The GenomeComputer Company
Back to tutorials

Tutorial

Open your .genome in Codex

Set up a clean Codex workspace and prevent the model from guessing what is inside your genome files.

5 minCodex

What you will learn:

You will have a reliable Codex workflow for file-based genome exploration.

Codex is the right place to work with a .genome bundle because it can operate inside a project workspace, inspect files, search annotations, summarize structure, compare evidence, and help you write better follow-up questions. The key is to treat the bundle as a workspace, not as a magic answer button.

Before you ask about a trait, gene, or paper, make the model prove that it understands the files. This avoids one of the most common failure modes in AI genome exploration: the model answers from general genetics knowledge while acting as if it inspected your actual data.

Codex workspace workflow

  1. Start a new Codex workspace dedicated to your .genome.
  2. Add the .genome bundle or the specific files you want to inspect.
  3. Tell Codex that this is an AI-ready genome bundle, not a generic document dump.
  4. Ask it to list the files it can actually see.
  5. Ask it to identify which files contain annotations, sources, prompts, and technical data.
  6. Only then ask your first biological or research question.
  7. Keep the same session for related follow-ups so the file map remains in context.

Workspace orientation prompt

I am opening a .genome bundle in Codex. It is an AI-ready package of my genome data, annotations, sources, and prompts. First inspect the files in this workspace and tell me exactly what you can access. Do not answer any biology questions yet. Create a short file map, then ask me what topic I want to explore first.

Ground the session

For the rest of this conversation, separate every answer into: evidence found in my .genome, annotation or source context from the bundle, and broader research context. If you cannot connect a claim to one of those, label it as uncertain.

Useful session rules

  • Ask the model to name files when it uses them.
  • Ask it to quote or summarize source fields when available.
  • Ask it to say when a file is too large, missing, truncated, or unreadable.
  • Ask it to keep a running list of follow-up questions.
  • Ask it to rewrite broad answers into file-grounded summaries.

If the model gives a generic answer before inspecting the files, stop and reset the session. A good .genome workflow starts with file awareness. The AI should know what it has before it tries to explain what it means.

Recover from a generic answer

Pause. Your answer was too general. Go back to the uploaded .genome bundle, identify the relevant files, and rebuild the answer using file-grounded evidence first. Then add research context separately.