Most projects start with "we need a website." This one started with a teaching problem. Clinicians and students working through a rehabilitation plan need to answer a specific question quickly and correctly: which movements load this muscle, how is each one performed, and how much of it is actually required. The answer existed, scattered across textbooks, anatomical references and exercise libraries that do not agree with each other.
So we built the answer as an instrument rather than a brochure. It is an anatomical atlas you drive by clicking a muscle on a three-dimensional figure; the muscle tells you what it does, which exercises train it, and what volume it needs.
Why a tool and not a site
The school was explicit that nobody would be visiting a front page. The brief was a thing a clinician opens mid-session, uses, and closes — so there is no marketing copy on it, no navigation to learn, and the landing state is the figure itself. It runs on a Queantic subdomain because a domain of its own would have added a cost and a renewal to manage in exchange for nothing the users needed. That is the right answer surprisingly often, and we would rather say so than sell a domain.
The work nobody sees
The hard part of a reference tool is that the sources disagree, and a clinical audience is exactly the wrong audience to hand an unreconciled dataset.
- The anatomy is a Z-Anatomy derived model — 92 muscles, 25 groups and 221 named structures, used under CC BY-SA 4.0 with the licence and attribution stated on the tool itself.
- 268 exercises came from the open Everkinetic dataset with their own vector illustrations.
- 44 exercises were written from scratch, to cover muscle groups the dataset simply had nothing for: the rotator cuff, the adductors, the neck, the shin and ankle, the hip flexors and the serratus anterior. Each is marked as added wherever it appears.
- 62 two-frame illustrations were drawn for this atlas — the 44 new exercises plus 18 of the dataset's own that had no drawing — posed and drawn in the same line style as the rest so a user cannot tell which is which by looking.
- 43 exercises had their muscle assignments corrected. The adductor machine was filed under quadriceps, an inverted row under chest, a reverse calf raise under the calves rather than the shin. Corrected entries are labelled as reviewed, because a clinician is entitled to know which claims we changed.
- Muscle emphasis is our mapping. The source data only ever names a whole muscle — "pectoralis major". Which head a movement emphasises, upper chest on an incline press and lower on a decline, is a judgement the source does not carry, so the tool states that it is ours.
Every one of those is a provenance label on the screen, not a footnote. A teaching tool that cannot tell a student which of its claims came from a licensed source and which came from the people who built it is not fit for the room it is used in.
Open, and still anonymous
This is the unusual part of the engagement. The school was happy for the tool to be publicly reachable — it is genuinely useful and there was no reason to put a login in front of it — but did not want to be named as the client. Those two wishes are easy to state and awkward to honour, because a public site normally carries its owner's identity in every corner of it.
So it does not. The tool carries its licences and its sources, which it owes the people who made them, and carries no institutional branding at all. Its content assets are served through a same-site gate rather than as plain static files, which does not stop anyone who legitimately loads the page — nothing can, for content rendered in a browser — but does stop a bulk mirror of a few hundred commissioned illustrations. That is the honest description of what that control achieves, and it is the description the client was given.
Why a public agency should care about this one
Because it is the shape of problem agencies actually have. A tourism board with a reconciled listings database, an economic development office with a properties index whose fields a site selector trusts, a district with an ordinance archive — these are all the same job: take sources that disagree, reconcile them, label the provenance, and put them behind an interface somebody can use under time pressure. It is also a reference from a public institution, which is the thing most solicitations score and the thing a new vendor usually cannot offer.
We will put you in touch with them to take a reference, with their permission asked first. What Queantic does for public agencies is set out on Queantic for public agencies.