We are currently living in the golden age of Artificial Intelligence. Enjoy it while it lasts, because we are rapidly approaching a cliff. I call it “Peak LLM.”
Right now, Large Language Models (LLMs) like ChatGPT, Claude, and Gemini seem like magic. They can write code, draft legal briefs, and summarize complex market trends in seconds. But beneath the hood, the magic is largely an illusion. At their core, LLMs are not independent thinkers; they are highly sophisticated information retrieval technologies with a basic layer of reasoning tacked on. They do not create knowledge ex nihilo. They synthesize what they have been fed.
And what they have been fed is the greatest resource in human history: the pre-AI internet.
Today’s models were trained on a digital Eden. They ingested decades of human-written articles, soulful artistic perspectives, deeply researched reports, and unpolluted forums. This golden era of the internet was funded by older revenue models—subscriptions, healthy advertising ecosystems, and commissioned art.
But what happens when the machines designed to retrieve this information bankrupt the people creating it?
We don’t have to guess. We’ve seen this exact playbook before. Twenty years ago, news publishers produced high-quality, vetted content. Then, search engines and social media platforms arrived, acting as the ultimate intermediaries. They disrupted the core revenue streams of publishers, hoarding the ad dollars while using the publishers’ content to keep users scrolling.
The result was entirely predictable. Newsrooms downsized immensely. The volume of high-quality, deeply researched journalism plummeted. In its place, the platforms were flooded with low-quality blogs, clickbait, and poorly researched propaganda. Ultimately, the tech platforms degraded the very ecosystem they relied upon to be useful.
We are watching the exact same self-destructive cycle play out today with Generative AI, but on a much grander scale. Because LLMs are so adept at summarizing and mimicking human expertise, they are actively dismantling the business models of the experts they learned from.
You don’t have to look far to see the damage in real-time. Take the world of specialized corporate research. Historically, human analysts at advisory giants like Gartner spent months interviewing executives and crunching data to produce high-quality insights. LLMs read those insights, learned from them, and now substitute that expertise for end-users at a fraction of the cost.
The market has noticed. In May 2025, Gartner’s stock was trading at roughly $450 a share. Today, it is hovering around $150—a brutal 65% drop in less than a year, accompanied by slashed revenue projections and waves of layoffs. As these traditional advisory models collapse and firms downsize, the stream of original, human-vetted market research will dry up.
The same applies to the creative class. If businesses stop hiring original photographers, illustrators, and painters in favor of cheap AI-generated images, the pipeline of new artistic perspectives and aesthetic innovation will flatline.
This brings us to the fatal flaw of the AI revolution. As original human analysts and creators are driven out of business by AI substitutes, the fresh data going into the internet will decline sharply in both quality and diversity.
Instead of reading soulful, intelligent, and original human output, the web crawlers of tomorrow will scrape the internet and find... other AI-generated content. We will enter a feedback loop of synthetic data. LLMs will train on the outputs of other LLMs. Like a photocopy of a photocopy, the fidelity of the information will inevitably degrade. The internet will become an echo chamber of recycled, synthetic sludge, devoid of the very human ingenuity that made AI useful in the first place.
This is Peak LLM. The models are currently performing at their absolute best because they were trained on a snapshot of the internet that was relatively unpolluted by AI. As the inputs into the internet become increasingly artificial, less original, and the older revenue models supporting human creators collapse, the quality of what LLMs retrieve for us will plummet.
The tech industry is operating under the assumption that AI will simply continue to get better, scaling infinitely toward Artificial General Intelligence. But they are ignoring the economics of the data pipeline. You cannot build infinite intelligence while degrading the core input that your technology feeds on.
In a profound sense, the ability of LLMs to replace human effort is self-limiting. If we want AI to remain a useful tool for humanity, we must figure out how to preserve and compensate the human creators, analysts, and thinkers who provide its fuel. Because if the machines eat all the humans, eventually, they will starve.
