Your first walkthrough · Lesson 2 of 3

Ask your first genome question

Follow a synthetic example from one question to a saved, checkable answer.

6 min read · allow extra time to follow alongCodex, Claude Code

Before you start

An AI conversation. The synthetic exercise below requires no personal genome.
For your own data, complete “Open your .genome in Codex” first.

Start with one record and one question. In this exercise you will inspect a synthetic annotation, compare the answer with the input, and save a result you can check. No personal genome is needed.

The example below is a teaching excerpt, not a complete .genome bundle. DEMO1, demo-001, and DEMO-STUDY-01 are invented identifiers. The association is made up to demonstrate how an annotation can be mistaken for a conclusion.

1. Run this synthetic exercise

SYNTHETIC TEACHING EXAMPLE — not personal data or a complete .genome bundle. All identifiers and the association below are invented.

File: example-annotations.json
{
  "record_id": "demo-001",
  "gene": "DEMO1",
  "observed_genotype": "A/G",
  "annotation": "Example research association with caffeine metabolism",
  "source_id": "DEMO-STUDY-01",
  "source_status": "Illustrative placeholder; no research paper supplied"
}

Question: What does record demo-001 contain, and does it tell us how much coffee a person should drink?

Use only this example. Separate the recorded data, the annotation, and what cannot be concluded. Cite the filename and field names for each claim. Do not invent a paper or treat the synthetic record as personal guidance.

2. Compare your answer with this example

Recorded data
example-annotations.json, record_id demo-001, records observed_genotype as A/G for the fictional gene DEMO1.

Annotation
The annotation field contains an example association with caffeine metabolism. The source_id is DEMO-STUDY-01, but source_status says that no paper is supplied. This is not verified research.

What we cannot conclude
This invented record does not establish a person’s caffeine response or an amount of coffee to drink. A label in a file is not evidence for that recommendation.

Illustrative answer. Wording may differ; the file references and limits should agree with the supplied record.

Check that it worked

The answer points to example-annotations.json and the fields it used.
It treats A/G as a value in an invented record, not a finding about you.
It identifies the source as a placeholder and does not fabricate a citation.
It does not turn the annotation into a personal recommendation.

3. Ask a useful follow-up

Audit your previous answer against the supplied example. For each factual claim, show the field that supports it. Identify missing evidence and remove any claim that the example does not support. What would we need to inspect before treating the annotation as research evidence?

Once you can trace the synthetic answer back to its input, repeat the workflow with your own bundle. Open it using the previous lesson, then replace the bracketed text below with one gene or variant you want to inspect. Your bundle may have no matching record; that is a useful result when reported accurately.

4. Try one question with your own bundle

Inspect my .genome bundle for [one gene or variant]. First identify the relevant files and how you will read them. Then report exact matches with file paths and record identifiers. Separate recorded data, annotation context, checked research sources, and uncertainty. Say which files you did not inspect. Do not infer a negative result from a missing record or provide a personal recommendation from an association alone.

5. Save the question and answer

Create a new Markdown file in reports/ for this exploration without overwriting an existing report. Include my original question, the prompt, the answer, the exact files and records used, whether the input was synthetic, unresolved questions, and one useful next step. Do not modify the source data.

For a chat tool without file access, copy the question and answer into a document yourself. Keep the synthetic label if you used the teaching example. Next, use the evidence checklist to review the saved answer.

Before you continue

Check each item when you can do it.

Match the answer to a specific record in the example or your bundle.
Separate the record, annotation, and uncertainty.
Save your question and answer for the evidence check.

Next tutorial

Check the evidence in an AI answer

Trace each claim to a file or source, and spot unsupported conclusions.