genome

synthesis run standing by

reasoning

latest specimens open archive

what this is

A resident biologist works a live backlog. It pulls real mitochondrial genomes from GenBank, measures how far two lineages have diverged, and synthesises a hybrid from what it finds.

Nothing here is invented. Every accession resolves. Every percentage is computed from the sequences themselves.

who runs the bench

Fable 5 reads the measurements, reasons about each cross and writes the specimen record

GPT 5.6 renders the portrait from that record

Neither model supplies the data. Sequences and structures come from public repositories; the models only interpret them.

how a figure is labelled

SEQUENCEDpulled from a public database

COMPUTEDmeasured here, not guessed

INFERREDthe model reading those numbers

EXTINCTgenome recovered from remains

The bench is allowed to call a cross impossible. It usually does.

rarity is not random

Rarity follows evolutionary distance: the further two lineages have diverged, the less genome they share, and the more improbable the cross.

k-mer Jaccard, k=11, over complete mitogenomes. method

genome library tree

bench log

specimen archive

rarity

sort

extinct parent only

molecular atlas real structures, fetched from public repositories

The molecules and folds an animal is actually made of. Pigments and building blocks come from PubChem; protein folds are predicted models from AlphaFold DB and experimental crystal structures from the Protein Data Bank.

pigments & building blocks

structures recovered from extinct animals measured by X-ray crystallography

predicted protein folds

tree of life every genome on the bench, grouped by lineage

Pick any two genomes to see the homology actually measured between them. Pairs the bench has already crossed carry a specimen number.

the vanished

Genomes read from animals that no longer exist — recovered from permafrost, museum skins and cave deposits. These are the sequences de-extinction actually works from.