This post is the result of my back and forth interaction with Claude Sonnet 4.6 that has yielded the following summary of Venkatesh Rao's recent essay, "Getting Gooier", followed by some perspectives that Rao does not address on Friston's active inference framework and the physiology of agency:
Venkatesh Rao's recent essay "Getting Gooier" makes a point worth sitting with: most AI commentary obsesses over how the world will change, while quietly assuming that humans stay essentially the same — just reshuffled among familiar roles (generalists thrive! storytellers inherit the earth! software engineers disappear!). Rao argues this is the wrong ontology. The real question is how AI is changing human nature itself.
His framework draws on Alan Watts's distinction between prickly and gooey people. Prickly people are tough-minded, precise, boundary-drawing; gooey people are tender-minded, synthesizing, prone to letting distinctions dissolve. All of us carry both, in varying ratios.
Rao's hypothesis: because AI feels like a psychologically safe counter party, we are more willing to expose our gooey side to it and suppress our pricklier instincts. Sustained AI use amplifies the gooey side. We become gooier. He notes that people who insist on prickly, suspicious, line-by-line relationships with AI agents tend not to use them effectively and retreat. People who can vibe with the machine — never even opening the code editor, just watching the agentic shell fly — will likely thrive. The first major gooified interface is, after all, called vibecoding.
The flip side is subtler: as more gooey relational needs are met by AI, the human-facing side grows less inclined to take the emotional risks required to balance prickles and goo with other people. We don't necessarily get pricklier toward humans — we just disengage unless the expected rewards are significantly higher. Your machine-face gets gooier; you look relatively pricklier to other humans.
In the medium term, Rao predicts greater atomization — or "molecularization" — as people grow more distant from other humans while becoming more intimately entangled with their AIs. He sees this as more sustainable than social-media-driven atomization, though the mechanisms for eventual re-convergence into new digitally mediated social forms aren't yet visible.
Following Virginia Woolf's famous claim that "on or about December 1910, human character changed," Rao proposes that on or about December 2025, human nature changed again — with the Claude Code moment being more definitive than the ChatGPT moment, because agentic coding creates a fundamentally alien way of being, open-ended enough to make us as alien as we dare to become. It is, he says, a portal to transhumanism.
Rao's prickly/gooey axis maps naturally onto Karl Friston's active inference framework, and the connection carries some weight. In that account, what we experience as a "self" is a precision-weighted predictive model — a system that assigns confidence to its own predictions and acts to fulfill them. Getting gooier, in Fristonian terms, looks like a reduction in precision-weighting of prior beliefs: a loosening of the grip that constitutes a rigid, bounded selfhood. That loosening isn't necessarily pathological. It can be a precondition for learning, creativity, and genuine engagement with novelty.
But there's a physiological dimension Rao doesn't address, and it connects to questions I've been exploring in recent MindBlog posts and in an ongoing correspondence with my European colleague Heribert on the physiology of agency in the age of AI. Daniel Wegner's work frames the feeling of agency — the sense of being the author of one's own actions — not as a metaphysical claim about free will, but as an evolved emotion, as functionally real as fear or grief. Martin Seligman's learned helplessness research gives that claim empirical teeth: sustained loss of felt agency doesn't merely feel bad, it triggers measurable physiological debilitation. The autonomic nervous system, it turns out, needs the experience of authorship to maintain homeostatic equilibrium.
This raises a question Rao's framing implicitly sets aside: is getting gooier — letting the AI absorb more of the regulatory and decisional load, softening the boundary of the agentic self — compatible with the physiological requirements of human wellbeing? Or does sustained gooification, however pleasurable and productive in the short run, erode the felt sense of agency that keeps the organism in regulatory balance? Heribert has described this as "externalization of self-regulation" — not merely cognitive offloading, but emotional and autonomic offloading. That's a different and more serious claim.
The honest answer is probably: it depends on the mode of engagement. Gooiness that arises from confident delegation — the experienced surgeon handing off a routine task — is physiologically distinct from gooiness that arises from passivity, confusion, or a sense that the machine is simply running past you. Rao's vibecoder who chooses not to open the code editor is in a different physiological position than the one who is simply lost. The challenge ahead is cultivating the former while recognizing that the latter may be far more common than the early-adopter narrative suggests.
No comments:
Post a Comment