New in 0.5.0 and 0.5.1 — the main thing
The body as fifteen systems, not as a stack of test results
The thyroid used to appear on the page three times — among the deviations, in the line about a drug, and in the list of what to test — and nowhere was it said that these are one subject. A body system is now the unit things assemble around. The tab is called Radar: the figure, two rings per system, one block each, and the full card one click away. Since 0.5.1 every segment also carries a panel of positions somebody wrote, standing in front of the base.
scholion system lipids · demo profile
System: Lipids · patient's register
this build cannot answer for this list — the genome is not
readable on this profile — see the genome status
1. Laboratory now 100/100 — markers measured: 3 of 4
never taken: Cholesterol, total
2. Movement 100 → 100 (0) against 2025-11
3. Genetics base GenCC 2026-09-06 — genes: 38;
positions of the curated panel: 7; read: 0, unread: 0
positions whose phrase is signed by the panel's author,
and by no clinician: 7 (signed 2026-09-13)
4. Prescriptions · 5. Clinician's target · 6. What to test · 7. Questions
Next step, in three baskets
test · read in the genome · ask the clinician
no genome is attached — a full genome would close for this
system: 40 genes of the list not read here
The synthetic demo profile of a fictional person, shipped with the package: scholion system lipids. That profile has no genome, so nothing of the genetic half is read — and the card says so rather than reporting silence as health. The same card comes back from a click on a radar segment, from an organ on the figure, from the local API and from the assistant's tool.
◎
Two rings, never merged
How much of the laboratory panel has been measured, and how much of the genetic half has been read. They are different quantities: somebody who has taken the whole panel and had no gene read should not see what somebody in the opposite position sees. A system whose genetic list is not composed shows that ring dotted — an absence, not a zero.
🧬
The base of the genetic half has a version
1469 genes across the fifteen systems from the Gene Curation Coalition export: who asserted the gene–disease link, how strongly, under which mode of inheritance, and on what date. Two groups disagreeing about one gene are shown disagreeing rather than averaged. An assertion classified Limited, Disputed or Refuted is never presented as a finding, and one copy of an allele in a recessive gene is carriership and a question — never a risk line.
✍
In front of the base, a panel somebody wrote
214 positions across fourteen of the systems, authored rather than generated. The unit of a row is a position, not a gene — an rsID with its HGVS, the allele the author named, the phrase for one copy and for two. A row whose phrase for the state actually found is missing is kept and printed as pending: somebody put the position here, and what follows from that genotype is still to be written.
- A system is never called clear where it was not read. «Nothing found» over a gene nobody read is a statement about the file, not about the person, so the count of unread genes travels inside the verdict rather than in a footnote under it. A gene short reads cannot read at all — a gene beside its pseudogene, a triplet repeat — is named as needing a separate method, and counts as neither read nor clear.
- A position the genome file does not list is the reference only when the alignment says so. Until the catalogue positions have been genotyped from the aligned reads, such a position prints as not read — and the panel where the gap shows names the step that closes it.
scholion recompute finds that step, says what it still needs, and runs it with the progress visible.
- Genetics does not enter the 0–100 score. A genotype cannot be refuted by the next blood draw; a score of laboratory deviation can. Pulling one into the other would produce a number that quietly drags a system down for years and that no measurement can disprove.
- Three baskets, because they are three different jobs. Test — the laboratory. Read in the genome — something you run yourself. Ask the clinician — something you take to an appointment. Tipped into one list of «recommendations» they read as a prescription. An empty basket says why it is empty: «nothing» without a reason reads as «all is well».
- Two densities, one verdict. The card prints in a patient's register and a clinician's; the second adds rsIDs and genotypes, coverage as a figure, the class of each link, the source and its tier, the date and provenance of every point. They differ in detail and never in the verdict — there is no reassuring version for the patient.
- The product still does not write the curated sentence itself. «What this genotype means for this drug» is a medical statement: it is printed only with a named source, and entries dropped for lacking one are counted as a figure. Every row that ships is signed — by the author of the panel, against the source it names, and by no clinician — and the card says exactly that, with the count and the date. A signature by the author of a panel is not a clinical endorsement, and nothing here pretends otherwise: counter-signing a row, correcting it, or striking it out is what is being asked of a clinician.
- One design across the application. Every page is built the same way: a title, the question it answers, the few numbers worth seeing first, the sections, the details folded and the sources in one line at the bottom. Six tabs — Overview, Radar, Labs, Genome, Medicines, Lifestyle — with the Guide and the Assistant in a menu; on a phone the header takes one line and no page scrolls sideways.
