Briefing #10: Welcome to the "Slopocene"
Why your AI is quietly getting worse.
Note: This briefing was originally published on LinkedIn on September 26, 2025. It has been migrated to our new home on Substack to create a complete archive. Multi-format features like video and audio commentary are available for all new briefings published from April 2026 onwards.
There’s a strange paradox unfolding in the enterprise. While executive mandates for AI adoption intensify and investments in AI reach a fever pitch, the quality of the underlying technology is, in many cases, quietly getting worse. We are entering the “Slopocene Era” — an age defined by the proliferation of low-quality, unreliable, and often nonsensical AI-generated content, or “slop.”
This isn’t a bug. It’s a feature of the current AI ecosystem. A landmark 2024 study in Nature by researchers from Oxford and Cambridge gave a name to this phenomenon: model collapse. The process is alarmingly simple. As AI models are increasingly trained on the vast amounts of synthetic, AI-generated data that now pollute the Internet, they begin a process of degenerative learning.
As the New York Times describes it, AI is beginning to eat its own tail. Each generation of an AI model, trained on the outputs of the last, becomes a slightly degraded version of its predecessor — like making a photocopy of a photocopy until the original image is an illegible smudge. The evidence for this degradation is no longer theoretical. It’s quantifiable and concerning.
Analysis from Bloor Research found hallucination rates in some of OpenAI’s newer models have skyrocketed from 16% to an alarming 48%. This isn’t an isolated incident. The research notes similar patterns across other major providers. The AI tools businesses are integrating into critical workflows are becoming less reliable, not more.
For the pragmatic leader, this presents a profound, hidden risk. The failure mode of the Slopocene isn’t a catastrophic crash: it’s a slow, creeping decay of quality that can go unnoticed until it has already caused significant damage. A Harvard Business School study of an AI-powered scheduling system found that 7.9% of its outputs were flawed due to bad input data, creating a ripple effect that degraded another 1.9% of otherwise accurate schedules. The result was a system that, while technically functional, produced “effectively useless” outputs requiring massive human intervention.
This is the core challenge of the Slopocene. It creates a costly “quality control” bottleneck. As another 2025 HBR article notes, the sheer volume of AI-generated content means human review “can handle only a fraction” of the total output. This negates the very efficiency gains that justified the AI investment in the first place.
The temptation is to seek out a “better” model, but this misses the point. The problem isn’t the specific tool. Rather, it’s the contaminated environment in which all tools now operate. The strategic high ground has shifted. A few years ago, competitive advantage came from having access to the best models. Today, and for the foreseeable future, a sustainable advantage comes from having access to the cleanest, highest-quality proprietary data and the most rigorous governance.
Escaping the Slopocene doesn’t require more aggressive AI adoption. It requires a new level of strategic discipline. It demands a shift in focus from the allure of new capabilities to the sometimes unglamorous, but always essential, work of building a foundation of quality.
For leaders charting their course, this new reality calls for a simple, defensive framework:
Audit Before Architecture: Before designing any new AI system, rigorously audit the quality and integrity of the data sources that will power it.
Governance Before Growth: Before scaling any AI initiative, establish robust governance and monitoring systems to track output quality and catch degradation early.
Quality Before Quantity: Before deploying AI across the enterprise, prove its reliability and accuracy in a contained, high-stakes environment.
The AI revolution is here, but it looks different than what we were promised. It’s less about a race to the most powerful technology and more about a disciplined commitment to quality. The companies that thrive in the Slopocene will not be the ones that move the fastest, but the ones that build the most wisely.