466,000 views. Thousands of reposts. Heated debate across X, Reddit, and the inboxes of product teams everywhere.

And the most meta part?
I used an AI browser to write a viral thread about AI browsers.

Specifically: Comet, Perplexity’s new agentic browser. I let it draft, research, pull logs, benchmark competitors, and even generate the visuals.

If you’ve been following the AI browser wars, you know this isn’t just about features. This is about the interface layer between humans and machines. Where action happens. Where execution lives.

But here's what everyone's asking: HOW did an AI browser create content about itself?

Time to pull back the curtain. This is the full playbook 👇

Before we go any further, let’s break down the key players. If you’re still using Chrome, this might feel like sci-fi. But here’s the new browser landscape, built not just to load pages, but to do work:

Arc: My daily driver for years (I still love Arc). Not AI-native, but a UX powerhouse. Spatial browsing, tab muscle memory, obsessive design. Built by The Browser Company. Cult following. Founder energy. Still iconic.

Dia: The Browser Company’s surprise AI spin-off. Launched into curiosity, and a little confusion. Chat-with-tabs is the flagship feature. Beautiful UI. But for now, it’s more conversation than execution (and I’m excited to see new features like vertical tabs).

Comet: Built by Perplexity. More than chat. Comet is about delegation. It reads your email, clicks your buttons, drafts your reports, and acts on your behalf. It’s not just “AI inside a browser.” It’s a browser that works like an AI agent.

After 30 days running Comet side-by-side with Arc, Dia finally dropped its beta.
A year ago, I’d have killed for a Dia invite. But after living in Comet? It felt… incremental.

That’s when it clicked. This is a shift in how work gets done, and the battleground is the browser.

That's when I knew I had a story.

From a bigger picture these aren’t just new UIs, they’re fighting over a much bigger prize: The interface layer between humans and AI.

I’ve said it before: Websites are becoming tool calls.

And if that’s true, the browser becomes the orchestrator of the AI stack. The place where intent turns into execution.

For years, Arc was home base for my workflow. The design was so good it felt like a productivity cheat code. But it wasn’t AI-native. It couldn’t act on my behalf.

Then Comet landed in my workflow.

I stopped browsing and started delegating.

And in a week where The Browser Company finally launched Dia, I saw just how fast the baseline had moved. The standard wasn’t slick UI anymore, it was agentic power.

Stage 1: Data Collection (a.k.a. feeding the beast)

Most viral content dies in the shallow end. It’s surface-level, regurgitated, and 80% guesswork.

That’s not how I like to operate.
If I’m building in public, I want receipts. I want friction logs. I want context no one else has.

So I stress-tested Comet across 30 days of real workflows.
Inbox triage. Growth reports. Sentiment mining. Personal knowledge recall. Competitive audits.

And I logged everything. With Comet, it’s easy. Just ask Comet to create a note for you (in your web editor of choice).

If you’ve followed me for a while, you know I’m militant about systems and notes. Every moment of friction, every spark of “holy sh*t, that worked,” got timestamped and stored.

Here’s why that matters:

All the top models (GPT-4o, Claude 3, Gemini) are powerful.
But what sets them apart isn’t the model. It’s what you feed them.

That’s why native access to your browser, Drive, Gmail, ClickUp, and notes matters.

In short:
Context ≠ bonus.
Context = advantage.

So before a single sentence of that thread was written, Comet already had my logs, my benchmarks, my past Arc love letters, my email threads, my analytics accounts.

The day Dia launched, I flipped the switch.

I dropped a prompt into Perplexity Labs, the new deep research engine they launched right as I was neck-deep in Comet testing:

If you haven’t used Labs yet, imagine a researcher that scans the web, synthesizes sources, creates visuals, and hands you a publish-ready breakdown. All without bouncing between tabs.

But something to note:
Labs couldn’t yet access to my personal stack (emails, notes, usage logs, analytics). Comet did. So I ran them in parallel:

And whenever I wanted a true insight? I piped my Comet notes into Labs.
Perplexity says Labs + Comet will soon work together, but in Beta, at least for now, they do not.

I dropped the first prompt into Labs for competitive research.

First, Comet launched with 50+ targeted queries from a cross-section of media and voices:

It surfaced sentiment swings, technical breakdowns, community trust metrics.

It stitched these together into a coherent landscape view: what each team was building, where users were excited, and where trust was breaking.

Then came the structured analysis.

Comet auto-mapped Comet, Dia, and Arc across:

Finally, it dropped in the visuals:

I iterated with Labs + Comet, building on the initial Labs report. I asked a series of personal questions like:

Years of my Arc obsession:

Despite having just one day with Dia, my notes appeared (saved in Drive). Comet surfaced what worked: UI polish, intuitive layout, genuine design strengths. Fresh takes from new users across platforms.

I did not want this to flatten into "Comet good, Dia bad." Just perspective to tell the story straight.

And I asked Comet to give me a retelling of how I’ve been using it in different agentic workflows.

This is the easiest way for a busy person wanting to learn AI in as little time as possible:

Sign up to start learning.

A good thread doesn’t just explain.
It tells a visual story.

I told Comet:

While Comet is not able to actually able to take screenshots, it’s able to point me in the right direction, create visuals and graphs, and to identify areas that I can take screenshots of.

What I got back was a collection of images in three general categories:

1. Product Examples
2. Sentiment & Strategy Charts
3. Receipts from My Workflow

Quick note: I did a second visual pass after refining the text thread.

At this point we had enough data and background to create a thread.

Here's how the chain played out:

This wasn’t a one-off experiment. It was a preview. Comet is in beta. Dia is in beta. These are the worst these tools will ever be. And like any beta product, they are not perfect. I could write a whole post on what to improve.

But looking forward, we’re not just getting better toys to play with.
We’re getting a new operating layer where research, context, creation, and distribution all happen in one flow, from one place, powered by agents that actually understand your stack.

For years we’ve been debating which model was smartest. Now that models are commodities the real edge is where that model lives. What it can access. How it executes.

And more often than not, that answer is the browser.

If you’re reading this, you already get it: the operators who master these browser tools and agent workflows will dominate the next cycle.

If you want an AI-powered growth audit tailored to your stack, or just want to trade strategies with someone who’s been living in the trenches, let’s talk.

Book Your AI Growth Audit Here 🔥

Cheers,

Matt Berman

🎁 Get rewarded for sharing! My team grew NEW accounts to over 50 million views in just a few months. We made an AI viral hook generator so you can follow the same hook frameworks that we do.

Invite your friends to join, and when they sign up, you’ll unlock our AI Viral Hook Generator—the ultimate tool for creating scroll-stopping content.

{{rp_personalized_text}}

Copy and paste this link: {{rp_refer_url}}