What is tacit knowledge?

Tacit knowledge is what people know but cannot easily put into words: the skill, judgment, and know-how learned by doing rather than by reading. Riding a bike. Reading a room. The sense that a deal has gone quiet in the wrong way.

Its opposite is explicit knowledge: anything written down — a manual, a spec, a wiki page — that transfers to whoever reads it. The philosopher Michael Polanyi coined the term and compressed it to one line: we know more than we can tell.

Explicit knowledge is written down and travels by document: manuals, runbooks, pricing sheets, code comments. Tacit knowledge is held in people and travels by apprenticeship: you watch, you try, you get corrected, and slowly you can do it too.

The test is simple. If a new hire could pick it up from a file, it is explicit. If they have to sit next to someone for six months, it is tacit. Neither kind is better — a pilot needs the checklist and the ten thousand hours. But they fail differently: explicit knowledge survives turnover, tacit knowledge resigns with its owner.

The workaround everyone uses but no runbook mentions. Knowing who to actually ask, whatever the org chart says. The sales lead's feel for a deal going quiet in the wrong way. Why the Friday deploy freeze exists. Which client will accept a rough draft and which one will walk. None of it sits in a system; all of it runs the place.

Because the person who holds it cannot fully articulate it. Polanyi's example was the cyclist: ask them how they balance and they cannot state the physics, yet they do not fall. Skilled performance rests on details the performer attends from, not to; they are not available for dictation.

The incentives make it worse. Documentation asks the busiest expert to spend their scarcest hours serving people who have not asked yet. And even the parts that do get written go stale as the work changes underneath them. So most organizations settle for the honest workaround: sit next to Maria for a while. It works — until Maria leaves.

Michael Polanyi, a chemist turned philosopher, coined it. His 1958 book Personal Knowledge argued that all knowing rests on personal skill and commitment; The Tacit Dimension (1966) delivered the line that stuck.

Business found the idea three decades later. Ikujiro Nonaka and Hirotaka Takeuchi's The Knowledge-Creating Company (1995) put tacit knowledge at the center of how firms innovate, and made converting it into explicit form (they called it externalization) a management discipline. An entire knowledge-management industry followed. The problem it set out to solve is still open.

Not from your documents, because tacit knowledge was never written into them. A large language model is trained on the written record — the explicit layer. That gives it the world's general knowledge and none of your organization's unwritten know-how.

Retrieval does not close the gap. Point an AI at your wiki, your drive, your tickets: it will surface everything anyone ever wrote, and nothing anyone only ever knew. You cannot index what no one recorded.

The gap is easy to miss, because the model is fluent: it sounds like it knows your business. What it actually knows is every business in general. The workaround, the client history, the reason the rule exists: the people who hold them have to put them into words before any system can use them. Once told, they stop being tacit. That telling is the work.

Dark context is the slice of tacit knowledge that matters for AI: what an organization knows but never wrote down — the situated context a general AI model cannot see. Put plainly, dark context is tacit knowledge seen from the AI's blind spot.

Tacit knowledge is the seventy-year-old academic category, and it covers everything from bicycle balance to surgical craft. Dark context names the organizational slice of it that an AI working for you needs and lacks: a general model has all the general knowledge and none of yours, and that gap is your dark context. We — Ola Möller and Andriy Zhukov, at Dark Context — coined the term by analogy to dark matter: the unseen mass that holds the visible structure together.

What is tacit knowledge?

Tacit knowledge is what people know but cannot easily put into words — the skill, judgment, and know-how learned by doing rather than by reading. Michael Polanyi coined the term and compressed it to one line: we know more than we can tell.

What's the difference between tacit and explicit knowledge?

Explicit knowledge is written down and transfers by reading — manuals, runbooks, wikis. Tacit knowledge is held in people and transfers by apprenticeship — watching, doing, being corrected. Explicit knowledge survives turnover; tacit knowledge resigns with its owner.

Can AI learn tacit knowledge?

Not from your documents, because tacit knowledge was never written into them. A language model learns from the explicit record, so it holds the world's general knowledge and none of your organization's unwritten know-how. It can use that know-how only after someone puts it into words.

What are examples of tacit knowledge at work?

The workaround everyone uses but no runbook mentions. Knowing who to actually ask, whatever the org chart says. The feel for a deal going quiet in the wrong way. Why a rule exists at all. None of it sits in a system; all of it runs the place.

How is tacit knowledge related to dark context?

Dark context is the organizational slice of tacit knowledge, seen from the AI's blind spot: what an organization knows but never wrote down — the situated context a general AI model cannot see. The term was coined by Ola Möller and Andriy Zhukov at Dark Context.