Begin with a scientific question.

You bring the research question. An AI collaborator prepares the Sounio program, checks it and explains the evidence. You do not need to write code to prepare this request.

This page prepares instructions, not numerical results. Execution needs an authorized agent connected to a compatible Sounio environment; a chat-only model cannot verify a run.

Describe your study

Choose a research workflow

Your scientific question

What do you want to understand, estimate or compare?

Available data and where the authorized agent can read it

Describe the file, columns and provenance. Use synthetic examples here, not patient records.

Variables, units and measurement uncertainty

State units, time basis, missing values and known correlations. Unknown is an acceptable answer.

Assumptions to preserve or discuss

Explain the scientific model or say that it must be proposed for your approval.

What you want to receive

For example: a table, plot, uncertainty explanation and reproducible methods note.

Reference method and acceptance criteria

Describe a trusted reference, controls or a tolerance you have justified. Do not invent a target.

The builder assembles text locally and does not submit these fields. Do not paste real patient records or confidential data into public or unapproved model services. Use an institution-approved environment for authorized study data.

Paste this first-stage request into your AI collaborator. Paste its exact reply below, validate it locally, then review and confirm your brief before preparing implementation instructions.

Inspect first-stage clarification request

2. Review clarification and approve implementation

Paste the AI clarification reply

Paste the model's complete receipt unchanged. You do not need to write JSON or code. Use the scientific field hints and sidebar to refine your brief; after any edit, request and validate a new reply.

Implementation is locked. Validate the reply for the current brief, then confirm it.

I have reviewed my current scientific brief and the matching reply, and approve preparing implementation instructions.

This local gate checks receipt syntax and correspondence only. It does not establish scientific truth, model authorship, source access or tool use. Confirmation does not execute anything or approve a model automatically.

Inspect approved implementation prompt and source-bound context

Prompt template · not an executed study

What the AI must establish

Scientific details to clarify

Which mixture, pressure, temperature and initial radical seed define the experiment, and in which units?
Is the protocol isothermal or constant-volume adiabatic, and how is ignition defined?
Which uncertainties are measured, which are assumed, and which covariance information is available or missing?
Which species, time window and artifacts should be reported, and how should missing inputs be marked?

Required evidence

Retain the input protocol, units, data provenance and missing-data decisions separately from model assumptions.
Request species concentrations, ignition criterion and uncertainty-source attribution with their declared interpretation.
Require actual commands, compiler identity, source revision, exit status and raw output for every authorized execution; otherwise report not executed.
Separate a reproduced historical checkpoint from an adapted protocol and from independent scientific validation.

Scope of this recipe

This adaptation is unexecuted; the historical golden uses lean_single and does not establish current Madaros behavior.
The implemented first-order uncertainty attribution assumes independent sources; do not label its bands as quantiles or silently discard covariance.
Keep the mechanism and ignition-front sensitivity explicit; do not extrapolate to reactor or flame safety.