Seismo Pulse

Seismo Pulse gathers a crisis in progress into one live picture, for critical-infrastructure operators where an outage is not an option. The same picture has to be useful to the person managing the event and to the one checking in on it.

The challenge was never adding more information. It was deciding what each of them needs to see, and when.

Seismo Pulse is aimed at organisations that have to keep something running: hospitals, airlines, cantons, federal offices, the postal service. That list is not closed. They all have crises and none of them means the same thing by one. An incident at an airline can be a crash. At a grid operator it can be a defect on a pylon that someone has to note before the end of the shift.

So it could not be one screen with a fixed set of panels. Every capability is a module, and a module has to pass two tests: it can be switched off for a customer without breaking what is left, and it owns its own kind of thing, whether that is incidents, tasks, journal entries or contacts. Modularity is a word every product deck uses. Here it was the first constraint rather than a feature.

The situation picture is the module they all share and the one every customer starts from. What differs is what stands around it, and what an entry in it turns out to mean.

The module set is designed to keep growing. Every customer conversation either adds one or sharpens one, and a module nobody has seen yet cannot be specified. So it gets built as a prototype first and written down after. That makes prototyping the method here rather than a phase: it is how each next module gets decided, one at a time.

Nobody in this domain buys a concept off a slide. So instead of presenting pictures of an idea, we demoed something running.

We prototyped heavily, AI-assisted on MUI, and took working prototypes into the demos. Clicking through the thing rather than describing it is what won stakeholders over. Doing that at volume forced the next question: how to keep producing them without quality dropping on the way to code.

The obvious move was to point Figma’s MCP server at the design file. It half worked, which is worse than failing. A Figma file is organised for the person who designs it, not the way markup is, so the model pays for how a design is assembled long before it reaches how to build it.

What replaced it is duller and works. Tokens leave Figma as a raw export, scripts rewrite them into the structure Cando’s developers already use, and the bridge is kept in sync, which is what makes it infrastructure rather than a one-off conversion. It has a price, and design pays it: one hand-picked colour and the thing the bridge exists to guarantee is gone.

A feature now starts from the ticket, a screenshot of the Figma frame and the design system as markdown. The developers’ own estimate of the difference is 3×, and the practice left the design team entirely: sales and product owners prompt on top of the real codebase, so a pitch stands on where the product actually is rather than on a disconnected artefact.

• –Led the concept and UX: how incidents, threats and assets resolve into one picture that still reads at every level.
• –Designed the modular structure: what counts as a module, what a customer can switch off, and how the same module works organisation-wide and inside one situation.
• –Drove the prototyping practice, AI-assisted on MUI, that put working software in front of stakeholders instead of slides.
• –Built the design-to-code workflow behind it, and carried the practice out to sales and product ownership.

Seismo Pulse is in active development at ecmt in Winterthur. The module structure is there so a new customer is a configuration rather than a rebuild. And the prototyping practice did more than win the demos it was built for: it left the design team, and the workflow behind it is now how features start and how the product gets pitched. None of which was the plan.