Solving the structure took years. Downloading it takes minutes. Explaining it still takes a commission and three weeks.
A stained section on the stage, lit from below. The two colours this page is set in were found in a laboratory rather than chosen.
There is no shortage of ways to look at a molecule, and no shortage of people who can draw one beautifully.
Mol* and 3Dmol.js will render any published structure in a browser, built and maintained by EMBL-EBI and RCSB. The Protein Data Bank gives the coordinates away, and AlphaFold has added 214 million predicted structures beside them. Over 4 million researchers build their figures in BioRender. All of it is free to use. None of the work behind it was free to do, and every deposited structure is somebody's years.
The scarce thing sits further down. A mechanism-of-action animation is skilled work by people who trained for it, and it is priced accordingly: USD 10,000 to 15,000 for ninety seconds, USD 20,000 to 40,000 and upwards once there is more than one scene. It is worth what it costs. What it cannot be is repeated. It clears review at 21 days on average and returns ten times or more, so in practice the explanation gets settled in the brief, by whoever wrote it, before anyone has seen it move.
Condenser is the drafting instrument for that stage. You pick a real structure, write the beats, and read the whole explanation back as many times as it takes, while changing it is still cheap.
The science does not change between these four. Only the sentence does. That is the whole observation the tool is built on.
| Reader | What they are trying to do | What they have now |
|---|---|---|
| Medical affairs | Explain a mechanism to a clinician who will interrupt with their own question halfway through | A linear compendium deck that cannot follow the conversation anywhere |
| Clinicians | Get consent that is actually informed | Words. Adding a visual raises what a patient recalls by 7.8% to 29.6%, and most clinics have none |
| Researchers | Explain their own structure in a grant, a talk, a paper | Flat figure tools. Nothing that puts an argument on a three-dimensional structure |
| Teachers | Get a class from a flat drawing to a spatial idea | The stated barrier is skill, not access: the models exist and building one is a specialist job |
A tool that builds mechanism-of-action explanations should be made to submit its own.
Condenser's primitive is not an atom and not a bond. It is a beat: one sentence, plus the state of the structure while that sentence is true. What is marked, what has receded, where the camera sits.
Every beat carries two sentences and one science. One for a clinician, one for a patient. You write both once, and the reader chooses which one they are reading. Today those are two separate commissions.
An explainer is a sequence of beats, and the reader drives their own position through it by scrolling. Nothing plays. Nothing loops. The reader's scroll is the narrator's pace, which is the only reason motion is allowed anywhere near this.
Four beats across one structure. The sentences change, the marked region changes, and the science underneath is the same object throughout. The dashed line is the reader, and it is the only thing on the page that moves.
A condenser takes diffuse light and puts all of it on one point of the specimen. The tool is named after the job.
Structures come from the Protein Data Bank by their four-character ID, at the moment you ask for them. Four subjects ship authored, with four beats each in both registers: DNA at 1BNA, insulin at 4INS, haemoglobin at 2HHB and an antibody at 1IGT. Any other entry can be typed in, and then you write the beats yourself.
The visual grammar is a stain. Haematoxylin holds the structure, eosin marks whatever is being said right now, and everything unmarked recedes so there is one place for the eye to go. It is the reason this page is set in the same two colours.
Nothing is pre-rendered. The molecule you scroll past is being drawn in your browser from the deposited coordinates, which is why you can drag it. As the tool puts it: nothing here is a picture of a molecule, it is the molecule.
A rack of stained sections, catalogued and ready to be read. The specimen shelf in the tool does the same job with four subjects on it.
Studio is three columns: the specimen shelf, the live structure, and the beat editor. You pick a subject by its name in the world rather than its accession code, because the person explaining insulin to a patient does not think in accession codes.
For each beat you write a heading and the two sentences, choose how the structure is drawn, and choose what is marked in eosin. Beats duplicate, reorder and delete. The whole set copies out as JSON.
Read turns it into the artefact. The structure pins to one side, the beats scroll past on the other, and the structure changes state as each beat takes the screen. One toggle switches every sentence between the two registers without touching the science underneath.
One sample, run up a plate, resolving into bands you can read separately. Two registers out of one science is the same trick.
A commissioned animation is priced as a deliverable and paid for as a loop.
Studios are briefed after the creative is fixed, so medical, legal and regulatory review arrives at the end, when a change is structural. Reviewers describe pieces coming back ten times or more carrying thousands of accumulated comments, against an industry average approval cycle of 21 days.
Condenser does not remove that review, and nothing does. What it changes is what you are holding when you walk into it: a draft you have already rewritten ten times, rather than a first attempt that cost USD 40,000 to discover you wanted a different one.
The blue track and the day axis are drawn to the sourced 21-day average. The Condenser track is illustrative: no timing study has been run, and it is the shape of the comparison rather than a measurement. Every figure on this page is recorded in FACTS.md beside the build, sourced or illustrative, one line each.
It is a narrative builder, not a viewer and not an editor. Mol* is a better viewer and it always will be. Ketcher draws structures better than anything built here would. Condenser does the part those two leave out, and it should be judged only on that part.
The science in the four authored subjects has not been reviewed. It is written to the best of our knowledge from standard references, and that is not the same as verified. Every accession code and every claim needs checking by someone who knows the field before any of it is used in front of anyone.
It does not replace a medical illustrator. A studio piece will be better than anything this produces, and on a regulated submission it is what you should commission. Condenser is for the drafts written before that brief is, and for the many situations where no commission is coming at all.
Enzwa holds no life-sciences credential. This is not the studio's field, and that is rather the point of building it: the method was tested on a subject where nobody here could fall back on knowing the answer.
Four subjects authored, four beats each, both registers written. Or type any Protein Data Bank code into the box and write your own.
Best on a laptop or larger: the studio is three columns and the explainer is a split.
Open Condenser1BNA4INS2HHB1IGTYOURS