Genome Explorer

.genome is an open file format for genome data, designed for self-exploration. It uses the same underlying data as a VCF, but restructures it so tools like Codex and Claude code can process it reliably without relying on partial parsing or guesswork.

Under the hood it's not one flat file but a structured, queryable bundle (format version .genome/1.0). Your variants are stored in fast columnar tables alongside the annotations that give them meaning — trait associations, the supporting research behind them, gene-level context, polygenic scores, and pharmacogenomics — so a tool can answer questions against your whole genome directly.

Compared to raw VCF files, .genome reduces token usage by 3–10× while reducing factual errors by 10–20×.

Raw sequencing providers usually deliver technical files — FASTQ, BAM/CRAM, VCF, or a static report. Those files are valuable, but they're hard to inspect, hard to keep updated, and not designed for AI tools to reason over directly.

Most providers tell you what variants you have. Genome Computer tells you what you have and what it means: variants, annotations, trait evidence, pharmacogenomics, polygenic scores, provenance, and prompts in one portable format you can use with Codex, Claude Code, or any tool of your choosing.

For sequencing orders, we build from a gVCF instead of only a consumer-style VCF. A VCF mostly lists the places where you differ from the reference genome. A gVCF also preserves the confidently sequenced regions where no variant was found, which helps distinguish "no variant here" from "we do not know because this position was not confidently covered."

The difference is that you're not just receiving raw output from a sequencer. You're receiving a self-explorable genome you can keep, query, re-annotate, and explore as the science changes.

A .genome file is a portable container for your genetic data. The point is that it travels with you and works inside any tool that can read a file, so you're not locked into one platform's interface.

Concretely, you can:

• Drop it into an AI coding/agent tool — Codex, Claude Code, Cursor, or anything that can ingest a file — and treat your genome as queryable context. The file format (.genome/1.0) is structured so the model can parse variants, archetype data, and phenotype context directly.
• Ask open-ended questions — "What does my genotype suggest about caffeine metabolism?", "Which of my variants relate to sleep?", "How should I read my stress-recovery axis?" — and get answers grounded in your actual data rather than generic advice.
• Test hypotheses — pull in a paper or a new GWAS finding and ask the tool to check it against your specific genotypes, so you can see whether a result actually applies to you.
• Stay current — re-run analysis as new research lands (ClinVar/PharmGKB updates, new associations) without re-sequencing. The same file gets re-interpreted against newer knowledge.

All three are true whole-genome sequencing — they read your entire genome, not a fixed chip of ~650K sites like an array. The "x" is depth: how many times, on average, each position in your genome gets read. More depth = more confidence, especially on rare variants.

3x — reads your whole genome lightly. Confident on the common-variant layer that powers your identity, polygenic scores, ancestry, and trait insights — and already far beyond any array. It can see rarer variants but not yet call them with clinical confidence. A real entry into whole-genome.

30x — the clinical-grade standard. Enough depth to confidently call rare and pathogenic variants: carrier status, actionable ClinVar findings, full pharmacogenomics. This is "your real genome" with no imputation caveats — the sweet spot for almost everyone.

100x — research-grade, maximum confidence. Diminishing returns for everyday genomics, but it pulls ahead on the hardest cases: low-frequency and mosaic variants, structural variation, and difficult-to-read regions. The deepest, most future-proof read.

Whichever depth you choose, it converts into the same .genome bundle — same format, same tools. The difference is how much of your genome answers with confidence: 3x lights up your identity and trait layer; 30x adds the full clinical and pharmacogenomic layer; 100x maxes out certainty on the rarest signals.