From industrialization to personalization

The age of copies

Everything around us is a copy: the same car model, shirt, medicine, software, insurance product, textbook, house design. Industrialization was fundamentally a technology for making copies cheaply. This gave billions of people access to things that previously only the wealthy could afford.

Look around you: most of the things we use are copies.

For the last two centuries, progress has largely meant learning how to make the same thing repeatedly, reliably, and cheaply.

Before industrialization, many products were expensive because they required individual human labor. Mass production changed that. Instead of making one thing for one person, we learned to design one thing and reproduce it millions of times. This was revolutionary.

Ford’s moving assembly line demonstrated the power of mass production: by standardizing parts and production processes, it made cars far cheaper and more accessible. The innovation was not just the automobile; it was the system for producing the same automobile millions of times.

This tradeoff worked extraordinarily well: less personalization in exchange for dramatically lower cost and greater access. It worked so well that we take it for granted.

Why the economy rewards sameness

Creating the first version of a product is expensive. Research costs money. Design costs money. Engineering costs money. Expertise costs money.

But once the first version exists, producing the next copy has to be much cheaper.

Consider a simple software product. A company might spend $2 million on research, design, engineering, testing, legal work, and launch. Selling it to one customer would require recovering nearly all of that cost from that customer. But if it can sell essentially the same product to one million customers, the original creation cost becomes only $2 per customer.

Example:

• Cost to design and build the first version: $2,000,000
• Cost to provide one additional digital copy: $0.10–$2.00
• Fixed creation cost per customer at 1 customer: $2,000,000
• Fixed creation cost per customer at 1 million customers: $2.00

The economic incentive is obvious: spread the cost of creating the first version across as many customers as possible. This is why companies are designed for scale.

They create one product, one operating model, one process, one message, and distribute it to as many people as possible.

The modern corporation is largely an optimization machine for spreading the cost of creating one thing across millions of copies.

Computing was the first break from this model

Computing began changing this because digital products do not need to be reproduced in the same way physical products do.

Historically, a newspaper gave everyone approximately the same product. From the same building, they created the same front page, same articles, same layout, and so on. You wanted to sell as many copies as you could.

Then computing, storage, and distribution became absurdly cheap.

Instead of creating one newspaper for one million people, technology companies could create one million different newspapers for one million people. Newspapers had to print essentially the same product for everyone. Digital distribution removed printing and distribution constraints; cloud computing made storage and computation cheaper; suddenly Facebook, Google, Spotify, Netflix, etc. could assemble a different product for each person.

The newsfeed became a product assembled specifically for you, repeated billions of times.

Spotify does not simply distribute the same music catalog in the same order to everyone. Its recommendation systems create personalized home screens, playlists such as Discover Weekly, daily mixes, radio stations, and suggested next tracks based on a listener’s behavior, tastes, context, and similarities to other listeners.

Netflix can personalize not only what it recommends, but how it presents it: the rows on the home page, the order of titles, and sometimes the artwork shown for a title can vary by viewer. Two people can open the same service and experience meaningfully different “front pages.”

Amazon’s store is not one fixed catalog page. Search rankings, product recommendations, bundles, “frequently bought together” suggestions, offers, and merchandising can differ by customer, based on browsing, purchasing, and contextual signals.

Google Search made information retrieval increasingly personalized and contextual. Results can vary by query intent, language, location, device, freshness, and prior signals—so “the same search engine” can produce a different useful answer for different people.

The product was no longer the same as everyone else's. The product became something built just for you.

Cheap computing made personalized information possible. But information was only the beginning.

Now intelligence is becoming cheap

AI is pushing down the cost of something much more important than storage or computation: intelligence. By intelligence I mean activities such as:

• writing
• analysis
• research
• programming
• design
• planning
• decision support
• teaching
• customer service

Intelligence is not free, and the frontier can still be expensive. As of September 2026, OpenAI charges $10 per million input tokens and $50 per million output tokens for GPT-6 Astra, its most capable model. But capable intelligence is already available for dramatically less: GPT-5.6 Sol costs $4/$20, Terra $2/$12, and Luna just $0.20/$1.20 per million input/output tokens. The important part is the direction: intelligence is becoming cheap enough that we can apply it to tasks where paying for individualized human attention would never have made economic sense.

Historically, personalization was expensive because personalization required human attention. A lawyer could personalize advice, but only by spending more hours. A teacher could personalize education, but only by spending more time with each student. A software engineer could build custom software, but only at high cost.

AI changes this equation because the cost of personalization goes down dramatically; while Industrialization made copies cheap, AI is making uniqueness cheap.

What happens when custom becomes cheap?

This creates a different economic possibility. Instead of “Build one thing and sell it to everyone,” we can increasingly say: “Build a different thing for every person.”

It goes far beyond a social media newsfeed. The first generation of personalization selected from a limited set of existing options. The next generation may actually create the product or service differently for each person.

Example one: education

Today, education is largely standardized because the cost of developing a curriculum is too high.

A teacher cannot realistically create a completely different lesson, explanation, exercise, and pace for every student because it was simply too much work.

With cheap intelligence, a student could have a personalized curriculum. Not static, but adaptive to the student's progress. Imagine a curriculum that adjusts itself every day based on an individual student's progress.

The cost of individualized teaching moves from years to minutes.

Example two: software

Most companies buy software designed for thousands of other companies. They then change their own processes to fit the software. Think about any CRM, ERP, ECOMMERCE or any other tool you use at work.

Custom software has historically required large teams, years, and a lot of money. With the reduced cost of intelligence, software creation becomes dramatically cheaper; companies may increasingly have software built around their own exact workflows, terminology, customers, and problems.

Instead of asking, “Which software should I buy?” companies may eventually ask "How do I solve this specific problem I only have?"

Example three: professional services

Let's look at Finance. Those services are already personalized, but personalization is expensive because every situation is different. The world is changing faster than ever before, and strategies that were set for many years, now are good only for a few weeks.

Today, your financial advisor spends weeks reviewing your portfolio and developing new strategies to rebalance it. Therefore, the service costs are high and only viable for high-net-worth individuals.

With AI, that's not true at all. A service that previously made economic sense only for wealthy individuals or large corporations could become accessible to anyone.

From bits to atoms

This transformation will happen faster in digital products than physical products. AI can personalize a report instantly. It cannot manufacture a unique car instantly (yet).

Physical personalization still depends on manufacturing, logistics, materials, regulations, and other factors. In sectors like medicine, food, clothing, and manufacturing, there will be changes in how things are produced.

Bits will personalize first. Atoms will follow.

If intelligence becomes dramatically cheaper, the structure of companies may change too. Large companies currently benefit from being able to afford specialized employees, engineers, lawyers, managers, and more.

If a small company can access much of this capability through AI, then David can fight Goliath at the same level. Imagine a company with two people that can serve thousands of customers while still being profitable.

From copies to uniqueness

For roughly two centuries, the economic miracle was learning to make copies more cheaply. The same thing for everybody. That model raised living standards by making previously expensive goods accessible to billions of people.

This era adds another possibility.

When intelligence becomes cheap, mass-producing the same thing no longer makes sense.

We spent two centuries learning how to make the same thing for everyone. We just entered an economy built around the opposite assumption: how to make one thing just for you.