Beyond Jevons Paradox

Jevons Paradox is one of the most cited economic principles in discussions about artificial intelligence.

It states that when a resource becomes cheaper or more efficient, we often end up using more of it rather than less. The classic example is coal. As steam engines became more efficient, coal consumption did not decline; it increased. Lower costs encouraged more applications, which ultimately drove higher total demand.

Many people apply the same logic to AI. As artificial intelligence becomes cheaper, faster, and more capable, we will use more of it. That is almost certainly true, but it is only the first-order effect.

Jevons Paradox explains why cheaper AI will lead to more AI usage. It does not fully explain how AI changes which ideas people consider worth pursuing in the first place.

The deeper transformation is not simply that we can do more. It is that more ideas survive long enough to be attempted.

Our brains are resource allocation machines

The human brain is not just a thinking machine. It is also a resource allocation machine.

Every day, it decides where to spend limited resources such as time, energy, attention, money, knowledge, and effort. Because those resources are finite, the brain cannot seriously evaluate every possible action available to us.

The world presents far more opportunities than any person could ever pursue. At any given moment, there are countless things we could learn, build, fix, automate, improve, or explore. Consciously evaluating every possibility would consume all our available attention.

Instead, the brain eliminates most options before they become conscious plans by estimating their cost.

This does not only mean financial cost. It includes total personal cost: time, effort, energy, knowledge, risk, complexity, opportunity cost, and the disruption a project might cause elsewhere in life.

When an idea appears, such as building an application, writing a book, starting a company, learning a language, or creating a new service, the first question is rarely whether it is interesting. The first question is whether it is realistically doable.

That evaluation often happens almost instantly and largely outside conscious awareness. How long would this take? What would I need to learn? How many evenings or weekends would disappear? What would I have to stop doing? Would it interfere with work, family, or everything else already competing for attention?

If the estimated cost is too high, the idea is discarded, not because it is bad or because we do not want it, but because the brain concludes that it does not fit within the constraints of our lives.

The internal cost filter

Over time, this process forms what could be called an internal cost filter.

The filter prevents us from wasting attention on ideas that appear impossible, unrealistic, or too expensive to pursue. It is not irrational; it is a necessary optimization mechanism. Without it, we would constantly become distracted by possibilities we could never act on.

The filter is also shaped by experience.

Suppose someone has a full-time job, a family, and only a few hours of free time each week. They may have ideas for a résumé service, a proposal generator, a stock-analysis platform, a personal automation system, a budgeting application, or a language-learning tool.

None of those ideas is physically impossible. They are simply unrealistic within the resources available.

Over time, the brain learns that lesson and eventually stops evaluating similar ideas in detail because previous experience has already predicted the outcome. The next time a comparable idea appears, it gets discarded almost immediately, not because it lacks value, but because experience suggests there is not enough time to make it real.

This means many ideas disappear before we even recognize that we have rejected them. They never become plans or experiments; they remain vague fantasies for a moment and then vanish.

AI changes the filter

AI does more than improve execution speed. It changes the input to the internal cost filter.

A project that once required six months may become a weekend prototype. A task that previously demanded a team may become realistic for one person. Research, planning, coding, design, documentation, testing, and iteration can all be compressed.

Documentation becomes faster, boilerplate disappears, design iterations happen in minutes, testing accelerates, and research that once required hours can often be completed in a fraction of the time.

None of these improvements is revolutionary on its own. The larger effect comes from their combination.

Every hour removed from execution changes the brain's estimate of what can realistically be accomplished. Eventually, enough small reductions accumulate that entire categories of projects cross the line from “not worth attempting” to “worth trying.”

The idea itself has not changed; the estimated cost has. When the estimated cost changes, the filter changes its answer.

Ideas that once triggered an automatic “no” begin receiving a “maybe,” and some eventually become “yes.” That is not merely a productivity improvement. It is a change in how people perceive their available future.

The ideas were already there

The interesting part is that AI does not necessarily create the original ideas. Many of them already existed.

A person may have wanted to build a particular service for years. They understood the problem, saw the opportunity, and could imagine the product, but they also knew that building it would consume months of evenings and weekends. The idea was therefore classified as unrealistic and mentally discarded.

AI does not need to invent that idea. It only needs to reduce its cost enough for the brain to reconsider it.

The shift is not that someone suddenly becomes more creative. It is that they stop rejecting ideas they already had.

That distinction matters. AI may appear to produce an explosion of new ideas, but part of what we are seeing is an explosion of previously suppressed ideas returning to consideration.

The ideas were always present. What changed was their status: they moved from fantasy to option.

The feedback loop

Once one of those ideas becomes real, the process accelerates.

Completing a project does not only produce a product. It also changes how future opportunities are perceived.

Every finished project teaches new patterns, creates reusable components, exposes new problems worth solving, and increases confidence that the next idea is achievable. Execution therefore produces more than output; it produces more ideas.

A developer builds one small application and discovers three related problems. The first application also creates authentication code, payment integration, user-interface components, deployment infrastructure, and knowledge that can be reused in the next project.

The second project therefore costs less than the first. That lower cost causes more ideas to pass through the internal filter, creating more experiments, more reusable components, and more knowledge, which reduce future costs even further.

The cycle reinforces itself: more capability leads to more attempted ideas, and more attempted ideas create more capability.

This is not simply increased productivity. It is accelerated exploration.

Why entrepreneurs notice it first

Entrepreneurs are likely to notice this effect early because they constantly operate near the limits of execution.

They are not necessarily the only people with many ideas. Almost everyone has ideas for things that could be improved, built, automated, or organized differently. Entrepreneurs are simply more likely to test the boundary between an idea and an implementation.

Historically, even highly productive entrepreneurs faced hard limits. They might have had fifty promising ideas but enough time and capital to pursue only one or two. The rest stayed in notebooks, remained in their heads, or disappeared entirely.

AI changes that equation. A single person can now test more concepts, build more prototypes, explore more markets, and abandon weak ideas earlier.

The entrepreneur does not merely become faster; the entire portfolio of realistic options expands.

That creates a new problem. Instead of running out of implementation capacity, the entrepreneur begins running out of attention. The bottleneck moves.

Scarcity does not disappear

AI does not eliminate scarcity. It relocates it.

Yesterday's bottleneck was implementation. As implementation becomes cheaper, selection becomes more important.

The difficult question is no longer “Can I build this?” It becomes “Should I build this?”

Later, that bottleneck may move again. Suppose AI can create ten complete prototypes overnight. The human's job is no longer implementation but deciding which of those completed implementations deserves further attention.

Which one solves a real problem? Which one creates value? Which one is worth maintaining, and which should be abandoned?

Eventually, AI may also assist with evaluation by measuring engagement, running experiments, comparing outcomes, generating alternatives, and recommending which direction appears most promising.

As those capabilities improve, the bottleneck moves yet again, from evaluating solutions to defining worthwhile objectives in the first place.

Humans increasingly specify goals rather than implementations. Instead of writing software, they describe desired outcomes, and instead of designing every workflow, they define what success should look like.

The work moves upward in abstraction. Scarcity remains, but it migrates from execution to judgment, attention, priority, and direction.

Beyond productivity

Most conversations about AI focus on productivity.

They ask how many hours can be saved, how much cheaper software will become, how many tasks one person can complete, or how many employees a company will still need. Those are important questions, but they miss the larger effect.

The biggest change may not be that people perform existing work faster. It may be that they attempt projects they would never previously have considered realistic.

History contains similar patterns.

When photography became digital, people did not simply take the same number of photographs at a lower cost. They took vastly more photographs.

When cloud computing became inexpensive, companies did not merely reduce infrastructure costs. They launched products that would never have justified purchasing physical servers.

When publishing became digital, production became accessible to people who would never have worked with a traditional publisher.

These technologies all shared the same characteristic: they reduced the cost of execution. The result was not only cheaper production but an explosion of experimentation, because projects that previously failed the internal cost filter suddenly became practical.

AI may amplify this effect more than previous technologies because it reduces the cost of many different activities at once, including thinking, planning, designing, coding, writing, researching, analyzing, and iterating.

That combination matters. A lower server bill changes one part of a project, while AI can change almost every part.

The second-order effect

Jevons Paradox explains why cheaper AI leads to more AI usage. The second-order effect is that cheaper execution changes which ideas survive our internal filtering process.

Lower execution costs do not merely increase output; they increase the number of ideas that survive long enough to be tested.

That means society does not simply produce more of the same things. It explores a larger portion of the possibility space, generating more experiments, more startups, more niche products, more scientific hypotheses, more creative work, and more solutions to previously ignored problems.

Most of those experiments will fail, and that is expected. The important point is that they now exist when previously they would never have been attempted.

A society that tests ten times as many ideas does not merely work ten times faster. It increases the probability of discovering ideas that would otherwise have remained invisible.

That is a different kind of productivity boom. It is not only more output per hour but more explored possibilities per mind.

A different way to think about AI

AI should perhaps not be understood primarily as an automation technology. It may be more useful to view it as a technology that lowers the threshold between imagination and execution.

For most of human history, ideas have been abundant while execution has been scarce.

AI shifts that balance. As execution becomes dramatically cheaper, the scarce resources move elsewhere. Attention becomes more valuable than implementation, judgment becomes more valuable than production, and choosing the right objective becomes increasingly more important than deciding how to achieve it.

In other words, the value chain moves upward. Humans spend less time translating ideas into reality and more time deciding which realities are worth creating in the first place.

That does not mean implementation becomes irrelevant. It means implementation is no longer the dominant filter preventing ideas from being tested.

The internal cost filter remains, but its thresholds change. Ideas that once looked absurd become expensive, ideas that once looked expensive become practical, and ideas that once looked practical become trivial.

As each threshold moves, human ambition adjusts with it.

From execution to possibility

Jevons Paradox remains an important way to understand AI. As AI becomes cheaper and more capable, we will undoubtedly use more of it.

But that is only part of the story.

The more profound change is psychological. Our brains continuously filter ideas according to the resources we believe they require, and when AI lowers those costs, ideas that once died instantly begin to survive.

Some become experiments, some become products, some become companies, and some change industries.

The greatest impact of AI may therefore not be that it allows humanity to perform existing work faster. It may be that humanity begins exploring a vastly larger fraction of the ideas it already has.

Many of those ideas will fail, some will change industries, and a few may change civilization. All of them will exist because the internal cost filter gave them a chance they never would have had before.

That is not just a productivity revolution. It is a possibility revolution.