Me and Gutenberg

Visiting the Gutenberg Museum in Mainz, Germany was a bucket list experience for me. I’d been at WordPress VIP for four years and the block editor in our CMS is called Gutenberg, so I kept trying to get them to host one of our team meetups close enough for us to visit the museum.

In March, we had a product team meetup in Berlin and a Lufthansa strike meant I might not make my connection home through Frankfurt. So I canceled that first flight and took a train to Frankfurt, leaving at 6am so I could spend a few hours in Mainz.

Why did this mean so much to me? After college, I worked with magazines and four-color CMYK printing. But really, it all started with a tour of Marvel Comics.

My dad wrote Stan Lee a letter asking for a tour of their offices for my 11th birthday. That Friday, instead of going to school, my dad, my sister and I rode the subway into Manhattan and we spent the day on the 6th floor of 575 Madison Avenue meeting artists and writers and editors. It was a surprise, but I clearly remember figuring out what was about to happen because the floor indicator over the elevator door had a picture of Spider-Man’s face on it for the sixth floor. Wow.

Besides dozens of free comics, two things I got that day that changed my life were John Buscema and Stan Lee’s book “How to Draw Comics the Marvel Way,” which taught me how to draw, and an acetate overlay from an ad for Elric that was drawn by P. Craig Russell. There were four acetate sheets, each with one of the yellow, magenta, cyan and black layers, stapled on top of a white cardboard backing. So you could flip them up and down and see how CMYK brings to life a fully-colored picture. It blew my mind.

A similar CMYK example from the Exploratorium in SF.

After college, I worked in desktop support for magazine companies: running Novell servers, fixing printers, troubleshooting fonts in Quark XPress. Soon, I got to be an art director myself and create magazines with my friend Jason. Other friends of mine like Glenn worked in digital pre-press and their offices had posters like “Whoever Dies With the Most Fonts Wins” and external hard drives full of fonts.

That’s how I got publishing in my blood, all before the web happened and CMYK gave way to RGB.

Around the time of my museum trip, Matt recommended a book by Matthew Butterick called Practical Typography. One of the great things about Automattic is that when they recommend a book, everyone in the company can buy it and expense it. We support creators and publishers in so many ways.

Butterick’s book is fantastic and claims to explain typography in ten minutes, which is bold. Even more bold was his claim that he can do it twice:

All it takes is ten minutes: five minutes to read these rules once. Then five minutes to read them again.

So March was a magical month for me, and had me thinking a lot about publishing history from NYC in the ‘90s all the way back to the 1400s.

And here I was, finally at the Gutenberg Museum.

Around 1450, Johannes Gutenberg created the first printing press. It used small interchangeable cast metal letters to transfer each line of text onto a page. Much of the typography terminology we still use today is from the Gutenberg days: leading, logotype, ligatures, upper case and lower case.

Before Gutenberg, scribes copied books by hand on scrolls and paper, or print-makers printed texts from hand-carved wooden blocks. Either process took a long time; even a small book could take months to complete. Because carved letters or blocks were flimsy and the wood susceptible to ink, the blocks had a limited lifespan. — Wikipedia

None of the devices there were from the 1400s. The original presses weren’t built to last for centuries, and maybe no one at the time knew how important they were. But a few of the original run of Gutenberg bibles do remain and they were on display in Mainz.

Upstairs, the museum had a room where you could do hands on work with inks and metal type. I had to make it to my flight, so I walked around and took some pictures.

That was an amazing day.

You’d think typography doesn’t matter now that everyone flicks through websites and apps on their phones, but it does. In the book Practical Typography, Butterick points out how important highway sign font choices are. Highway Gothic, the 1940s font choice for U.S. highway signs, was carefully designed to fit different text lengths and for fast legibility at high speeds, but that was replaced by Clearview in the 1990s to improve legibility at night when bright retroreflective headlights cause letters to blur together. Type saves lives!

When we built the collaborative design platform Ceros in 2012, our engineer Rus converted fonts into JavaScript glyph functions that could render 20,000 characters as they were dragged across a multiuser canvas at 30 frames per second.

When I was in Mainz, one of the most talked about web technology breakthroughs was a TypeScript text layout engine called Pretext that was 300-600 times faster at measuring and laying out multiline text in a browser, including wrapping text flow around objects as you drag them.

And one of the global media companies we’re working with at VIP asked me to help make WordPress the core of their print workflows in addition to their digital publishing. I cannot wait to see what WordPress and modern AI can do for newspapers.

Thank you, Johannes!

Who will release the first great Search Console for AI?

Publishers can see the clicks AI sends them.

They can’t see the answers it builds without them.

The next big breakthrough for publishers won’t be another AI model. It will be a Search Console for AI.

Comscore’s new AI Intelligence Report shows how quickly discovery is fragmenting. Between January and June, ChatGPT’s share of AI prompt volume fell from 70% to 50% while Gemini and Claude gained.

Optimizing for a single model is already a losing strategy. Yet publishers still can’t see what happens between someone’s prompt and an AI-generated answer.

Measurement is the blind spot. A publisher’s website and CMS provide a durable foundation for AI discoverability, but your site analytics show only the people who click. They don’t show the prompts, impressions, mentions, and citations that don’t get clicked.

Google has begun closing that gap. Its new generative AI report in Search Console shows impressions from AI Overviews and AI Mode. But it still doesn’t show clicks, click-through rates, the queries that triggered an appearance, or where your citation appeared within an answer.

Publishers need that information.

Will one great AI console solve the problem?

No.

The fragmented AI market means publishers will eventually need a unified view across all of the major services. But that view can only be built after each platform exposes the first-party data only it can see.

So keep making your content authoritative, structured, and original. That isn’t just practice. It’s the engine you need.

But today, publishers are still tuning it by ear.

The first AI platform to offer a full dashboard won’t solve cross-platform measurement. But it will set the standard and earn publishers’ trust.

Running on empty

With 10 days left in September, I had used up 96% of my Claude Enterprise tokens.

Oh no!

(I use the enterprise plan because WordPress VIP works with big companies like Salesforce, Meta and Jefferies and government websites like NASA and The White House.)

I tried to look at this as a positive sign. A milestone. I spend so much time talking about using AI at a high level, but I rarely have time to code and I always feel I’m stuck in first gear. Like an imposter. So it was a kind of accomplishment.

But seriously, what was I going to do for those 10 days?

Catch up on podcasts?

Take long walks?

Learn a new language?

Retire?

Fortunately, we had a four-day weekend trip planned to go to the Bourbon & Beyond music festival. It was my first visit to Kentucky and we had a blast, along with more than 200,000 other people. For an event of this magnitude, it was unbelievably well-run. Nice work Danny Wimmer Presents.

I’m in that building on the right.

We enjoyed every band, but the ones that really amazed me were Lindsey Stirling, Foster the People, Mumford & Sons, Squeeze, Hootie and the Blowfish, Kaitlin Butts, The Verve Pipe, Clover County and Vertical Horizon.

Two of my favorite bands: Foo Fighters and Squeeze

I am now less than 2 hours away from my token usage being reset.

Not enough time for a concert. Maybe I’ll take a long walk.

Be careful what you wish for

Search engines have always worked like a mall directory. They showed you where all seven sneaker stores were and how to get there.

Today’s answer engines are more like genies. You need sneakers? They just hand you the sneakers. No listings. No browsing. No shopping. No visits. No clicks.

This is worse news for publishers than it is for marketers and commerce sites. If you’re a car company or pest-control service, the answer agent shortcuts the shopping process, but they still send you a customer who’s ready to buy.

But if you’re a publisher? Your content is the product. How do you make money when the answer engine shortcuts the whole reading journey?

Last week, I talked with Brian Boyle at The Daily Upside about zero-click, GEO, crawler controls, and the strange new job of making bots love your brand.

We also talked about the publishing industry being in a decline.

One size doesn’t fit all

Cloudflare just segmented AI traffic into three types: Search, Agents, and Training, and they’re letting their customers turn access on or off for each type. That’s better than a single on/off switch for bots, but for a high-end website that isn’t enough control.

The real control businesses need is for each service within those categories. They need to decide which crawler gets which content, for what purpose, and under what terms.

This isn’t hypothetical. One of WordPress VIP’s enterprise customers might want to allow every bot to access their site except Perplexity, because they’re negotiating a deal with them. Another maybe has a deal with Perplexity and loves them, but needs to block OpenAI. And many customers just want all bots redirected to TollBit to collect micropayments for accessing their content.

Giving three on/off switches isn’t enough. Enterprise customers need fine-grained controls, down to the individual service.

Antifragile rocks

For decades, ranking on Google was survival. Now ADWEEK reports that some of the biggest media businesses are not-so-quietly making plans to disappear from it.

This isn’t just one angry newspaper. The Wall Street Journal lists USA Today, Reuters, Politico, The Economist and People Inc. among the publishers who are considering cutting off Google. And Cloudflare, which sits in front of a fifth of the web, is flipping its defaults to block the crawlers that scrape for both search and AI training.

What changed?

Publishers have always known this dependence was a risk. This is not a revelation. What changed is that the relationship has stopped being a two-way exchange. It’s lopsided.

Neil Vogel, CEO of People Inc., calls Google an “intentional bad actor” for using the same crawler for both search and AI, but he admits People still can’t block it without losing the search traffic it depends on. People Inc is “probably heading towards more confrontation than productivity.”

(I’ve watched Vogel speak a few times recently and he inspires confidence. He seems to have a firm grasp of the industry and an actual plan.)

This Google trap is why the open web matters. It isn’t because of nostalgia or a blind love of all things open source. It’s because ownership and portability increases resilience.

Traffic comes and goes. Ownership compounds.

But the sharp media newsletter A Media Operator makes the case that “No One is Going to Block Google” and the math backs them up. When a publisher still gets over 40% of its traffic from search, walking away from Google is a top-line bloodbath. Google’s Liz Reid reinforces the symbiotic relationship: “Search is not a useful product if the web is not a thriving product.” And Future’s Mike Peralta asserts: “[AI is] nothing but a hallucination without our content.” The industry is also too fragmented for everyone to leave Google all at once, and that’s the only version where blocking actually moves Google.

So will publishers push the big red button? I doubt it, and it doesn’t really matter if a few of them do.

The preparation is the actual prize here. Building newsletters, direct audience, licensing revenue, and owned infrastructure makes you stronger whether you ever block a single crawler or not. Silicon Valley has a word for this: antifragile, systems that get stronger under stress.The exercise of getting ready to survive without Google is exactly what makes you the kind of business that never has to.

The goal isn’t to leave Google.

The goal is to build a business that could if it has to, and no longer has to care whether it does.

AI isn’t killing websites, but their job is changing

AI isn’t replacing websites. But it IS changing what they’re used for.

For years, websites have been the destination.

Today, they’re your system of record that powers AI interactions everywhere else.

I got to join Lauren Goerz on The Dialogue Architects Podcast to explore all the ways that AI is reshaping enterprise content management, digital experiences and the future of work.

We covered enterprise websites as the source of truth in an AI-first world, how answer engines are changing content discovery and totally disrupting traditional customer buying journeys, and why governance and accessibility matter even more as adoption grows. We also talked about jobs to be done (JTBD) as a better way to divide work between people and agents, and what conversational interfaces could mean for enterprise software.

Thanks Lauren for a fun, wide-ranging conversation. You do great research and run a fantastic show.

We did a lot on this one!

Watch us here:

I’ve seen things you people wouldn’t believe…

Our big “Future of the Web” research report says that people hate sites that are robotic. They hate digital experiences, things that feel synthetic.

Does it really need to be like this?

The first sites I built in the ‘90s were all by hand. There were no CMSes. Then we hooked up databases to websites. In 2000, browsers added a way to let people make content bold and add links and images. From there, filling websites became a big machine, an assembly line. That’s who I work with. Large sites with large teams making thousands of new stories every day for millions of visitors.

You NEED a machine for that.

Heck, if you want to compete, you NEED AI for that.

But does it have to feel like robots made it?

Two-thirds of people can’t name a single brand that’s doing AI well in their own messaging. They can’t name one.

So the bad news is: everything is terrible.

I guess the good news is: maybe YOU can be the one who isn’t terrible?

And if you do manage to cut through all of the AI slop and reach people with a story that’s organic and human, one that feels like a person made it, maybe your audience will stick around and your business will win.

AI-native developers vs. legacy security models

Your developers are completely AI-pilled. They live in Claude Code, Cursor, Windsurf. But your security model doesn’t.

That’s why Secure MCP matters for WordPress VIP. Your agents run as a signed-in VIP Dashboard user. They can only do what you allow them to do. Trying to perform sensitive operations or accessing high-risk resources only happens if you pass our Step-up Authentication. Every command you run gets recorded in our immutable MCP Call Log. Secure MCP gives teams more than 50 platform tools through one governed endpoint: apps, environments, domains, logs, metrics, users, and WordPress content workflows too.

And there’s a kill switch that pauses all access in under 60 seconds.

That kill switch is really important, because AI governance is meaningless if it can be ignored.