For researchers
One experiment. One continuous record.
Research is usually spread across a chat window, a document, a paper notebook, a statistics package and a manuscript. The result survives; the reasoning behind it does not. Olto keeps design, protocol, execution, analysis and provenance connected, so the record of how you got there is a by-product of doing the work rather than a second job afterwards.
No signup required. The demo runs the same generation path the product does.
The workflow
From question to record
Seven stages, each one carrying its context forward. The Passport at the end is assembled from what the earlier stages already recorded, not written separately.
- 01QuestionState what you are testing and why, so the reasoning is part of the record rather than something reconstructed later.
- 02DesignControls, replicates, sample size and the variables you are holding. The gaps are named before the bench, not after.
- 03ProtocolA 12-section structured method, versioned, with a fingerprint that changes when the method does.
- 04ExecutionA guided run: steps, timings, deviations and observations, captured live or entered afterwards and labelled as which.
- 05DataDatasets attached to the run that produced them, with the instrument and lot that were actually used.
- 06AnalysisStatistics computed in your browser, on your data, with the test named and its assumptions stated.
- 07RecordA Reproducibility Passport: the method, the execution, the provenance and a fingerprint anyone can recompute.
What computes what
Where the model is, and where it is not
A model drafts and suggests. It does not compute the fingerprint, the rigor score, or your statistics. Knowing which is which is the difference between a number you can cite and one you have to check.
| Capability | Kind | What it does |
|---|---|---|
| Protocol drafting | AI-assisted | Drafts a structured first pass from a research goal, for you to correct. It is a starting point, not a method. |
| Missing-control detection | AI-assisted | Reads a design and names controls it cannot find. It suggests; it does not decide. |
| Rigor Score | Deterministic | A text-pattern instrument over the written method. Every signal is published and computed the same way every time. |
| Protocol fingerprint | Deterministic | A content address over the canonical form of the method. Recomputable by anyone, from a published specification. |
| Statistics | Deterministic | Runs in your browser. Raw data does not need to leave your machine to be analysed. |
| Version lineage | Deterministic | Every revision of a method, and which version a given run actually executed. |
31 of 454 server routes call a model. That ratio is re-measured by a test rather than recorded by hand, because the previous figure went stale on the homepage. It counts routes that call a model; it is not a guarantee that no displayed number is model-derived, and the protocol risk score is one that is.