When intelligence becomes cheap

The more interesting question

There is one mistake we keep making about artificial intelligence. We ask what happens when it becomes smarter. The more interesting question is different: what happens when intelligence becomes cheap?

For most of human history, a competent answer carried a meaningful cost. Legal advice required a lawyer. Software required a programmer. Research required an analyst. Translation, design, and planning required people who had spent years learning their disciplines. The price was not only money. It was time, coordination, and access.

I can open a laptop in Ukraine, describe a problem, and receive in seconds something that once took hours of professional work. The answer may be wrong. It may be average. It may be surprisingly good. For the larger economic shift, perfection is not required. The cost only has to fall far enough that behavior begins to change.

AI is our quartz movement

The watch industry offers a useful analogy. Mechanical watchmakers once competed intensely on accurate timekeeping. Then quartz technology made accurate time cheap. An inexpensive watch could tell time more precisely than a costly mechanical one.

Mechanical watches did not disappear. They changed what they were selling. Accuracy was no longer enough, so value moved toward craftsmanship, history, scarcity, identity, and trust in the maker.

AI is a quartz movement for intellectual work. It does not need to reproduce every quality of an expert human mind. It only needs to make a large amount of competent generation cheap. Once that happens, the value moves somewhere else.

The advantage moves upward

A decade ago, the distance between a technical founder and a non-technical founder was enormous. One could test an idea on a Saturday evening. The other needed to find a developer, explain the idea, negotiate a price, wait for implementation, discover misunderstandings, and begin again.

That difference still matters, but it is changing shape. A founder can now ask AI to propose a database structure, create an interface, explain deployment errors, generate queries, write tests, and help debug the result. What used to be impossible can become merely difficult.

The technical founder does not lose every advantage. The advantage moves upward—from writing the function to deciding whether the function should exist. Which user matters? Which feature solves something real? Which detail only looks impressive in a demo? Should the process be automated, or performed manually twenty times first? What should be ignored?

These questions look less technical. They are harder.

Generation is cheap. Verification is not

AI can produce a detailed security report, a confident market analysis, or thousands of lines of code at almost no marginal cost. The recipient still has to determine whether any of it is true.

This creates verification debt. Information can now be generated faster than responsible people can inspect it. A coding agent may work for an hour and leave behind four thousand new lines. Work appears to have happened. Perhaps it did. Perhaps the team also inherited four thousand lines that someone must understand before they can trust them.

The bottleneck moves from production to selection. AI can generate one hundred business ideas. Which one deserves the next year of your life? It can produce fifty names. Which one is memorable? It can draft twenty landing pages. Which message is actually true?

More output is not more progress

AI feels productive because it responds with visible artifacts. Ask for a strategy and twelve pages appear. Ask for competitors and a table fills itself. Ask for content and a calendar becomes crowded. The abundance is real, but the feeling of movement can be deceptive.

Entrepreneurs already know that activity and progress are different. A company of one hundred people can move more slowly than a company of ten. More meetings do not guarantee clarity. More features do not guarantee value. AI multiplies this old problem because output becomes effortless.

A new kind of company failure is becoming possible: not a company that cannot produce enough, but one that produces so much it can no longer tell what matters.

Stay close to the unscalable work

AI naturally pulls us toward abstraction: automate onboarding, generate the content strategy, build a customer acquisition system, let an agent handle support. Sometimes the intelligent move is to do the apparently unintelligent manual thing.

Call one customer. Watch how they use the product. Answer ten support messages yourself. Build something by hand for one person. Read the strange complaint instead of asking for a summary of five hundred complaints.

The amount of work we can automate is increasing. The amount we should automate does not rise at the same speed. Manual work is often how a founder discovers the shape of the system worth building later.

Answers are abundant. Gaps are not

AI performs best when the question is already clear. Entrepreneurship usually begins earlier, when the real question is still hidden. Customers ask for one thing and pay for another. A crowded market may be full of products that solve the wrong part of the problem. A toy may have a small group of unusually devoted users. A business that looked impossible may become viable when one cost collapses.

Interesting opportunities live in the gap between what everyone assumes and what is actually true. From a distance, a field looks complete. After enough attention, cracks appear: Why is this process still manual? Why do customers export everything to a spreadsheet after buying expensive software? Why are people unhappy in a market everyone calls solved?

AI can help investigate these questions. Someone still has to notice them. Curiosity is not getting cheaper.

Thinking needs deliberate friction

AI can write almost any business text: an email, proposal, job description, investor update, landing page, or article. This is useful. But writing was never only the production of text. Writing forces you to discover what you think.

An idea can feel complete while it remains inside your head. The moment you try to place it into simple sentences, its missing half becomes visible. The blank page is irritating because it exposes confusion. AI can remove that irritation—and therefore remove the exercise.

Machines removed much physical effort from ordinary life, so people who want strong bodies now exercise deliberately. Thinking may follow the same path. We may choose to write before asking for a draft, read the original source instead of only the summary, calculate before opening the calculator, and explain the problem before requesting a solution.

Not because the tool is bad. For the same reason the elevator is not bad, but sometimes you still take the stairs.

Trust becomes more expensive

Polished work once carried a hidden signal: someone had invested time. Perfect grammar, professional design, and a forty-page report were costly enough to suggest commitment. AI weakens that signal because corporate polish can now be generated instantly.

Proof becomes more important. Show the product. Show the numbers. Show customers using it. Show the code working. Explain what you personally observed. Put a real name behind an opinion and make the claim specific enough to check.

This may be good news for entrepreneurs. For years, internet marketing taught small companies to imitate large ones: use polished stock images, say 'we' when the company is one person, and publish generic thought leadership. When anyone can generate corporate polish, being recognizably human becomes differentiation.

Building from Ukraine gives me a different view from someone working in San Francisco, London, or Singapore. That difference is not noise for AI to smooth away. It may be the part worth keeping.

A non-expert with judgment

Cheap intelligence does not only create spam. It gives ambitious people access to capabilities they did not previously possess. A person can enter unfamiliar territory, use machines to acquire temporary competence, recognize where the machines may be wrong, and bring careful evidence to someone with deeper expertise.

This is a powerful model for the future entrepreneur—not a person who knows everything, but one who can navigate what they do not know. The distinguishing skill is not blind confidence in the output. It is expert-level care around the output: honest limits, good questions, verification, and respect for the people who understand more.

Intelligence was never the whole game. Businesses often remain unbuilt because someone avoids calling the customer, delays the launch, prefers impressive features to the one users need, or waits six months after understanding exactly what must happen. AI does not automatically repair courage. It can even become a more sophisticated form of procrastination.

What becomes expensive

Judgment becomes expensive because someone must decide which answer is right. Taste becomes expensive because someone must decide which of a thousand possible outputs deserves to exist. Trust becomes expensive because convincing information can be generated almost for free.

Curiosity becomes expensive because answers are everywhere but good questions are not. Courage stays expensive because a machine cannot take your personal risk. Attention stays expensive because generated information becomes unlimited. Responsibility becomes expensive because a human still has to say: yes, ship this.

AI will make small teams surprisingly capable. Five people may produce what once required fifty. That does not guarantee the five-person company will be good. Each person will make more decisions, and bad decisions will accelerate alongside good ones.

Do not compete with the machine where the machine makes things cheap. Move toward what remains scarce. Understand a particular customer better. Find the strange problem before everyone notices it. Develop taste. Build reputation. Verify things. Take responsibility. Form opinions from experience. Learn to ask better questions. And sometimes, while everyone else automates everything, do something manually.

The next opportunity

The future will contain far more intelligence than the past. Intelligence by itself may become less impressive, just as accurate time became less impressive once everyone could have it.

For entrepreneurs, the most important question is not whether AI will become intelligent enough. It is this: when everyone has access to intelligence, what will you have that is still difficult to get?

That is where I would look for the next opportunity.