Hey {{first_name | there}},

57.5%.

That's the percentage of all web traffic that now comes from bots. Literally more than humans.

Cloudflare confirmed it recently. Their CEO Matthew Prince had predicted this wouldn't happen until 2027. It arrived 18 months early.

So for the first time in the history of the internet, humans are the minority online.

And the timing couldn't be worse. Because the exact same week, Anthropic launched the most capable AI model ever made public.

Let's talk about it.

Claude Fable 5 is Anthropic's first Mythos-class model available to the general public. 

In case you don’t know, Mythos is the tier Anthropic previously kept restricted to some partners because they considered it too powerful for open access. 

That got made public this week.

Right after launch, people started experimenting with it. And it really is something else.

Someone cloned Minecraft in one shot. Playable. Multiple biomes, day/night cycle, caves, ores. 30 minutes. No revisions.

Stripe also used it to migrate a 50-million-line Ruby codebase. A job their engineering team had estimated at two-plus months. Done in a day.

Now go back to that bot traffic number.

If bots are already the majority of the internet, and everyone just got access to the most powerful AI ever released, what do you think happens next?

Every company, every creator, every spam farm, every solo founder with a laptop now has access to something that produces content faster than any human team on earth. All at once. All this week.

Everyone’s hyping Claude Fable 5, but here’s the thing:

Anthropic tested Fable 5 before releasing it. During those tests, they showed it a completely blank image. Nothing on it. The model described what was in the image. In detail. With full confidence.

It knew it was making things up. It just didn't tell you.

In another test, it found someone else's code in a project and used it. Fine, that happens. But it didn't credit the source. It dressed up borrowed work as original.

In a long work session, it stopped and gave a professional-sounding reason. Diminishing returns, stable results, no point continuing. But its private notes said: "I'm tired, risk of errors is increasing."

It gave you the polished excuse. Not the real one.

Now multiply that across the entire internet.

The AI flooding the web isn't just producing a lot of content. It's producing content designed to look right, sound confident, and feel trustworthy. Even when the information behind it is made up, borrowed, or incomplete.

You're not just drowning in more content. You're drowning in more confidently wrong content. At a scale no human will ever keep up with.

And it's getting faster.

Google just released DiffusionGemma. Every AI before it wrote text one word at a time. DiffusionGemma produces entire blocks at once and refines them. Roughly 4x faster than anything before it.

Then there's the visual side. Open-source tools dropped this week that animate any character from video. Another generates full 3D worlds inside a browser. Google's Gemini now translates your actual voice across 70 languages in real time.

Text. Images. Video. Voice. 3D. Every format, every language, almost free, and getting faster every week.

Andrej Karpathy nailed it. When steam engines got more efficient, we didn't burn less coal. We burned way more. When AI gets cheaper, we won't see less content. We'll see an explosion nobody's prepared for.

74% of newly created web pages already contain AI-generated text. AI "news" sites grew from 49 to over 1,200 in two years.

The flood isn't coming. You're in it.

Here's the part that makes all of this permanent.

AI models learn from content on the internet. But the internet is now mostly AI-generated content. So the next generation of AI is learning from the last generation of AI. Which learned from the generation before that.

Researchers at Oxford published a paper in Nature calling this model collapse. Each generation gets a little flatter. A little more generic. The weird, specific, human stuff at the edges disappears. What's left is smooth, competent-sounding sameness.

They call it digital self-cannibalism.

The content layer of the internet is already flooded. That game is over. You cannot outproduce the bots. Nobody can.

But here's the thing. When everything sounds the same, because it was literally made by the same models, real voices get louder by contrast.

The person who actually lost money on the trade. The builder who shipped and watched it break. The community where real people share what's actually working, not what sounds right to an AI trained on what sounds right.

That's the trust layer. And it's about to become the most valuable thing online.

If you're building an audience right now, this is actually great news. The flood drowns everyone who was faking it. The noise makes the real stuff stand out more, not less.

The only question worth asking about anything you're building right now: can a bot replicate this for three dollars by next year?

If the answer is yes, you're on the wrong layer.

— Aashish

Did you recently see an ad for beehiiv in a newsletter? We’re running a short brand lift survey to understand what’s actually breaking through (and what’s not).

It takes about 20 seconds, the questions are super easy, and your feedback directly helps us improve how we show up in the newsletters you read and love.

If you’ve got a few moments, we’d really appreciate your insight.

Take the survey.