypothesize
propose a cause
- III
Experiment
put it to the test
- IV
Evaluate
weigh the evidence
- V
Iterate
begin again, better
…and the fifth stage returns to the first.
Research, without the waiting.
Three movements
02
Run experiments
Composing the training and evaluation runs that would settle the claim, and carrying them out under conditions recorded well enough to repeat.
03
Learn from results
Reading outcomes against the original claim, discarding what failed, and letting the surviving evidence shape the next question.
NoteChousorus is currently in development. Nothing described here is offered as a finished product or a demonstrated result; these are the capabilities the system is being built to hold.
Tools wait for instructions.
Researchers pursue questions.
Most machine-learning work is not blocked by ideas. It is blocked by the interval between having one and finding out whether it holds — the setup, the sweep, the waiting, the reading of results, the small correction, and the waiting again.
Our vision is a system that owns that interval: one that keeps its own notebook, chooses its next move from evidence rather than instruction, and treats a negative result as information instead of failure. Autonomy here means responsibility for the whole loop — with every step left legible to the humans who set the question.
Chousorus
Named for an ancient Phoenician tradition of the skilled creator — the craftsman, inventor and maker.
Early access
Enter the workshop.
Chousorus is currently in development. Join the early-access list.