Loading
Light, color and a convincing surface can all mislead. Change the display while holding the numbers still.
Explore the recorded preview
0909 / 10A field becomes easier to understand when light reveals its gradients. It also becomes easier to misunderstand when that light makes the field look like an object it never simulated.
FLOW's apparent depth encodes dye concentration. It is not the height of a free liquid surface. That sentence is a small part of the interface, but it prevents a large substitution in the reader's mental model.
The solver produces fields. The renderer reconstructs and presents them. A useful boundary between those responsibilities is simple to state: changing a display setting must not change the numerical state.
The explorer below makes the boundary visible with an analytic field. It contains 48 fixed Gaussian concentration samples. The lighting control changes the appearance of the marks, while their values and sum remain unchanged. It is a teaching figure, not a frozen output from the fluid solver.
Frozen analytic Gaussian samples, not a solver recording. Lighting changes the appearance of each mark. The 48 concentrations remain identical; the apparent depth is not a physical free surface.
The readout is intentionally unexciting. The same sum remains there after the picture changes. Its job is to keep the visual effect from smuggling in a claim that the underlying material changed.
The fluid display derives lighting normals from the reconstructed concentration field. Normals determine how a light interacts with a displayed surface. They do not add momentum, alter viscosity or create a physical free surface in the solver.
That separation needs testing. A renderer can accidentally mutate a texture, reuse a working buffer or trigger an extra numerical step while responding to a view change. FLOW's presentation checks compare raw fields around camera, theme, view and quality changes. The intended result is numerical identity alongside a visibly different presentation.
The lesson is not to make everything visually flat. It is to make the visual transformation accountable to the quantity it represents. A surface can help a reader see concentration gradients as long as its caption does not call the relief a new physical measurement.
Reduced motion exposed a different kind of display problem. A gesture could leave a paused field with too little visible progress and no clear next action. Respecting a motion preference is necessary, but a blank-looking pause is poor instruction.
The current internal Make a Flow action applies a bounded twelve-tick preview, corresponding to 0.1 model-time units at the declared tick size, then stays paused. A visible Step control advances one tick; Play permits continuous motion. The guidance describes the actual paused state without claiming that every later pause happened after exactly one step.
This is a different design decision from starting an animation silently. A finite progression gives the reader something to inspect and a clear way to continue. It also makes the interaction testable: count committed ticks, verify pause, wait, and confirm that no autonomous ticks appear.
These internal controls are not the public recording controls. The public preview offers recording playback; the article should not tell a visitor to press a live solver button that is unavailable there.
Two panels need direct A/B labels and the relevant settings, not a sentence several screens away. A matched-time caption makes the comparison legible. A still frame's time must not be presented as though it labels every video frame.
The series itself keeps the same design as Particle Life and Mind Map: the same article frame, cards and navigation. Creativity belongs in the explanations and explorations, not in making readers relearn the site for each project.
More scalar passes, higher resolution and more elaborate presentation can cost frame time. A clean desktop screenshot does not qualify sustained phone performance. Thermal behavior, input latency and frame intervals require actual runs on the relevant devices.
A beautiful image is therefore one observation among several. Keep the legend, the pause state, the numerical identity and the accessibility path in the same design conversation.
Make the change honest: a different light should reveal another reading of the same data, not invent another physical experiment.
Next: a first visit to FLOW, and a checklist for what to believe afterward.
More to read
A world that wraps is not a wind tunnel. Following a parcel through periodic, closed and channel domains.
A useful first visit, a reproducible reading and a checklist for deciding what the evidence actually supports.
A ten-unit run lost 5.115% of its dye. Keep the failed observation, change the transport method, measure again.
