The quote

“Our whole design org is eight people. All eight are on Mobbin.”

A textbook-perfect hero testimonial. Concise, human, as b2b as you can get. It made our Figma file, sitting bold and loud next to the name and photo of a real designer at Plaid.

Except he never said it.

Let’s rewind. The Mobbin homepage refresh team had agreed, unanimously, to feature a hero testimonial quote pulled from our customer stories (that weren't published yet, but that's another story.)

Since I'd workshopped all the stories in trusty bestie Claude anyway, I got him? her? Let's go with him – to help.

The eight-people line was one of them. I sent it to the team. Everyone loved it, and into the design it went.

I'll summarize my sleuthing: Claude drafted the line as a placeholder for a completely different company. The number was real – eight people at that company really are on Mobbin. It flagged it as "placeholder". But when it travelled from .md to chat, chat to Slack, Slack to Figma, somewhere along the way we lost the label and were left with the pretty-sounding quote.

Ironically, the way it got caught is the way it happened. I pasted a screenshot of the design into Claude for a quick copy audit, and it flagged its own placeholder. Oh horror.

So we fixed it: the eight-people line got attribution to its actual owner, in their actual words (which were better anyway):

“Eight people in our org are on Mobbin, and each one signed up independently.”

And John got John's proper quote:

Our cast

How many people does it take to ship a homepage? My first project at Mobbin was our onboarding email refresh and we made the mistake of having way too many cooks (each with excellent taste, which turns out to be the problem.)

So our supporting cast this time round was:

• 1 CEO: Jiho, in charge of Mobbin DNA

• 1 marketer: Rebekah

• 1 head of growth: Jovin, who magically appears whenever a fact needs checking, a chart needs charting, or Rebekah needs unblocking

• ? designers: for some reason none of us thought to scope this out prior to the project

After round 1 of presentations featuring my enthusiastic but ill-advised attempts to vibe design and vibe code my ideas, this very quickly became 1 designer (Alex).

Which very quickly became 1 designer + 1 more senior designer (hi Cleon, who was originally told "no pressure on this homepage stuff just feel free to chime in if u r interested").

Which then became 1 designer + 1 more senior designer + 1 head of design (David).

(If you can't already tell, Mobbin is a design-first company.)

And our AI agents.

I'm sure my teammates use others, but I'm a Claude person so we'll stick with Claude here.

And with our 3 designers' powers combined, they created...

...this Figma file of dreams or nightmares, depending on where you are in the (designer-engineer) spectrum.

(Not featured: 269 comment threads)

31 heroes died for this

Cristina Cordova posted something that explains the next six weeks of my life:

Everyone thinks they can do marketing. Nobody weighs in on the database schema. Everybody weighs in on the landing page.

She's right – and our page was better for it.

Things usually go something like this: headlines are proposed. Everyone has feelings. Feelings, everywhere. Everyone attempts to articulate why they like or dislike a headline. Rinse and repeat.

Instead of walking you through the process, I'll show you a glimpse of the graveyard.

The bar every contender had to clear, from my very first pitch doc: does a stranger know what Mobbin is and does within 5 seconds?

Let's go.

...it goes on.

After everything, the winner was the line from the first week.

The hero was only half the fight, by the way. The subhead under it went through its own versions:

• Do we say "Mobbin"? (uncommon to refer to yourself in third person, said David; it's our identity line, said me)

• Do we want to use a colon? I like colons, but I deferred

• "iOS and web" vs "iOS, Web, Apps, Sites"

• "new every week" or "updated weekly"?

For that last point, I said "updated weekly" sounds like maintenance, rather than new additions. My message got a bunch of emojis and no replies, so the previous line stayed.

Most of our decisions ended exactly like that, by the way. You'd think that things pass with an approval, but most of the time it's the opposite: things pass when nobody has a strong enough opinion to protest loudly.

You can't win an argument with taste

This is one of my favorite lines to say in 2026: "Everyone and their mother's AI has an opinion."

You and I and Claude can have all the feelings in the world – "feels flat", "feels weak", "a bit convoluted", but this generally does not solve anything unless everyone else feels the exact same way as you.

What ended arguments half the time was pulling up what another company shipped. For example, checking Linear's live homepage mid-debate and discussing their framework. Auditing Cursor, Framer and Attio navigation menus to kill a mega menu idea.

We did have one unfair advantage in the reference department: we are Mobbin. We settle many of our discussions by pulling from our own product.

I'd be workshopping in Claude and ask directly, "run mobbin mcp - pull top saas homepage hero examples and summarize patterns"

...Mobbin's MCP would pull web screens and analysis directly into chat.

AI sometimes gets things wrong, though, like with the customer quote in chapter 1.

At some point, Claude surfaced a "strong hero reference" from Vercel, but I sensed that something was a little off. I opened the actual site to check, and yep – it had pulled a section from the middle of the page instead of the hero.

Lesson here: AI can do amazing things. Two years ago I'd have laughed if you'd told me I could workshop an entire homepage in AI AND it could even pull references to validate decisions and inform choices while I worked. But – always always fact check and trust your gut.

Interestingly, AI also fact checks you. It was able to pull up things I'd forgotten that happened and correct a few I remembered wrong. While writing this very article, I told Claude the fake quote got caught by a teammate in Slack. I was sure of it. It searched our Slack, found nothing, then walked me through what actually happened: I'd merged two different incidents into one better story.

So my takeaway is, for now: fact check AI, and let AI fact check you.

And if this whole workflow sounds embarrassingly basic to you – like everyone else has some 147-agent fleet while we're over here pasting screenshots into a chat window – I'll leave you with something Elena Verna (head of growth at Lovable, an actual AI company) wrote: even she feels "like everyone else has cracked the AI code while I'm still using some basic tutorial version over here."

She asked people who claim AI changed their life to show her their workflow. Turns out they’re mostly summarizing Slack.

Nobody has the one-and-done, mind-blowing, AI-will-enter-the-Matrix-and-upgrade-your-life process. This is the process. We’re good.

Product-led or brand-led?

The actual war of the project was, at its heart: do we want to sell features or narrative?

We knew that we wanted something in between, but how do we get there?

My original pitch was a 30/70 page: 30 brand, 70 product. (I like to define brand vs product as: why should people care vs what do people get.)

I wanted brand to breathe in the problem section, the origin story, and the copy tone throughout.

This is how that went.

• I pitched an origin letter: cut. Parked for a proper About page someday.

• Descriptive subheads: cut, because "my hunch is most people don't actually read" (said Cleon. I mulled this over for the briefest of moments, then conceded.)

• Problem section: survived, phew.

That last one deserves its own eulogy. Version one started with naming what design research used to cost.

I loved it. It got proposed for deletion four times, shrank every round, and at one point actually disappeared entirely. It wasn't officially cut during a discussion or flagged in a comment; it was just...gone, and I had to ask for it to get added back. (All good now though, because what made final was the best iteration of it.)

One of the stories I fought hardest for got reworked 4x, almost cut, then ended up as two lines right at the close of the footer.

So a big part of the 30 in my 30/70 became, in the end, two lines in the footer.

I'm so happy they made it in though, and I'd still defend them. They're the only place where the page tells our origin story.

Words vs pixels

I started my career in advertising, so I'm very used to the copy <> design dance. For people who aren't, it can be simply put as: 1 + 1 = 3. The two need to work together, and concessions must be made for that to happen. But when they do click, magic happens.

One example. Our hero has little app icons sitting inside the text (so cute! go look). When Cleon first built this style, the headline didn't fit the layout, so he put in a new variant and popped it into Slack as a WIP. The copy change was disclosed in an italic footnote.

I think my heart almost stopped.

The entire concept had changed with that small tweak, but really nobody was being careless or unintentional. The layout needed shorter words, and the shortest words meant something different.

That's kind of the point of this chapter, because so far I might be making it sound like the designers kept almost breaking the words. It's the opposite, which is that the friction with them made the words so much better.

They also see things I can't. Alex once flagged that "find" and "work" were hanging awkwardly off the ends of a line. He also had three layout options drafted before I'd even processed the sentence.

By purging subtitles, we were forced to really prioritize what we wanted to communicate and use copy economically. (Also, our audience is mostly designers. Cleon was right.)

Layout notes, design dances and copy expanding and cutting are almost never attacks on either craft. Every layout decision is also a copy decision, and vice versa. The same discussion thread that nearly stopped my heart also produced the best-looking hero we've ever shipped.

It shipped!!!

The subheader was still being debated suspiciously close to launch. But looking at the progression, we were getting surer and surer – the debates and flags kept getting smaller. When all the attention is on the tiniest details, that's when you know you're about to cross the finish line.

Also, this kind of work does take its toll. As much as we all love our work and craft, looking at the same page for weeks on end drives anyone a little batshit crazy.

For the record: we have the absolute best designers at Mobbin. They took six weeks of feelings, fake quotes, discussions over right inner paddings and rendering quality, and made a page that looks like none of it ever happened – the homepage came to life better than I'd dreamed of.

Now comes the part Cristina also called: the scoring is asymmetric. If the new homepage performs, we know we built a great product. If it doesn't, marketing will be hearing about it.

I'll let you know if I get called in.

Top tips for designing a homepage in 2026

To be clear, the craft is real and learnable. Just Google it or ask ChatGPT about page structure, hierarchy, what a hero does, top copywriting formulas. There are 101 guides on those things. What nobody can give you a process for is the part this article is about: getting a group of people who care about different things to agree on how to apply it all to your homepage, specifically.

With that said:

1. There is no process. The process will depend on your team makeup, what your company and brand value, and who gets final say. The faster you accept the mess, the faster you can move through it.

2. Align on the brief before anything else. It's the main tiebreaker that is tangible. "This one hits the brief" ends debates; "I like this one" never will.

3. Commit, but make it reversible. "We can change it if it doesn't perform" is the sentence that helps everyone let go.

4. Your first good idea might be the winner. You'll spend six weeks confirming it anyway.

5. Always try the next version. If it's getting stronger and simpler, keep going, you're on to something. If everyone's starting to get muddled and you're birthing a Frankenstein, stop and go back to the last version nobody opposed.

6. When the arguments shrink to "do the app logos look a little blurry?", it's time to ship.

See the homepage in action: mobbin.com

Send me a message or connect on LinkedIn or X