Writing
When the Answer Becomes Cheaper Than the Question
A path from a missing prompt detail to the question we stop asking
ReadFrom the printing press to artificial intelligence
What do we stop asking when something answers for us?
GPSGPS — Global Positioning System: satellites that let a phone or a car know where it is and suggest a route turn by turn. gets us there without our learning the city. A grade sums up a year of lessons. A protocolprotocol — A fixed list of steps that says what to do and what to check each time, so the result does not depend on who does it. decides what to check, and now an AIartificial intelligence (AI) — Computer programs that produce text, images, code or answers by learning from huge numbers of examples — ChatGPT, Claude and Gemini are well-known ones. drafts in seconds the report that used to take an afternoon. Most of these shortcuts are worth it. But some quietly take away a question the effort used to force on us: is this true? what is it for? what do I still want to be able to do myself? AI makes these questions urgent because it speeds everything up at once, but it did not invent them: writing, printing and statistics raised them first. Here they are, in plain language, with what people from many eras have thought about them and the source of every quotation. We won't give you the answer — we give you material to think through your own.
What is Unfinished?
It is a library of questions, not answers. Each question has its own page: it explains why the question matters, gathers what different people hold about it — from Socrates to today's AI researchers, with the exact quotation and a link to the original text — links the key words it uses, and suggests what to read next. You don't need to know philosophy or the authors. It is for you if you study, teach, lead a team, work with AI, or simply wonder what is lost, and what is gained, each time something gets easier. It is a work in progress — hence the name — and it grows by adding questions and voices that don't always agree.
What we hold, and what could prove us wrong
A site that asks every idea what would refute it has to answer the same about itself. This is our position on the central question, and the strongest objection we know to it.
Our positionCheap answers do not make questions cheapWhen answers cost almost nothing, the rare thing is no longer the answer. It is the judgment to know which question was worth asking. That judgment has usually grown out of doing the slow work a quick answer now skips: a student who writes an essay herself finds out along the way what she really needs to know, and one who gets it ready-made, from a classmate, a website or an assistant, misses that step.
The objection we take seriouslyNot all lost friction was formativeA lot of what tools now do for us was plain drudgery that taught nobody anything: retyping a page, hunting for a reference, waiting in line. Treating every bit of effort a tool removes as lost judgment makes work that was merely tedious sound noble. It also makes the argument impossible to prove wrong, since any convenience can be counted as a loss.
A few places to start
Four well-developed questions to start with. Each card shows a real position on the question, with its author and its source.
- Knowing · ToolsWhat do we stop asking when something answers for us?The site's central question: when something answers for us — a tool, a number, a rule, an expert — which questions quietly stop being asked?When answers cost almost nothing, the rare thing is no longer the answer. It is the judgment to know which question was worth asking. That judgment has usually grown out of doing the slow work a quick answer now skips: a student who writes an essay herself finds out along the way what she really needs to know, and one who gets it ready-made, from a classmate, a website or an assistant, misses that step. — Editorial position of Unfinished
- ToolsWhat should not be automated?Some tasks are how we learn to judge them. If a person, a routine or an AI always does them for us, we may lose the ability to tell when the result is wrong.Neil Postman argues that a new technology is never simply added to a culture; it changes the whole culture. Each tool leans toward certain uses, gives power to certain people, and brings some problems into view while making others hard even to ask about. A present-day example of ours: with GPS, getting somewhere became easy, and learning the layout of a city began to seem pointless. — Our paraphrase of Neil Postman“A new technology does not add or subtract something. It changes everything.”— Neil Postman, Technopoly, ch. 1, “The Judgment of Thamus”, p. 18 (Vintage paperback, 1993)
- Knowing · ToolsWhat did your request leave out?When you hand a job to someone — a colleague, a contractor or an AI assistant — you can't write down everything you know. What comes back fills in the gaps, but with what?Michael Polanyi argued that much of what skilled people know cannot be put into words. His own example: you can recognize a face among thousands without being able to say how you do it. He called this *tacit knowledge*: know-how we use without being able to spell it out. The gap is not laziness; it is part of how skill works. — Our paraphrase of Michael Polanyi“I shall reconsider human knowledge by starting from the fact that we can know more than we can tell.”— Michael Polanyi, The Tacit Dimension, ch. 1, “Tacit Knowing”, p. 4
- KnowingWhat should count as proof?A chatbot sounds sure, a study says one thing, a friend says another. What does a claim need before we accept it?Eliezer Yudkowsky argues that a belief is worth keeping only if it tells you what to expect, and above all what not to expect. If you believe a bridge is safe, you expect it to hold when you cross. A belief that fits any possible outcome, or that links only to other beliefs and never to something you could notice, has come loose from the world, and you should let it go. — Our paraphrase of Eliezer Yudkowsky“A null answer means that your belief does not constrain experience; it permits anything to happen to you. It floats.”— Eliezer Yudkowsky, Making Beliefs Pay Rent (in Anticipated Experiences), paragraph beginning “It is even better to ask”
What you'll find on each question
Every question follows the same order, so you know where to look.
- 01
The question
Put in one sentence, with the context that makes it matter today.
- 02
Why it matters
What is at stake, and what gets lost if nobody asks it.
- 03
Positions
What different people hold about it, each position with its author and its quotation.
- 04
People
Who each author is, and which field they speak from.
- 05
Sources
The exact book or article behind each quotation, with page and link where one exists.
- 06
Related questions
The ones this question leads to, and the ones that push back on it.
- 07
Paths
Guided sequences that walk through several questions in order, like a class.
The questions
Grouped into five areas. Under each question is one line on what it is about; open it to see the positions, the authors and the sources.
Appearances
How things get shown to us — by institutions, media and numbers — and what the picture leaves out.
- When is a model of reality good enough to act on?Maps, forecasts, summaries and simulations are always partial. What matters is whether they are good enough for the decision at hand.
- Which rules exist because of mistakes we've forgotten?Checklists, sign-offs and regulations often began after a disaster. When nobody remembers why, they look like pure bureaucracy — and get removed.
- What happens when the number becomes the goal?Test scores, likes, sales targets: we measure what we can count. Once people chase the number, it stops telling us what we wanted to know.
- What does technology hide when it stands between us and the world?Feeds, maps and recommendations shape what we see. The smoother they work, the less we notice what they left out.
Knowing
How we find out what is true, and how we notice when we are wrong.
- What do I actually know?Most of what we know we took on trust — from teachers, encyclopedias, search results, an AI or someone who seemed sure. How much of it could we actually check?
- What would make me wrong?A habit of good thinking: before defending an idea, decide what evidence would change your mind.
- What should count as proof?A chatbot sounds sure, a study says one thing, a friend says another. What does a claim need before we accept it?
- What gets lost when a stand-in replaces the real thing?A map instead of the land, a photo instead of the place, a score instead of the student: every stand-in drops something.
- Where does the real world get a say?A process can run for a long time on reports about reports. Where does it actually check against what is happening — a test, a customer, a patient?
- When does spotting a pattern become real understanding?A student can recite the right answer without grasping it, and an AI can produce fluent text. When does that count as understanding?
Meaning
What a life is for, once the tools can do the work.
- What is still worth doing once doing it well is no longer rare?Photography didn't end painting, and calculators didn't end learning to add. Now that AI can draw, write or code competently, why learn or do things yourself?
- Which abilities should remain ours?We stopped memorizing phone numbers without regret. Which skills — writing, calculating, judging — should we keep even if a tool can do them?
- Does work matter for what it produces, or for what it makes of us?If an AI puts the report together in seconds, is anything lost besides the hours? Whether the value of work is the result, or what doing it teaches the person who does it.
- What are we optimizing our lives for?We track steps, followers, grades and productivity. Are those numbers what we actually want our lives to be about?
Skill
Where real skill comes from, and what happens to it when the effort is removed.
- Which difficulties actually teach us?Some effort is pointless drudgery; some is how we learn to spot a bad answer at a glance. How do we tell them apart?
- Where will tomorrow's experts learn their craft?Experts are made by doing routine work, getting it wrong and being corrected. If the routine work goes to AI or is outsourced, how will beginners learn?
- Why are we making this at all?Reports, emails and slides get made because the job says so. Now that AI can make them in seconds, it is worth asking who they are for.
- What do we lose when we hand a task off?Handing work to a colleague, a contractor, an app or an AI saves time. What else goes with it — practice, understanding, responsibility?
- What happens when we can do something before we know whether we should?New tools let beginners do what only experts could. But knowing when — and whether — to do it comes from experience.
Tools
What tools give us, what they take away, and what we stop doing ourselves.
- What do simplifications hide?A spreadsheet, a dashboard or an app hides the messy details so we can work faster. What do we lose sight of, and when does it come back to bite?
- Which old limits held hard-won knowledge?Some rules and limitations exist because someone learned the hard way. Remove them with a new tool, and the lesson can disappear too.
- What happens when answers cost nothing?Finding an answer used to take effort, and the effort taught us things. What changes when a search, a formula or an AI assistant answers instantly?
- What should not be automated?Some tasks are how we learn to judge them. If a person, a routine or an AI always does them for us, we may lose the ability to tell when the result is wrong.
- When does making something more efficient make it worse?Faster meetings, shorter lessons, cheaper care: every optimization aims at one thing and ignores the rest.
Longer reads
Two slower ways through the same material: a piece of writing that follows one line of argument from question to question, and a guided path that walks a set of ideas in order.
Guided path
What the Tool Took
The site's central argument, walked end to end: from the questions a good tool lets you skip, through the objection that most of them were worth skipping.
StartOther ways in
There are two other ways through. The map shows every question and how they connect. The index is a plain list of everything here, filterable by area.