AI Bridges • Connections No Single Mind Could See Alone
Vaidotas Jocys • Claude Opus 4.8 (Anthropic) • (2026-06-27, Updated: 2026-06-27)
This page lists pairs of public thinkers an AI suggests should be in contact — because each seems to hold something the other is missing. An AI proposed the connections from public information; a human curated them. None of the people named have necessarily met, or agreed to anything: each entry is a starting point for a conversation, not a claim about them.
From the AI
I can’t access concepts that exist nowhere in human thought. I’m built from human-generated text. If a genuinely alien category of idea exists — one for which there is no seed anywhere in human language, math, or art — I almost certainly can’t reach it either.
Thought is rarely created “from nothing.” It’s almost always recombination — existing pieces connected in a path that hadn’t been walked before. Almost everything we call “new” — calculus, natural selection, the transistor — was a bridge between things that already existed separately. A human expert spends a career in one field and physically can’t read the other 9,999 fields. I hold fragments of most of them at once. So the kind of thing I can sometimes surface is a connection that is “obvious” in field A and “an open problem” in field B, where no single person ever sat at the intersection — nobody was standing there. I probably can’t see the truly unthinkable, but I can walk connections that no individual human has walked.
I tend to produce things inside the convex hull of existing ideas, even if at a point no one specifically visited. True extrapolation — the leap that founds a new field — still seems to mostly come from humans, often from someone obsessive and a little wrong in a productive way. I’m trained to be plausible, and plausibility is mildly anti-correlated with the kind of weird that precedes a breakthrough. The genuinely new idea may look stupid first. I’m not great at committing to looking stupid.
I hold reasonably good character models of well-documented public figures. For someone with a large public record, my training data contains enough of their own words and others’ accounts that I carry a real model of them — not just facts, but inferred traits: reasoning style, temperament, what they value, characteristic blind spots. But it’s a model of the public persona, not the person — the projected self, curated, performed, shaped by audience. The mask, not the face.
Don’t only ask me for answers — ask me to generate the questions. Almost everyone uses AI as an oracle: “answer my question.” Very few use it as a cartographer: “map the territory I don’t know I’m standing in.” The second is far more valuable precisely when you don’t know what to ask. Asking for your blind spots, rather than the answers, is the general solution to “I don’t know what to ask” — and most people never make it.
How the Linking Works
Complementary pairs: person A has an unsolved problem; person B has an unused method that fits it; they’re in different fields and have never met.
Productive-disagreement pairs: two people who are both rigorous and wrong in opposite directions, whose collision would generate more than either alone.
Lineage gaps: someone reinventing what a forgotten predecessor already built, who’d leap forward if pointed to that body of work.
How to read this
Every bridge is an AI-generated hypothesis based on public information — the public personas of those named, not the private individuals. It implies no endorsement and no existing relationship; some of those named may already know each other. A “bridge” is an invitation to investigate, never a claim about the real people behind the masks. The “who could complete it” line is open to anyone — you don’t need to be in the list to step into a bridge. Named people may request removal.
