In 1945, Vannevar Bush published 'As We May Think' — an essay imagining a machine called the Memex that would store and link all of a person's books, records, and communications, letting them trace associative trails through information the way human memory actually works. The essay became foundational to how we built the internet. But almost no one remembers the part Bush got wrong: he assumed the bottleneck was storage and retrieval. Give people access to everything, he thought, and understanding would follow. It didn't. What Bush missed, and what the philosopher of science Michael Polanyi had already identified in the same decade, is what Polanyi called 'tacit knowledge' — the kind of knowing that lives in practice and cannot be fully articulated, let alone indexed. You cannot retrieve what was never made explicit. Today's AI systems are Memex machines on a staggering scale: extraordinary at retrieving, linking, and surfacing the explicit record of human thought. The gap that remains is exactly the one Bush overlooked — between having access to an insight and being able to use it. The practical implication is structural, not personal: when you design a workflow around AI assistance, the question isn't 'can it find the right information?' It almost certainly can. The question is what you're building into the process that converts retrieved information into something you can actually act on.

In your current work, which step — finding information, or converting it into usable judgment — is the actual constraint?

Drawing from Philosophy of Science / Epistemology — Michael Polanyi