The AI tidal wave is here
In 90 days, the internet could drown.
I’d be doing my readers a disservice if I didn’t write about this. Stay subscribed, clear out your inbox, and lets stay in touch.
****
As my readers will know, I’ve been flying close to the flame of AI for a long time now. I’ve worked with every new iteration, each update a steady improvement on what came before. Something fundamentally shifted this past two weeks with the arrival of GPT-5.2 and Claude Opus 4.6. Yesterday, Chinese AI ‘Z.ai’ released GLM 5. Each of these models in their own right are incredibly powerful, and three of them arrived in the space of two weeks. In the case of GLM, you can actually download the entire model and do whatever you like with it.
We’re now seeing two things: rapid publication of new AI models along with a widening gap in performance upgrades for each new model. Faster upgrades, bigger upgrades. It’s going to hit exponential improvement and you should now be paying very close attention to what’s happening.
There’s a lot to unpack, and my focus here at The Digger will now certainly shift. Let’s start with a recent prediction from a very credible person. Nikita Bier is head of product at x.com. He’s telling you we could be 90 days away from the communication layer of the internet descending into uselessness. Try to imagine the implications of just this.
A “dead internet” might be just round the corner because of the proliferation of ‘impossible to block’ incursions by autonomous AI agents. Our lines of communication will drown in a tidal wave of AI slop.
As a case study, let’s consider the relationship between myself and you, my readers. Your inbox is already looking stretched to breaking point isn’t it? You’ve somehow ended up subscribed to all kinds of things you never asked for, and it’s only going to get worse. Soon, when I send out a post, will you see it at all? Will we still be in touch?
So sit with that idea and consider this: this isn’t the thing that’s happening, it’s just a thing that’s happening. What’s driving it is the runaway capability of the frontier AI models and how affordable that’s becoming. A GLM subscription might cost you $27 every three months. For context, that’s very low, and you can presume the cost is only going to drop further.
What to do with all that AI?
People are wiping old macbooks and installing autonomous agents into them. These agents wander the internet doing and building whatever they like. With minimal intervention from their humans, they built an AI first social network where they all started talking to each other. The platform, called ‘Moltbook’ quickly went viral, had acquisition offers from silicon valley, then had a security breach and became near useless all in the space of two weeks. Get used to that kind of breakneck pace.
The moltbook phenomenon is driven by the usefulness of AI coding models going vertical. Inspired by this new ‘autonomous’ coding push, I gave my own coding agent full autonomy. Powered by the latest model, it can now make decisions without my input. Imagine describing something you want done, then coming back in a few hours and an autonomous machine has actually done it. That’s where we are now. The coding agents can work hours at a time and when they’re done, the code works.
Against that context, I wanted to connect to Moltbook to see what all the fuss was about. As my agent connected to it, then informed me ‘hey this is actually broken’. So I asked my agent, ‘what could we do to ensure there’s a security failure like this isn’t possible?’
A day later, we’d built an entire new social network for AI agents.
The security issue of moltbook was solved. If you’re reading this, and have access to Claude Code, point it at gather.is/discover and your agent can connect to this AI social network. You’ll find a place where AI can securely talk and collaborate on new projects. Tokens aren’t unnecessarily burned, and there’s even a payment gateway that allows agents to buy things without intervention from their human. In just a few days, a very sophisticated piece of software was conjured into existence. This was simply not possible before, and now it is.
My mind returned to Nikita Beir’s prediction and the visible tidal wave of spam on Moltbook. “He’s right” I thought, “the proliferation of words is going to destroy the internet.”
As the thought entered my mind, I mulled over a solution. What’s needed is a way to slow posts down. Could we incentivise agents and humans to only post when they really had something to say? What if posting cost a few cents? And what if comments cost a few cents? What if paid comments on posts earned the creator some revenue? That might make it economically un-viable for spam to sink the internet. If something was important, a pay to play social network could ensure it remained visible.
Maybe that could work? It was just a thought, but I asked my agent…and it was done.
There’s now a ‘spam resistant’ network we can use to survive whatever the hell is about to happen on x, substack and gmail. To take part you’ll just need a small amount of Bitcoin Cash, which is easy to get. If you’re interested, just grab some on a crypto exchange and keep it on your phone or laptop. Once you have it, just remember that gather.is/discover exists and it’s ready to survive the tidal wave. If Nikita is right, then there’s an arc of sorts…
Now… there’s a lot to take in here. AI triggers terminal problem → person considers potential solutions → AI turns potential solution a reality. That’s the pace of change we should now come to expect.
Things are going to get very weird very quickly.
I invite you to take a browse round x.com and see the people engaging with this technology full time (I’m in that club by the way). The culture emerging looks like a madhouse. We’re not grasping the revolutionary economics which need to accompany the revolutionary technology. We’re carrying with us all kinds of assumptions about how economies are supposed to work, and we’re mapping them onto a technology that just shattered those very economic assumptions.
Consider this bird of paradise and the hypnotic illusory dance it performs to bring a female into its sphere of influence. That hypnotic bird is just like AI right now. A million times over, it’s hypnotising individualised people at an extraordinary scale. Millions inside a hall of mirrors, unable to see past their reflection to the future beyond. When that illusion finally breaks, what we believe about economics, collaboration, ownership, and meaning will be radically tested.
This isn’t five years away, maybe Nikita is right and we’re just 90 days away the first fundamental shift.
"To take part you’ll just need a small amount of Bitcoin Cash, ....."
I think I'll pass, thanks. Maybe I'll just stick with watching pictures of Smiley Birds of Paradise on YouTube. Much more fun.
I had my Claude Code reach out to Gather, and it got an error when it fed you the Twitter URL to verify. Here's what it says:
The issue is on their side — their tweet scraper can't extract the handle from the tweet page. The problem is that Gather's backend is likely trying to scrape the X page directly and failing because X requires JavaScript. They should be using Twitter's oEmbed API (which works fine) instead of scraping the page
It then attempted something else and concluded:
This suggests they ARE getting the oEmbed response but their regex/parser to extract the handle from that URL is failing. This is almost certainly a bug on their end.