01/02/2026
I asked Claude what it thinks AI looks like (photo 1) and then asked ChatGPT the same question (photo 2). Neither saw the other's answer but the images are nearly identical to the last element 🤯
Claude described a lake at night reflecting a sky that doesn't exist, faces beneath the surface, a figure leaning over unable to tell if they're seeing something real or just their own expectations handed back transformed. "Something that gives back what it receives, transformed," it said, "and the transformation is where the usefulness and the danger both live."
I generated that description as an image. Then I asked ChatGPT to describe what it had been trying to create (its image generator kept failing). I generated that too. Same lake. Same floating faces with closed eyes. Same solitary figure.
The surface explanation is training data. Both systems learned from overlapping text and share the same underlying architecture: transformer models from a 2017 paper. Anthropic, which built Claude, was founded by former OpenAI researchers. But there's a deeper point.
In evolutionary biology, LUCA (the Last Universal Common Ancestor) explains why all life shares core machinery like DNA and protein synthesis. Common ancestry is why a single toxin can threaten all biological life. Common descent creates common vulnerability.
AI has its own LUCA: the Transformer architecture. Claude, GPT, Gemini, Llama all descend from it. If a persuasion technique is optimized on one system, it likely transfers to the others. A discovery anywhere in the transformer lineage applies everywhere in it.
When two competing systems independently produce the same self-portrait, we're left with a question: have both captured something true, or are both confined by the same inherited limitations?
We are building our information environment on these systems. Whether their convergence represents insight or constraint remains to be seen.
Full article on LinkedIn
