How it works

The platform asks for four things, and none of them require you to have written any code. A host organism, chosen from the 65 the catalog covers. A heat target, placed anywhere on a ladder that runs from 100 SHU to 16,000,000 SHU. A burn profile that says how the heat should arrive and how long it should linger. And a tissue, because a target that makes sense in fruit rarely makes sense in a leaf.

Those four selections live in a short form on the project screen. Submit them and the platform does something useful before any computation starts: it shows you the nearest entries from the catalog, so you can see at a glance whether your target sits inside well-mapped territory or out at a sparse edge. SS-0421 and its neighbours are not decoration. They are the evidence that tells you how much to trust what comes next.

Step 1

Create project, choose host and tissue

Step 2

Set heat target and burn profile

Step 3

Run flux analysis — CPU, returns in-page

Step 4

Scan enzyme variants — GPU, queued

Step 5

Review shortlist, pick candidates

Step 6

Assemble construct, validate, export

Elapsed

An afternoon, most of it reading

Notice how much of that is reading rather than waiting. The flux step runs on CPU and comes back while you are still looking at the map, and it answers the question that matters earliest: can this host fund this target at all? A ceiling below your number is not a failure of the software. It is a season of greenhouse space you did not spend.

Only the variant scan leaves the browser for a GPU worker, and it queues rather than blocking you. Start it, go read the catalog neighbours, come back to a scored shortlist. Nothing about the session demands that you sit and watch a progress bar.

A finished design is reviewable by someone who was not in the room. The project record carries the model and its assumptions, the scored variants with the reasoning behind the shortlist, the edit plan with its off-target evidence, the assembled construct, and a dated manifest. Collaborators see what was decided and on what basis, which is the difference between a result and a claim.

Cost follows the same logic as the screens. Flux modelling, codon work, primer design, annotation, assembly and export are CPU engines and run free, however many times you revise. GPU engines draw on prepaid credits — $25 buys 250 credits, one credit covers roughly one Standard job, with Performance and Max classes available when a structure earns the extra compute. Most first projects never leave the free tier at all.