← Library

Why we stopped saying “thinking”

English uses one verb for what a person does and what a language model does. The conflation is not harmless. So we split the word.

Dan DeMers · co-founder, thinqOS · July 2026

People think. Digital minds thinq.

That is the whole proposal. Two words, pronounced identically, spelled one letter apart. Thinking, with a k, keeps meaning what it has always meant: cognition performed by an organic mind. Thinqing, with a q, names cognition performed by a digital one. The rest of this essay is the argument for why anyone should bother.

One verb, two minds

Every day, millions of people watch a status line that says the model is “thinking.” The label is not exactly wrong. Something cognitive is happening: a question is being weighed, alternatives are being considered and discarded, an answer is taking shape. But it is not what your brain does, and everyone saying the word knows it is not, and we all say it anyway because English gives us nothing else.

Borrowed words are how language usually handles new things, and mostly the borrowing is harmless. Ships sail without sails. Phones dial without dials. Nobody is confused. But “thinking” is not like “dialing.” Thinking is the word we use for the most inner thing we do. It comes loaded with everything we know about minds from the inside: consciousness, feeling, intention, stakes, experience. When we hand that word to software, the cargo comes with it.

The two errors

The conflation fails in both directions at once.

In one direction, it inflates. Say a model is “thinking” and you have quietly implied a thinker: something with a point of view, something that understands, something that means it. None of that has been established, and most of it is very likely false for the systems we have. Anthropomorphism is not a fringe mistake made by naive users. It is the default failure mode of using human words for machine processes, and it shapes real decisions. People extend trust to a thing they believe thinks. They defer to it, confide in it, and blame it as if it chose.

In the other direction, it dismisses. When a digital mind does something genuinely cognitive, forming a belief from evidence, revising it when the evidence changes, recalling it later with its source attached, the deflationary reflex calls that “just processing” or “just autocomplete.” That understates what happened in a way that also has costs. If it is all just processing, there is nothing to inspect, nothing to audit, nothing to hold to account. The dismissal is as lazy as the inflation, and it produces the opposite blindness.

One word, two errors. We either grant the machine an inner life it has not earned, or we deny the process a structure it demonstrably has.

Why this is getting worse

For a chatbot answering trivia, the sloppiness was tolerable. It stops being tolerable when digital minds act on our behalf.

Agents now read our mail, move our money, write our code, and negotiate with each other. Each of those actions flows from something process-shaped: context was assembled, beliefs were weighed, a course was selected. When that process misfires and the action goes wrong, the first question is the accountability question: what did it believe, where did that belief come from, and why did it act on it?

Notice that the question is unanswerable in the language of “thinking.” Human thinking is private. You cannot open someone's head and read the belief that made them act; you can only ask them and trust the reconstruction. So when we say the agent was “thinking,” we import that opacity too. The word teaches us to treat the process as a black box, to accept “the AI thought it was fine” the way we accept it from a person, as the end of inquiry rather than the beginning.

But a digital mind's cognition is not private. That is the single most important difference, and the word erases it. What a digital mind believes can be stored. Where the belief came from can be recorded. When it changed, and why, and what it displaced can all be kept. The process that led to an action can be replayed. None of this is true of thinking. All of it can be true of thinqing.

Precision is a safety property

This is why the distinction deserves its own word rather than a paragraph of caveats. Vocabulary is compressed policy. The words a field settles on determine what its practitioners notice, demand, and build.

If we keep saying “thinking,” we will keep building systems whose cognition is as opaque as the word implies, because nothing in the language demands otherwise. If the verb for machine cognition instead names an inspectable process, the expectation travels with the word. To say a system thinqs is to invite the follow-up: show me. And that test is definitional gatekeeping, drawn on purpose. We are defining thinqing as cognition that can produce its own trace: a system that can show its beliefs, their sources, and their revisions is thinqing, and a system that cannot is merely running, however fluent its output. The line is stipulated, not discovered, and stipulating it is the point.

Concretely, here is the trace the word demands. Ask an assistant why it booked you the 9 a.m. flight instead of the 7 a.m. A thinqing system answers with its working: it holds the belief that you prefer to land before lunch, formed in March from something you said, held with high confidence. It revised the belief that early departures are fine last week, when you complained about one. It weighed both against the meeting time and chose 9 a.m. Every line of that answer is a stored object with a source and a timestamp, not a rationalization composed after the question arrived. That trace, and not the fluency of the reply, is what the q claims.

The claim is deliberately narrow. Thinqing does not assert consciousness, understanding, or experience, and it does not deny them. It brackets the metaphysics, which is exactly what a working vocabulary should do. Whether machines can ever think, in the full human sense, is a question philosophy gets to keep. What we need today is a word for what the machines demonstrably do: form, weigh, revise, and recall beliefs, as a process that can be opened and examined.

A distinction that lives in writing

The two words are pronounced the same: thinqing sounds exactly like thinking. The q is written, not spoken.

Be clear about what that costs, because it is a real cost and not a hidden virtue. In speech the distinction does not exist at all: a podcast cannot carry it, a meeting loses it, and a transcription tool will silently turn thinq back into think. Autocorrect will fight the spelling everywhere it appears. A word you cannot hear spreads more slowly, full stop. We accepted that price deliberately rather than minting a new sound, for one reason: writing is where the distinction can do its work. A digital mind's cognition is itself a written thing, stored, structured, auditable, replayable. In writing, where inspection is possible, the q tells you which kind of mind is at work and reminds you that you can go look. In speech you lose the marker and gain nothing back, but you also lose nothing that was available there anyway; the inspection never lived in sound.

The obvious objection

Yes, we named it after our product. thinqOS has carried the q since the beginning, and we are not going to pretend the coinage arrived from nowhere. Every word arrives from somewhere. “Vibe coding” arrived in a tweet. “Enshittification” arrived in a blog post with a book to sell. The origin does not settle whether the word earns its keep; the gap it fills does.

Our claim is that the gap is real: English genuinely lacks a verb for machine cognition that neither inflates it into personhood nor deflates it into noise. If you agree the gap is real, the word is free. Use it without us. Use it against us; hold our own product to the standard the word implies. A coinage only becomes language when it escapes its coiner, and we would rather lose custody of the word than keep custody of the confusion.

How to use it

Write thinq when the cognition belongs to a digital mind: an agent weighing your request, a model revising a belief against new evidence. Write think when it belongs to a person. When both are at work on the same problem, use both, and notice that the sentence gets clearer rather than clumsier: you think, it thinqs, and the collaboration is better for the difference being visible.

The dictionary entry, with the full definition, etymology, and usage notes, lives at thinqos.com/thinqing.

People think. Digital minds thinq. The difference matters, and now it has a name.

---

Dan DeMers is a co-founder of thinqOS, the cognitive layer for AI. thinqOS is a product of AI4Outcomes.