Category: Articles

Back In The Game

A few weeks ago, I tweeted that I was 'back in the game'; but didn't go into details.

On 22 March 2021, as England’s lockdown eased and kids – specifically, my kids – returned to school, I returned to the familiar world of consultancy and company directorship, with the formation of Tintou Ltd.

I left Automattic in May 2020. I didn’t have a plan for what came next. But I did have an opportunity to put my family first, like I always said I would, as the world suddenly shut down, and everything got weird.

Often, the best way to solve a problem is to focus on something completely different, and let your brain work it out the background. After a few months concentrating primarily on my home-schooling duties, a few conclusions emerged.

What I wanted most was to get back to making things. My time at Automattic had mostly been about creating opportunities for other people to make things; and don’t get me wrong, I look back on it with a degree of satisfaction. But deep down, I was jealous of those taking advantage of those very opportunities.

I had submitted a few – let’s call them ‘ambitious’ – job applications over the summer. I was curious to see if the success of the WordPress working model, and my own credentials in its ecosystem, would resonate beyond it.

The answer was: not so much. It led me to question precisely what I had to offer, and where its value lay. I concluded that my WordPress knowledge and experience was actually more valuable in the WordPress space itself than ever: and it would be crazy not to make use of it.

But I also realised there were a few things I wanted to do, which no full-time job description would ever include. Some connected to WordPress and open source, some not at all. Some arising from specific dates, events and circumstances which may not come around again. And to be blunt, I’m not getting any younger.

So I decided to return to WordPress consultancy: probably the most productive, most flexible and most enjoyable phase of my so-called career. Specifically, in Product Ownership: not a widely-used label the last time I was doing it, but probably the best description of what I do best. And crucially, acting as my own employer, trading the stability of a full-time job for the freedom to pursue those other things.

I’m pleased to announce that my first engagement is with XWP: one of the top-tier WordPress agencies globally for many years, but one which has reinvented itself in the last year or two, with a few notable people leaving, and a few equally notable people joining. It makes for an exciting mix of stability and startup energy. And with the company’s commitment to the open web, and its 100% remote nature, I felt instantly at home.

I knew almost immediately that I’d made the right choice. Into my second week, I found myself grappling with an exciting, but somewhat amorphous project description, working with Chris Lema at Liquid Web. Eventually I unlocked it: digesting some complex third-party data reports, digging around in the WordPress options table, collecting the information needed to produce a crude but indicative wireframe. Which one of my new design colleagues then turned into… something beautiful.

We all knew it immediately. Couldn’t wait to show it to Chris. Happily, he agreed.

I could literally feel it on my face: a certain kind of smile, that I hadn’t smiled in years.

I’m back in the game. Let’s play.

A plea to API owners

‘API-first’ thinking has been around for quite some time now. It’s the notion that you shouldn’t think about building (for example) a website: you should think about building a data API, and then – as a separate thought process – building a consumption or presentation layer on top.

The theory goes that, if you get the data API right, then it’s easy to swap in a new front-end when you fancy a change. It may also allow you to share that data with another service, be it your own or – even better! – someone else’s.

I recently built a custom Alexa skill to address the defining user need of UK local government: which bin do I need to put out on which day? Courtesy of a contact at my local council, I discovered how to access the API which powers the council website. (They don’t offer it as a public service, but my reputation preceded me: in a good way, for once.)

It gave me the data I needed, but in a format which had clearly been conceived only to power the website. Dates didn’t include a year: just the day name, day number and month name. (So, Friday 1 January.)

It would no doubt be obvious to a website visitor, looking at the web page, which year was meant. But computers can’t infer that sort of thing: so, given that I had to sort by date, I had to write some additional logic to work out what the full year would be.

Easy enough. But frustratingly, the machine powering the API must have known the relevant year. I was writing extra code to extrapolate and restore data which could easily – in fact, probably more easily – have been included.

My Alexa skill worked fine for a number of months. Until this past week or two, when the API response changed. If a collection was today, its primary text response was now: Today.

That probably made for a better user experience for the website visitor: indeed, my Alexa skill did exactly that, in constructing the voice response. But my skill wasn’t expecting the change: and it broke. Again, the fix was pretty easy, as the date was being sent in another field; but it was another unnecessary interruption.

So my plea to API developers is as follows.

  • Think of the API as a stand-alone product… Assess its success in and of itself, rather than a means to another end. Make it as helpful, flexible and reliable as it can be. You’ll eventually find a reason to be glad you did.
  • … and think of yourself as Just Another Consumer of that API. Once the API launches, don’t do yourself any special favours.
  • Supply data in the most raw, standard, flexible form possible… In the case of dates, that might mean the number of (milli)seconds or days since 1-1-1970; or RFC 2822; or ISO 8601. They might not look pretty, but this isn’t about looking pretty: these are widely adopted standards, and virtually any modern computer and language can produce and work with them, without extra pre-processing. And even if they include elements which you don’t immediately need, someone else will.
  • … and do any tweaking in your consumption layer. By all means, change the date display to ‘Today’ – but do it in your web page, not in the API response.
  • Don’t change things once the API is out there… Developers, and the users of their apps, require reliability.
  • … but don’t be shy about extending it. If you decide you really need to format something slightly differently, maybe add a new line to the API response – keeping the old one as it was.

Most importantly though: please keep thinking API-first. It’s fantastic that any of this was possible in the first place. If we can get used to doing this in a developer-friendly fashion, it uncorks all kinds of possibilities, and rarely means extra work.

Lockdown could launch a long-overdue digital revolution in education.

For most of the last year, my two children have been home-schooled. Neither by my choice, or theirs. I never really got why some parents chose to home-school, and I’m totally baffled now.

The media reports have focused on disruption to exams. But I can’t help thinking of the simpler formative experiences that kids are missing out on. The sports days, the school trips, and so on. Everything in our calendar has been cancelled, and we’ve stopped even being disappointed.

One up-side, perhaps the only up-side, has been the enforced pivot to digital delivery.

During the initial closure of English schools in spring 2020, the commitment seemed a little half-hearted: that’s not a criticism, probably just a reflection of the initial novelty element, followed by a sense that it wouldn’t last that long. A bit of blitz spirit, and some unusually good weather helped too.

Second time round, in early 2021, things feel very different. We’re in it for the long haul now, and teachers’ expectations are much higher.

Both kids have multiple Zoom calls daily. The local secondary school already used a digital solution for classes and homework; the primary adopted Google Classroom. Work is set in PDFs and Microsoft Office documents; and in our case at least, returned as Google Docs or photographs. Sure, it could all be more efficient, and better organised, but it’s working well enough.

What we’re producing here is a generation of kids who are completely comfortable, not just with technology in general – smartphones and games consoles had that covered – but with the normal, practical technology their parents all use every day at work. And this is a good thing.

At some point, be it weeks or months, the worst of this pandemic will be behind us… and the kids will return to the classrooms.

Do we actually want them to return to an experience built on exercise books and joined-up handwriting? I don’t know about you, but most weeks I don’t find myself handwriting anything longer than my signature. (And even that’s getting pretty rare.) We can stop talking about digital skills in the future tense: they are an integral part of every job, trade and profession now.

I sincerely hope our school leaders and civil servants see this for what it is: the best opportunity they may ever get to leap forwards in our delivery of education. I know you all have a lot on your plates right now, like not killing your staff and pupils; but you need a plan for what happens when we eventually emerge from this tunnel, into the light. A plan which doesn’t abandon the skills our kids have acquired in the last year. Skills they will actually need as adults.

And I hope the digital industry – people like the Raspberry Pi Foundation, experts in open source software, and the production of computer hardware so cheap they literally give it away on the front cover of magazines – recognises that this is their moment too.

Where their focus has historically (and entirely admirably) been on steering kids into STEM – as is immediately obvious when you look at any of Raspberry Pi’s communication, and the ‘maker’ software they bundle into their OS – they must see the greater opportunity here.

It’s said that all jobs are tech jobs now. Well, right now, at this very moment, all education is tech education.

We need to decide if we want to embrace that, or abandon it, when it’s no longer our only option. And if we’re going to embrace it, we should already be thinking about how.

Automattic joins call on EU to defend the Open Internet

It came as a very pleasant surprise this morning, to see Automattic’s name alongside Twitter, Vimeo and Mozilla on an open letter to the European Union, ahead of next week’s delayed publication of the new Digital Services Act.

A consistent problem in recent years has been the failure of second-tier platforms to show up and make their voices heard, particularly in Europe.

Some felt they didn’t ‘do’ politics. Some simply weren’t interested in anything outside the US. Most were happy to ride on the coat-tails of their fastest-growing peers, the likes of Google and Facebook, whose greater revenues allowed them to build public policy teams – first in Washington, then in Brussels. They were our kind of people, and they represented the digital angle well.

Until, that is, Google and Facebook became the problem.

And as my former Automattic colleague Stephen Blythe memorably put it: if you regulate the internet as though only Google and Facebook exist, you will create an internet where only Google and Facebook exist.

We saw this with 2019’s EU Copyright Directive, introducing requirements which realistically only the largest, wealthiest platforms could satisfy. Measures intended to protect ‘the little guy’ in the creative sector could well stymie the growth of ‘little guy’ platforms, thus allowing Big Tech to grow ever bigger. Whoops.

And with the DSA imminent, YouTube’s chief product officer Neal Mohan appeared to call for more of the same last month, insisting that: ‘anything that the DSA does has to apply to all companies, all platforms, not just the large ones but the small ones as well’.

‘We’re writing to you today,’ the letter states, ‘as a group of companies to advocate for a regulatory conversation around illegal and harmful content that firmly roots Open Internet principles at the heart of the EU’s digital future.’

Choosing their words wisely, the signatories note that in Europe and around the world: ‘The laws that laid a foundation for the spectacular growth of the Internet – with its economic, social, and cultural revelations – are now being reviewed to assess whether they’re fit for the next generation.’

‘Reviewed’? Joe Biden was rather more forceful in January, telling the New York Times (see 1h26 in the video at the top): ‘Section 230 should be revoked, immediately, should be revoked, number one, for Zuckerberg and other platforms.’ Challenged by the Times’s tech-savvy Charlie Warzel, formerly of BuzzFeed, he repeated it, twice. The rhetoric may have softened since, but it seems certain that changes are coming on both sides of the Atlantic.

Today’s letter tries to reframe the contentious issue of content removal, calling for: ‘a content moderation discussion that emphasises the difference between illegal and harmful content and highlights the potential of interventions that address how content is surfaced and discovered. Included in this is how consumers are offered real choice in the curation of their online environment.’ Again, a wise choice of words, bringing in Europe’s favourite angles, consumer choice and competition.

It asserts that: ‘in the digital world, content policy can influence the shape of markets’ – again, playing to Brussels’s competition instincts; and very gently brings up the example of the Copyright Directive, noting its ‘regressive impact on smaller players’.

‘In sum,’ it concludes, ‘we ask the EU to defend the Open Internet.’

This is precisely the kind of constructive engagement that I tried to encourage more of during my time with Automattic. Indeed, the lack of such engagement was one of the primary factors in my decision to move on. If the company has finally found its voice, and some resourcing, I’m delighted.

Can we really claim that 40% of all websites run on WordPress?

As WordPress breezes past another market share milestone, and the next comes rapidly into view, it’s worth pausing a moment to consider what the most commonly cited figure really means, and how we should communicate it.

It counts ‘domains with WordPress’.

As I explained in an earlier post, W3Techs considers a site to use WordPress if it finds evidence of WordPress on the main domain, or any subdomain.

I cited the example of Craigslist: online since 1997, currently ranked no 142 by Alexa, clearly not running WordPress as its primary technology. But it has a blog, running on a subdomain. And that’s enough for W3Techs to count as a WordPress site.

So it’s wrong to say that ~40% of the world’s top ten million websites are built on WordPress. It’s more accurate, albeit less impressive, to say that 40% of the world’s top ten million websites use WordPress for at least part of their overall online presence.

It isn’t uniform.

W3Techs offers a tantalising breakdown of WordPress usage by popularity of site domain, on its pages comparing one CMS with another (eg WP-Drupal). As I write this, it shows:

The overall WordPress market share, across more than ten million sites domains is 63.8%: but the more you concentrate on the bigger and busier properties, the lower it goes. Or to put it another way: WordPress is stronger in the ‘long tail’.

But there’s nothing inherently wrong with that. Indeed, for a project whose stated mission is the democratisation of publishing, one might argue it’s a good thing: putting easy-to-use, powerful tools in the hands of the many.

And anyway, ‘more than half of the world’s top 1,000 websites using WordPress for at least part of their service’ is not to be sniffed at.

It may not reflect active usage.

One of the many charms of WordPress is how easy it is to get started. The project still boasts of its ‘famous five-minute install process’: these days, it’s often a lot faster – and in many cases, pre-installed before you arrive.

But it’s also very easy to abandon. As free software, often running in cheap hosting space or alongside other code, there is often no hefty annual invoice to remind you of its existence, inviting you to wonder if it’s still ‘worth’ paying for.

Let’s take that Craigslist blog as our example again. The last post on the site was April 2017, three and a half years ago. The site is still running the Kubrick theme, which first appeared with WordPress version 1.5 – released in February 2005. It’s hosted by Automattic, on, and probably only costs them a handful of dollars per year.

But it still works. As it’s on a hosted service, it gets regular software updates – and of course, with auto-updating now in WordPress core, the same would apply these days if it were self-hosted.

So it’s probably more human effort to take a site down, than to simply leave it up. The site may be dead to all intents and purposes; but whether it’s admirable dedication to the cause of permalinks, or sheer laziness, dead sites still count.

So, a little humility.

None of this is to question the success of WordPress in its space. Regardless of the measurement’s weaknesses, WordPress is still far and away the most used CMS out there. And even if a large corporate user only uses it as a secondary solution, it’s still a valid endorsement of its capability.

There’s plenty of pride to be claimed for WordPress, and with ample justification. But we owe it to the industry, and ourselves to be accurate in what we claim.

WordPress is now the undisputed leader in content management

The open source publishing solution has passed the latest milestone in its continuing growth. But where do its ambitions end?

The latest data from independent analysts W3Techs, considered the most reliable source in the industry, reveals that WordPress now sits at the top of the ranking for Content Management System usage.

Hang on, you may ask, wasn’t it top already? Yes and no. WordPress has been the technology identified on the largest number of sites – or maybe more accurately, domains – for many years now. And for quite some time, it has been more popular than every other identifiable technology combined.

But there has always been one category smiling down from the top step of the podium: ‘None’. Or perhaps more accurately, ‘None that we could identify‘.

As I explained in a previous post, W3Techs has no special access to data or websites’ back-ends. They simply look at externally visible clues and signals, and work it out as best they can.

Their list covers literally hundreds of CMSes: but still, more than a third of the time, they come up blank.

That could mean a site has been produced using an in-house CMS, or conceivably, 90s-style with no content management solution at all. That’s what people often understand when they see ‘None’.

But equally, it could mean the developers have done a great job of covering their tracks, hoping for security through obscurity. Or increasingly, it could mean they’re using a decoupled arrangement, with the CMS connection reduced to mere API calls. (And of course, in these cases, it could still be WordPress behind the scenes – as leading static site-generating solution Gatsby is now doing.)

When I’ve talked about market share up to now, I’ve always been careful to soften my assertions. It’s theoretically possible that all those unidentifiables are using the same CMS. ‘Manual’ is a content management strategy of sorts. But today, with the WordPress number now higher than ‘None’, we can be a little more assertive.

So where do we go next?

Obviously, the next milestone coming into view is the nice, round 40% threshold. And then it’s 50%, when a majority of websites are identifiably running on WordPress.

But perhaps a more interesting number to look towards is 51%, an explicit objective for WordPress co-founder Matt Mullenweg for many years. He referenced it directly in the announcement of Automattic’s acquisition of Woo, back in 2015, when WordPress market share was a mere 23%.

Team 51 is also the name of a team within Automattic, originally formed to build market awareness of WordPress and its capabilities. These days, of course, market awareness is less of a concern; and they now prefer to be known, externally at least, as the Special Projects Team. The mission statement on their team homepage is rather woolly; Jeffrey Zeldman put it better in a 2019 blog post:

Team 51 is a design agency within Automattic. With a preference for clients who do good in the world, and for assignments with the potential to expand how businesses and agencies view WordPress, we make websites — for charity, celebrity, and influencer clients — that show off what WordPress can be.

In practice, the team operates as Automattic CEO Matt Mullenweg’s private development team, offering design and migration services to projects he wants to work with, often on – let’s say – generous commercial terms. Nothing wrong with that at all, of course: but I often felt more could and should have been done to lever these as case studies.

Beyond that? Eyebrows were raised when, announcing Salesforce Ventures’ $300 million investment in Automattic, Mullenweg suggested that WordPress had ‘potential to get to a similar market share as Android, which I believe now has 85% of all handsets’. He later recast this as a ‘trailing indicator’ of user satisfaction, as others questioned the desirability of such market dominance.

But such things shouldn’t concern us for some time. WordPress has shown remarkable resilience over the past decade, maintaining steady growth month upon month, seeing off all challengers, whilst its historic rivals Joomla and Drupal fell into steadily decline.

50% doesn’t seem as far away, or as fanciful as it once did… and perhaps we should be thinking more about its implications.

Understanding that CMS market share number

You’ll often hear it said that WordPress is the world’s most popular content management system, powering xx% of all websites. But where does that figure come from... and what does it really mean?

WordPress’s share of the CMS market is a continuing good news story for the platform. It’s the centrepiece of many presentations and pitches, as incontrovertible evidence that WordPress should be taken seriously. 50,000,000 Elvis Fans Can’t Be Wrong, right?

The primary source for the headline market share number, the one you hear quoted most frequently, is the rolling study conducted by W3Techs, an independent software consultancy based in Austria.

From a sample set derived initially from Alexa’s list of the top 1 million, then 10 million websites from June 2013, and supported since April 2020 by the Tranco list produced by European academics, they use publicly available information to collate statistics on websites’ usage of technologies and related services.

Sites are analysed ‘approximately once per month’, and data is updated daily. You’ll find plenty of enlightening data free of charge: W3Techs also sends monthly reports running to several hundred pages to paying clients.

Their website isn’t the most user-friendly or intuitive; but it allows you to analyse recent trends, drill into certain details, and combine metrics for some interesting insights.

So for example, I can call up a combined report showing how many websites hosted by major data centre providers use WordPress; or a comparison of recent market share across selected CMS options.

Each site they monitor gets a page of data on their website, listing the technology detected – for example, There are free browser extensions for Chrome and Firefox, which will let you call up this information on demand via a toolbar popup.

The W3Techs team are quite open about the challenges and limitations of their work. Their goal is to be ‘as accurate as one can possibly get’, based on what they can see; and they say with some modesty: ‘We believe that we are not too far away from that goal.’

Having observed them for many years now, I’ll certainly say their numbers have the ring of truth about them, and aren’t subject to the same wild fluctuations seen in certain other sources.

However, the data can sometimes be cited or interpreted incorrectly.

You’ll often hear people quote it as ‘X percent of the internet‘. Of course, there’s much more to the internet than just browsing websites. Strictly, you should say ‘X percent of the top ten or eleven million websites‘ – but it’s surely fair enough to trim that down to ‘X percent of websites‘ in conversation.

But perhaps the most important point to bear in mind, particularly where WordPress is concerned, is that W3Techs do not consider subdomains to be separate websites.

This can produce some initially confusing results: Craigslist clearly isn’t powered by WordPress, but it’s listed as a ‘popular site using WordPress’. You’ll see it stated on its site profile page, though, that WordPress is being used on a subdomain by a secondary site.

I’ve confirmed with W3Techs directly: this means that is counted as a site running WordPress, even though ‘Craigslist proper’ isn’t. But I’m told I shouldn’t expect this to make a big difference: ‘very few of the smaller sites bother to use a subdomain, and even less bother to use more than one CMS for their website.’ (I’d love to see hard data on this, though.)

This also means that the many millions of websites at which do not have a mapped or custom domain applied, all count as a single site.

But since the analysis only covers the top ~10 million sites, and few of these sites still using the free URL generate much traffic, it’s unlikely to be affecting the data to a significant extent.

I just quit my job. Because there’s work to be done.

The time has come for me to move on from Automattic, and the WordPress VIP team. I'm open-minded about what comes next. And open to offers.
On stage with Automattic CEO Matt Mullenweg

I felt a sense of happy inevitability when I joined Automattic, the ‘foster parent’ company of WordPress, at the end of 2014. I loved working with the software, but more than that, I loved what it stood for – and Automattic provided the foundations it stood upon.

In Automattic, I could see a company committed to bold principles of user and employee freedom. One of the first companies to recognise and actively embrace the emerging realities of a hyperconnected world. A company with both the desire and the potential to shake things up.

I joined an Automattic of 300 people. Tomorrow, after five and a half years, I will be leaving an Automattic of 1,200 people.

In some ways, it’s a very different place: that’s good in some respects, not so good in others. In other ways, it’s remarkably similar: that’s also good in some respects, not so good in others.

I’ve been very happy to see many parts of the business – notably my own area, the VIP enterprise services team – growing larger, more mature, and more stable.

A few years back, I gave a talk at the annual all-staff meetup, drawing parallels with Ikea: founder Ingvar Kamprad’s 1976 corporate creed The Testament Of A Furniture Dealer maps surprisingly well to Automattic’s ethos. My conclusion was not simply that it’s possible to build a successful business on the principle of democratisation; but that if you want to deliver democratisation, you need to have a successful business model beneath it.

But there’s never been any doubt in my mind about where my own motivation lies.

Royal College of Art, 2016. Photo: John Maeda

Each year, Automattic runs an employee engagement survey. There’s a cheeky little kicker right at the end: ‘what’s the one thing that would make you leave?’ I always felt that was a dangerous thing to ask. It forced me to define what the test would be, if I ever suspected it might be time to move on.

My answer, each time, was that I would leave when I no longer felt I was in the best place to have the impact I wanted.

Returning to work earlier this year, having taken up Automattic’s very generous offer of a sabbatical after five years service, I found myself applying the test. With the clarity of mind that comes from three months away, and with a heavy heart, I concluded that my time had come.

I’ve been doing this thing long enough to know what I’m best at. But recent evolutions in my job description were moving me away from that, rather than harnessing it. Sometimes, what’s best for the team isn’t what’s best for every individual on the team.

I don’t have a next gig lined up yet; and I’m open-minded about what that next gig might be.

Perhaps it’s time to bring some truly internet-generation startup thinking to some larger, more established contexts. Perhaps I should be working through the second-order consequences of a world that has been transformed by open source and global connectivity. Perhaps a combination of those.

But whatever I end up doing, it feels like an exciting time for a fresh start.

Automattic’s magic formula of distributed working, radical transparency and global cooperation once seemed pretty bold.

Right now, as we ease out of the coronavirus crisis, and remember the various crises we all put on hold, it seems like the only reasonable way forward for the entire planet.

I’m available. We should talk.

So many lovely responses to my news on Twitter. Thank you all.

What I’ve been up to on my sabbatical

One of the perks of working at Automattic is the offer of a 2-3 month sabbatical after five years service. Most people take the opportunity to travel, to disconnect, or to achieve an ambition. I triggered mine pretty much on the fifth anniversary of my joining the company; and I've spent the first half of it doing... web development.

I’ve always been interested in politics and elections: even as a kid, I’d stay up late to watch the live coverage of by-election results. So as it became obvious that a general election was looming in the UK, more or less at the time I’d become eligible for my sabbatical, I saw an opportunity to get involved in the digital efforts of my favoured political party, the Liberal Democrats.

I had hoped to find a party machine into which I could slot for a month or two; but that wasn’t really how it worked out. Instead, I got in touch with leading Lib Dem digitalist Mark Pack, with whom I’d worked in the past, to seek his support for a few ‘independent’ experiments.

Continue reading “What I’ve been up to on my sabbatical”

How Leavers and Remainers get their news

Oxford’s Reuters Institute for the Study of Journalism recently published a report on the media habits of UK consumers: they took 2019 research data from YouGov, and broke it down by whether they voted Leave or Remain in the 2016 Brexit referendum.

With Brexit framing the entire election campaign, its lessons must have at least some relevance to the next few weeks.

For both camps, the most used news sources, in order, were the BBC, ITV and Sky: the three national TV broadcasters, although of course, they all also have online presences.

But in the context of content sharing ahead of the election, it’s especially interesting to note the mismatches between the brands people use to get their news, and the brands they trust.

Channel 47.1213%
Channel 46.2410%

Remainers have a high degree of trust in Channel 4 and the Independent, but they are highly unlikely to get their news from them. Both sides have a high degree of trust in the (pro-Remain) FT, but few people actually read it.

So if you’re targeting Remain voters, and you find favourable articles on these sites and feeds, it’s well worth giving them some social amplification.

The pattern of usage of different social networks is remarkably consistent between the Leave and Remain camps: Facebook and Twitter are first and second for both, but Twitter is a much closer second among Remainers. Leavers are slightly more likely to use YouTube for news. The two sides are equally (un)likely to share and comment on news in the average week.