One framework to rule them all: how I confirmed I’m a sucker for efficiency
SAPO
2016
In 2016, SAPO, one of Portugal's best-known digital brands, had more than 30 media websites, each built in its own way, on its own technology stack. Every stack needed its own maintenance, and every site was a separate job. When I was approached to design a new news website, SAPO 24, I challenged the engineering team to approach it differently: build it as the first site on one shared front end that every other media site could move into. Together with the engineers, and later more designers, we moved all of them by 2019, and the framework went from saving effort to bringing in new business. This story is about scale and efficiency, and about the part I found hardest back then: bringing people along.

SAPO 24 was going to be a general news site: politics, current affairs, the daily news. It would compete with the big news brands, with its own editorial team. It launched in September 2016, a partnership between SAPO and MadreMedia, which ran the editorial side.
Behind it, the company had a bigger problem, one that not many people were talking about. We had more than 30 media websites. They used different technologies, were built in different ways, and ran on different back offices. And people were leaving. Co-founder Celso Martinho had left, and many others followed. Most sites no longer had enough people to look after them.
Designing one more site the old way would have meant one more thing nobody could maintain. So the question became: what if SAPO 24 was the first site built on something every other site could use?

SAPO 24, the first of many websites in the framework
I'd done this before, on phones
It wasn't a new idea for me. Around 2008, I joined SAPO's mobile team, a multidisciplinary team, working on native apps for iPhone and then Android. Around 2010, my focus moved to the mobile web. Back then, if you wanted a site to work on phones, you built a separate one on its own address: m.sapo.pt and the like. We were a small team and couldn't keep a mobile version of every SAPO site going by hand, so we built one simplified framework for all of them.
It worked, but we didn't leave room for creativity. Almost every mobile site looked the same: change the colours, and that was it.
So this time, the shared part had to be strong, and every site still had to look like itself.
The advent of Atomic Design
I wasn't starting from a blank page in my head either. Front-end frameworks like Bootstrap, which Twitter released in 2011, had shown that a website could be put together from ready-made, reusable pieces.
Then, in June 2013, Brad Frost published a blog post called Atomic Design. It described interfaces as small building blocks, atoms, combined into bigger ones, all the way up to templates and pages, and it presented the whole thing as a method for creating design systems.
That post heavily inspired how I thought about this framework: what a design system is, how small pieces add up to templates, and how to make something like that scale. We added one more layer on top: templates that each site could extend, and pieces of structure it could override where it needed to be different.
Still, I didn't call what we built a design system. The idea was around, but design systems were only becoming a field of their own in 2016.
What every media site shares, and where they differ
Media websites share a lot. They all have home pages, landing pages, ads, articles, lists of articles and categories. Building all of that again for every site made no sense.
Each site also had needs of its own. The sports site, mostly about football, had live matches, pages for teams, lists of games and rankings. The entertainment site had every movie in every cinema, so it needed templates for movies, cinemas and trailers, and special sections for the Oscars.

These were exceptions, and they could live in the same system. What mattered was keeping everything truly shared in one place, so an improvement made for one site became an improvement for all.
SAPO already had a base design system called Ink, built by other people at the company: a responsive grid, common interface elements and interactive components. Content lived in WordPress, chosen together with the editorial teams, who were comfortable with it, and with our commitment to keep adding to it.
WordPress was only where editors wrote, though. What readers saw was a front end built from the ground up, made for media: shared templates, and overrides to those templates and to the CSS for each site, only where it needed to be different. The sports site needed many of its own. A site made only of articles needed none.
The base was plain on purpose. If it had a strong look, every site would look like SAPO 24. If it had none, it would be useless. The overrides were where each site became itself.
Also, I designed directly in HTML and CSS. By then, Photoshop already felt like a prison. It had never been built for this kind of work: it couldn't carry a change across pages, and it didn't work the way we already worked in code, in components. In HTML and CSS, change the typography once and it changes on every site. Leaving Photoshop behind was an important step in my journey.

Miranda, a beauty website
Co-op mode
Nelson Duarte was the engineer who stood out, the same Nelson I later built Pandemia Clara with. There were more engineers on the team, and as the work grew, more designers joined.
That's how I tend to work. I have an idea, and I try to sell it to the people around me that I trust. Then, to bring it to life, I need their motivation as much as my own.
It's always been like that. As a kid, I never enjoyed playing video games against my friends. I liked playing with them, against the computer or someone else. Music is the same: I could never start a band on my own.

With great power comes great responsibility
Centralising everything in one place had a price. If we changed something in the core, it changed on every website. If we created a bug in the core, we created it on every website too. The same system that let us improve 30 sites at once could also break 30 sites at once.
When it worked, it was a big win. Embedding an Instagram post in an article, for example, sounds small. On 30 separate sites it would have been 30 separate jobs. We built it once, and every site had it. It was also an invisible win. Nobody measures the work they didn't have to do. People only notice it when you remind them what it would have taken: fixing the same bug, or making the same improvement, on 30 websites, one by one.
But we made a lot of mistakes. Things that seem basic today, like code reviews, pull requests that go through a pipeline, or QA engineers testing before a release, weren't in place. Part of the reason was that this was media, not banking. We weren't handling people's money, so a mistake was never going to be a tragedy, and we never put real checks and balances in place.
Those were the times I wished the team was bigger. On the design side, I'd have liked someone to exchange ideas with. And the same was true in engineering.
Good for efficiency, great for new business
It took three years. By 2019 every media website had moved, and new sites were built on the same structure from day one. I stayed on it, maintaining and growing it, until 2021.
The last big piece of that move was SAPO's own homepage, sapo.pt, in 2019, one of the most visited websites in Portugal.

SAPO's homepage
Maintaining a media framework is never quiet, because editorial sites keep having side projects. Portugal has presidential, legislative and local elections, and each one needed its own pages. The commercial team would sell an advertising format that didn't exist yet, and we had to support it. A new partnership needed a new site.
That last one turned into the biggest change. Because we could build quickly, our partnership managers could go to other editorial teams with an offer that was ready to go. Partners shaped the look and feel of their site with us, and got a ready-made front end, a content management system and an advertising platform already in place, with the ad revenue shared, most of the time 50-50. The framework wasn't only a way to save effort. It started bringing in new business. Polígrafo and MAGG were two of those partners. MAGG was a site made only of articles: it needed no overrides, and its redesign took about 3 to 4 weeks.

Around 2019, a team of 4 could do what a team of 10+ did before. That year, the system supported 300M+ pageviews and 4.5M+ visitors, and it had made sites possible for 10+ partners.
The human part is always the hardest
Building it was the easier half. Getting everyone on board was harder.
Every team was reluctant at first. My first job was to convince the engineers to stop thinking of each website as its own silo. Then, over the years, came the people running each site, including the partnership managers who looked after the sites we ran with other editorial teams.
I went about it the wrong way. I showed finished work and tried to impress people with a new look: a new interface, a new design language. But their site was theirs, and a new look wasn't a good enough reason to give part of it up.
What would have convinced people to come into the framework was the fact that we would be solving problems they actually had. Sometimes they saw it for themselves: a team would see a feature on another site that already lived in our framework and say, I want that feature for mine too.
That's how I should have sold it from the start, by putting myself in their shoes. Starting from the problems each team had at the time, not from the technology or from efficiency, and showing them what they would gain.
What I took away
Selling an idea takes empathy. A better way of working only matters to people if it solves a problem they have today, and a good idea doesn't explain that on its own.
What I care about most in a system, whether it's a design system, a platform or any other digital structure, is efficiency and scalability: how much faster it lets an organisation work and build its products. I'm good at it, and this project is where I found out.
Back then, design and development went hand in hand. All my experiments were done in HTML and CSS, so there was no design file to drift away from the real thing. That's something I miss today. In 2026, keeping design files and code in sync is a struggle of its own.