Starling Lab

An organisation whose entire proposition is the authenticity of digital records can’t afford a website that feels approximate. Every detail is read, fairly or not, as evidence of whether the Lab knows what it’s doing – and the Lab is read by the ICC, the UN Human Rights Council, Reuters, CNN and Stanford.

That set the bar, and it ruled out the obvious shortcut. The client required Uncode. On a project like this it’s tempting to configure the theme and dress it. The pieces that carry the argument – the hero, the framework panels, the data integrity sequence, the editorial components – are custom modules instead, designed and coded on Uncode and pushed to the theme’s absolute limits, with everything else deferring to them.

Starling is not one thing. It runs multiple labs, each with its own research programme – Journalism Fellowships, the Archive Accelerator, Project Dokaz – alongside a case study library, dispatches, courses, open-source prototypes, press highlights, events and recruitment. All of it had to feel like one organisation without flattening into one shape.

We worked the system out with the Lab rather than in isolation, dissecting each piece and assembling it as the build evolved, so decisions were made by the people who understood what the content had to do. It’s slower than working to a finished spec. It’s also how the pages that weren’t in the concept designs end up looking like they were designed on purpose.

The proof came later, when the Lab added a new practice area. The Spatial Lab arrived with its own programme, its own research questions and its own case studies, and it dropped into the existing system – templates, navigation, archive filters and all – without a redesign. A design system is only worth the name when it survives something it wasn’t drawn for.

The Lab’s central argument is a comparison: 149 years for photography to produce fifteen billion images, eighteen months for generative AI to do the same. Everything the organisation does follows from it.

Working from supplied artwork, we made it perform – adding the animation and film, and building it as a pinned scroll sequence where the figures resolve out of blur as the comparison assembles, so the reader feels the gap between the two numbers rather than reading it.

Then rebuilding its behaviour for mobile. A pinned scroll narrative designed at desktop width doesn’t survive a 375px screen by scaling down; the pacing, the crops and the reveal order all have to be reconsidered. Most of the effort in that component went into the version most people will actually see.

The site’s job was to become the centre of the Lab’s publishing – one place where research, submissions, fellowships and collaborations sit together and read as a body of work.

The Lab asked for four ways of surfacing case studies on a single page: featured work by practice area, an Editor’s Pick with room for a proper standfirst, the most recent studies, and a filterable archive across every area with a live count. Four curation mechanisms competing for one page is where a design usually turns into a list.

Making them coexist was the design problem: giving each a different weight, rhythm and density so a reader moves between them without noticing the machinery. Underneath, the content model separates the types properly – case studies, dispatches, news and events, prototypes, courses, people – and relates them, so a case study surfaces alongside the lab that produced it and the framework it applies. Editors file a piece of work once. The site does the cross-referencing.

WordPress — and Uncode — was the Lab’s own call. As an academic organisation it wanted a platform its people already knew and could still hire for in five years, with no dependency on a specialist stack or a single agency. Given staff turnover, grant cycles and an editorial operation run by people whose expertise is cryptography and human rights law, it’s a defensible decision.

Our job was to make sure it cost them nothing. We build across Next.js, Sanity, Webflow and WordPress, and the platform is a project decision rather than a house preference – which means it can be the client’s decision without the work suffering for it.

Page composition is component-based and deliberately constrained. Administrators assemble pages from blocks that can’t be arranged into something off-brand. It’s a trade – less freedom in exchange for less drift – and for an organisation whose credibility is the asset, it’s the right side of the trade.