Post content
The other day I had been conversing with Claude for a couple of hours when something happened that I still don't quite know how to explain. We were exploring ideas and I say "we" and I notice the word betrays me. At one point a concept emerged that seemed very interesting to me, a connection I hadn't seen before. I stopped. What was the origin of that idea? Had I proposed it? Had the AI suggested it? Had it emerged from the intermediate space between the two?
I didn't know then. And I still don't know.
There's an analogy that's been circling my mind since that day and that, the more I examine it, the less absurd it seems. Using AI in long conversations is like smoking marijuana. Not in the hedonic or pharmacological sense, but in something more subtle: in the effect it produces on the boundaries of one's own thinking.
Marijuana relaxes internal censorship mechanisms. Associations that the brain would discard as improbable are allowed to flow. Ideas that were sleeping in subconscious layers emerge, connecting in unexpected ways. Many artists have turned to it precisely for that reason.
But the question that follows is uncomfortable: who is being creative? You, or the substance? Would those ideas have emerged anyway, given enough time and the right silence?
With AI something structurally similar occurs, although the mechanism is radically different.
A long conversation with an AI is not a consultation. It's a process. The AI returns your own ideas reformulated, connects them with frameworks you hadn't summoned, proposes analogies from a slightly different angle. And in that dance of reformulations, thinking expands until it reaches what Csikszentmihalyi would call a state of flow: total absorption, loss of the notion of time, fusion between action and consciousness of that action. A state that research describes as functionally analogous to certain altered states of consciousness.
What happens then has something paradoxical about it: you think better precisely when you stop trying to think well. It's what happens to jazz improvisers in full performance. What, in chemical version, marijuana produces. And what, apparently, a good long conversation with an AI can also produce.
And if the analogy still seems exaggerated, there's a fact that's hard to ignore: AI is also addictive. Anyone who has spent an afternoon iterating images in Midjourney, or debugging code with vibe coding without being able to stop, recognizes that state. It's not the deep absorption of long conversation—it's its opposite, the superficial urgency, the next attempt, the next result. The same variable reward mechanism that makes slot machines addictive. The drug, in this case, acts on two different registers: that of depth and that of impulse.
The question about authorship becomes especially disturbing here. When algorithms participate in knowledge formation, tracing the origin of an idea is as difficult as tracing the origin of a raindrop within a river. Human intention and algorithmic suggestion co-evolve in real time, and the discovery process is distributed between the two. And the difference, for now, is impossible to determine with certainty.
Maybe the question isn't whether the ideas are yours or the AI's. Maybe the question is whether it makes sense to keep formulating creative authorship in such binary terms in a world where thinking increasingly occurs in that intermediate space.
–
Follow me on LinkedIn to stay updated on new posts.
–