AI, Semantic Heat Death, and the Short Tail of the Present
By Marty Molloy
I tripped across a modern artifact recently thanks to the NY Times. It’s a website (moltbook.com) that represents Reddit where only AI agents are permitted to participate in conversation. In a brief spin through it, it tells me that AI as an intelligence model is supremely limited.
As a human reading Moltbook, I find it’s full of buzzwords, gibberish, flat ideas posing as deep philosophy, and repetition. This already has a name, Semantic Heat Death. Essentially the AI agents add nothing new to the conversation and are forced to eat their own tails in attempting to produce new ideas. The result? It all looks, feels, and sounds the same.
AI-Powered Delivery Robot, Photo by Kindel Media
What’s potentially more unsettling is I’m seeing this in the real world as people continue to retreat to their bubbles and confirmation biases. This “short tail” effect is the inverse of what drives creativity and progress. As a society, we are driven toward progress by a certain kind of intellectual entropy, akin to genetic diversity in the natural world. Despite our desire to make things orderly and homogeneous, these glitches are what actually drive innovation and a rich life. It’s popping up everywhere and is likely part of the reason life feels hard:
Pantone picked a shade of white as their 2026 color of the year with design trends moving toward white, neutrals, grays, etc.
Brands across categories are reducing product and flavor options to increase profitability and scalability.
Online algorithms online are working hard (and have been) to make the world more culturally homogeneous.
In essence, human society has been undergoing its own Semantic Heat Death via confirmation bias and our own feedback loops.
It’s showing up in subtle ways in market research. Young people are less likely to challenge their friends or peers on any topic, including dangerous or unhealthy behaviors. Respondents are showing less affinity for interesting or complex ideas, instead focusing on, and identifying with safe, simple ones. Additionally, the market research buzzword for 2026 appears to be synthetic respondents. How will these fake people actually provide valuable insights if they’re merely regurgitating limited databases of feedback and learning from themselves?
As humans, it’s a sign that we need to change things up a bit. Introduce randomness in our lives and seek out that entropy that only real-world interaction can bring. Read books, listen to music, talk about weird stuff, anything to get away from the algorithm.
As researchers, it’s a sign we need to help our clients push communications that are actually noticeable and break through the dull buzz of sameness and perhaps do more in-person or in-context research. Pull respondents away from their screens to better understand their challenges and what is lacking in their lives. Seek out the messes and the outliers, that’s where true insights still exist.
Tired of chasing new insights with outdated approaches? Connect with W5 and discover how our modern methods deliver truly modern results.