Optimized Toward What?

I started researching this because I wanted to answer a pretty simple question:

Is Amazon evil?

Not in the cartoon-villain sense. I wanted to know whether the company I use constantly is actually as bad as people say it is. Does Amazon destroy small businesses? Is Amazon Basics basically a giant company watching what sells on its marketplace, copying it, and eating the original seller’s lunch? Are the warehouses actually terrible places to work? Is Amazon really sustainable, or is that mostly marketing?

And maybe the bigger question: Amazon has become one of the most powerful companies in the world. Is what it’s doing actually good for capitalism in the long run?

I expected to come out of this either more comfortable with Amazon or ready to cancel Prime. Instead, I ended up thinking about robots, universal basic income, employee ownership, the Protestant Work Ethic, frogs, and fully automated luxury gay space communism.

So this got away from me a little bit.

Is Amazon evil?

The annoying answer is that I don’t think that’s a very useful question anymore.

There are plenty of legitimate reasons to criticize Amazon. In 2024, OSHA reached a corporate-wide settlement with Amazon after a multi-site investigation involving ergonomic hazards and injuries. The agreement required Amazon to make changes across its facilities, including better risk assessments, engineering controls, training, and monitoring.

The FTC and a group of states are also suing Amazon over alleged anticompetitive practices. The government alleges that Amazon uses its power over its marketplace in ways that hurt competition, raise costs, and make it harder for rivals to challenge Amazon. Those are allegations in an ongoing case, not facts I’m going to claim true, but they’re serious allegations.

At the same time, Amazon has unquestionably created opportunities for small businesses. According to Amazon’s own 2025 Small Business Empowerment Report, independent sellers now account for more than 60% of sales in Amazon’s store, and more than 75,000 independent sellers surpassed $1 million in Amazon sales in 2025. That’s Amazon talking about Amazon, so take the self-congratulation accordingly. Still, there are a lot of businesses making a lot of money using infrastructure Amazon built.

And that infrastructure is incredible. A small company can put a product online and suddenly have access to things that would have been completely out of reach for most small businesses not that long ago:

• payments
• advertising
• warehousing
• customer service
• national distribution
• international customers
• an absolutely absurd logistics network

That’s real value. It’s also dependency. Amazon can help a small business reach millions of customers while also gaining enormous power over a business that now depends on Amazon to reach those customers. Both things can be true.

Amazon Basics has the same problem. I’m less interested now in whether an Amazon Basics camera tripod is a “rip-off” than I am in the structure of the marketplace itself. If you own the store, control the search results, know what customers are buying, know what sellers are selling, and also sell products of your own, there are some pretty obvious conflicts of interest there.

And this isn’t just hypothetical. A small seller named Pirate Trading sold camera tripods under the Ravelli brand on Amazon. According to a Wall Street Journal investigation, the owner said Amazon eventually introduced AmazonBasics versions of six of his best-selling tripods. He said one shared his design and components, came from the same manufacturer, and Amazon sold its version for less than he was paying the manufacturer to make his.

That’s kind of the problem I’m talking about. Amazon didn’t just have more resources than the small business. It owned the marketplace where the small business was selling and could see what was selling well there.

Amazon disputed parts of the Journal’s reporting, so I’m not going to pretend I know everything that happened behind the scenes. But even without declaring Amazon guilty of anything, that’s a pretty good example of why I’m uncomfortable with the structure itself.

That doesn’t automatically make Amazon Basics evil. Competition is good, and cheaper products are good for consumers. But there’s a big difference between competing in a marketplace and competing against the company that owns the marketplace.

The environmental story is similarly annoying if what you want is a clean answer. Amazon is doing real work on efficiency. According to its 2025 Sustainability Report, carbon emissions per shipped unit fell 7% in 2025, it had deployed more than 52,000 electric delivery vans globally, and its overall carbon intensity had fallen 38% since 2019.

Great.

Its absolute carbon emissions also rose 16% in 2025.

Not great.

Part of that is because Amazon is growing enormously, including a massive buildout of data centers for AI. Amazon can get cleaner at doing each individual thing while doing so many more things that its total environmental footprint still gets bigger.

That’s where this whole article started changing for me. I realized that isn’t really just an Amazon problem. It’s an optimization problem.

Optimized toward what?

Amazon is an optimization machine. I don’t even mean that as an insult. It’s one of the most impressive things about the company.

Amazon obsesses over how long things take, how far packages travel, where inventory sits, how boxes are packed, how servers are used, how quickly products load on a screen, and how many steps exist between wanting something and buying it. Amazon’s own sustainability material literally talks about how it “almost obsessively” thinks about its boxes.

That’s kind of the whole company: optimize, scale, repeat.

And it works. Amazon generated about $717 billion in revenue in 2025, with about $80 billion in operating income. AWS alone generated about $45.6 billion in operating income. This is not a company that accidentally stumbled into success. Amazon is really, really good at optimization.

But optimization leaves out an important question:

Optimized toward what?

Suppose I give an employee an AI tool and it makes that employee 20% more productive. That’s a technological success. But what happens to the 20%? Maybe the company produces more, customers get lower prices, profits increase, the employee gets paid more, or the employee works less.

The technology doesn’t answer that question. We do.

I actually learned a version of this lesson almost twenty years ago.

I learned this from SEO

When I was around 20, I got a job working for a tiny SEO company. This was back in the write-articles-and-build-backlinks era of SEO. There was the owner, his assistant, and me. The owner was working on a repeatable process for doing SEO work for clients, and I helped develop it. Eventually we got the process down, the company hired more people to do it, it worked, the clients were happy, and the company grew.

During the development phase, we made $10 an hour. At 20 years old, before 2010, I was perfectly happy with that. I lived in a barn anyway.

Once the process was figured out, we switched from hourly pay to $35 per project. A project usually took around three to three and a half hours, so the deal was pretty simple: if you got into a flow and finished faster, good for you. You made more per hour.

I loved that.

Then something completely predictable happened: we got better at our jobs. We were doing this process 30 or 40 hours a week for years, so of course we got faster. We also figured out little shortcuts. I remember using macros in Notepad++ and Excel to automate some of the repetitive parts.

Eventually I could get a project done in maybe an hour to an hour and a half. On a good day I was making something like $35 an hour, which was incredible money to me then and, slightly painfully, more than I make per hour right now almost twenty years later.

But here’s the important part: I didn’t use that productivity gain to make as much money as humanly possible. I kept my income around where I wanted it and took a lot of the gain as time. I worked on music, built my own SEO and affiliate marketing websites, and spent time with my friends.

I had gotten better at my job, and the reward was that I got some of my life back.

Then the owner found out how fast we’d gotten and cut the rate from $35 per project to $15. We were pissed.

I’m leaving his name and the company’s name out because this isn’t meant to drag the guy. We’re still friends. I love him. He was also pretty young at the time and running a growing company, and I don’t blame him now for a business decision he made almost twenty years ago. I also don’t know what happened on the client side. Maybe he lowered their prices. Maybe he kept the difference. Maybe it was somewhere in between.

But I’ve never forgotten the lesson. We had made the process better. We had gotten faster. We had even found small ways to automate parts of it. Productivity went up. For a while, I got to keep some of that gain as free time. Then the pay changed, and I didn’t.

That’s when I learned that getting more productive doesn’t necessarily mean you get to keep the benefits of that productivity. Someone still decides where those benefits go.

Nearly twenty years later, we’re having basically the same conversation about AI and automation, except on a much bigger scale.

The automation dividend

If technology lets us make the same amount of something with less human labor, or more of something with the same human labor, we’ve gained something. There’s now some extra value available that wasn’t available before.

Call it an automation dividend.

I didn’t invent that term. People have used versions of it before. But I think I’ve settled on what I mean when I use it:

The automation dividend is the portion of technological productivity we choose to take as increased human freedom rather than simply increased output.

That dividend doesn’t have to be a check in the mail. It could show up as:

• higher wages
• lower prices
• more profit
• more production
• better public services
• broader ownership
• more free time

That last one seems neglected.

There is already some evidence that AI productivity doesn’t automatically turn into more free time. A 2025 NBER working paper looking at AI exposure and time-use data found that greater exposure to AI was linked with longer work hours and less leisure, especially where AI helped workers become more productive. The authors discuss limited worker bargaining power as one possible reason the gains may instead go to firms or consumers.

It’s one study and a working paper, so I’m not going to turn that into “science proves AI makes everyone work longer.” But it illustrates the question perfectly. AI makes Spreadsheet Guy faster. Great news, Spreadsheet Guy! You can go home at 3, right?

No. Make more spreadsheets.

The Bureau of Labor Statistics has also documented that productivity and compensation, which once moved relatively closely together, began to split apart in the 1970s. There are a bunch of reasons for that, and there are legitimate arguments about exactly how the gap should be measured. I don’t think “productivity went up and THE BOSSES STOLE ALL YOUR MONEY” is a useful reading of the data.

The simpler point is that productivity gains don’t contain some natural mechanism that makes sure workers get those gains as wages or leisure. Technology creates the possibility. Humans decide what happens next.

We’ve already taken some technological progress as time. Most of us don’t work six days a week anymore. Machines have removed huge amounts of household labor. Agricultural technology allowed a much smaller percentage of the population to grow food for everyone else.

We’ve also taken enormous amounts of the dividend as better stuff:

• modern medicine
• air conditioning
• computers
• recorded music
• video games
• refrigeration
• air travel

I like that stuff. I don’t want to go back to 1930 and work fifteen hours a week if the tradeoff is losing modern medicine and my Switch.

But maybe the slider has been sitting too far toward more output for a while. Maybe the next wave of technology should give us some of our lives back.

Here’s the weird contradiction

Technology is slowly decoupling production from human labor. Our economic system still largely couples consumption to human labor.

That’s fine when human labor is needed to make the things humans need. It gets stranger as the first part changes.

Imagine a machine can make enough food for everyone. The food exists. You’re hungry. But you still need to prove you’re economically useful before you’re allowed to eat it.

That’s weird.

We’re nowhere near a world where everything is post-scarcity. I’ll get to that, because scarcity matters a lot. But we already occasionally run into something I’d call abundance without access.

Food waste is an obvious example. I used to assume grocery stores threw away near-expired food mainly because the government forced them to. That’s not really true.

Except for infant formula, federal regulations generally don’t require food expiration dates, and many dates are about quality rather than safety. State and local rules can make this more complicated, especially with perishable foods. Liability isn’t as simple as I thought either. Federal law provides significant liability protection for good-faith food donations, including some direct donations by qualified donors.

Retailers already donate a lot of food, and that’s important to say. But according to ReFED, retailers still generated 3.98 million tons of surplus food in 2024. About 21.5% was donated. About 25% went to landfill or incineration.

There are legitimate reasons some food can’t just be handed to whoever happens to be standing outside: safety, storage, transportation, sorting, labor, timing. I’m not going to pretend logistics disappear because I have a philosophical point to make.

Still, the food exists and the need exists. Sometimes the result is that the food becomes garbage anyway.

Production and access aren’t the same problem.

Spreadsheet Guy and Frog Guy

Let’s imagine two people.

Spreadsheet Guy has a job. He works forty hours a week making internal reports for a large company. He’s pretty good at Excel. The reports are definitely looked at by somebody. Probably.

He makes $70,000 a year, has benefits, pays taxes, and buys things. Productive member of society.

Now meet Frog Guy.

Frog Guy doesn’t have a conventional job. He spends a suspicious amount of time outside. He built a pond, then got interested in native plants because he wanted more frogs around the pond. Now he’s reading about wetland ecology. He helps his neighbor fix things sometimes, he’s learning guitar badly, and he volunteers occasionally.

Yesterday he spent three hours watching a frog.

Frog Guy does a bunch of stuff, but nobody pays him for any of it. He’s unemployed. Depending on how he’s supporting himself, maybe somebody eventually calls him a burden.

Why?

David Graeber’s Bullshit Jobs obviously lurks around this question. You don’t have to accept Graeber’s whole argument to see that employment and usefulness aren’t the same thing. Plenty of useful things aren’t employment:

• raising a human
• learning an instrument
• tending a garden
• taking care of your elderly parents
• learning something difficult just because you’re curious
• making art
• starting a company

All of those things are work. Sometimes having fun is work. Anyone who has spent twelve hours getting destroyed by a complicated strategy game can probably confirm this.

I don’t have a problem with work. I actually like working. I have a problem with treating employment as the only kind of work that proves someone deserves access to society.

Necessary work isn’t the same thing as compulsory employment.

“I had to work for it”

This is the part that probably bothers me more than anything else in this discussion. Whenever you talk about making life easier for people, somebody eventually says:

Well, I had to work for it. Why shouldn’t they?

I don’t know. Why should they?

If something sucked for me and we figured out how to make it suck less for the person who comes after me, good. We don’t make people carry water five miles because their ancestors had to. We don’t smash washing machines because Grandma washed clothes by hand.

The whole point of progress is that people who come after us should have access to things we didn’t. Why would I want my fellow humans to have a worse experience just because I had one? That’s making suffering reciprocal for its own sake, and I don’t see the value in it.

Maybe you like working hard for things. Great. Work hard. Maybe the struggle itself makes the accomplishment better for you. Also great. Why does that mean somebody else shouldn’t eat unless they adopt your philosophy of work?

I don’t want to eliminate work ethic. I want to stop legislating it through deprivation.

There is a real economic argument that everyone needs to contribute because otherwise we won’t have enough. Sometimes that’s true. Then there’s a different argument: everyone needs to contribute because people who don’t work shouldn’t get things.

That’s not really an economic argument. It’s a moral judgment. And the more productive technology becomes, the more important that difference gets.

Scarcity is a legitimate constraint. Resentment isn’t.

Fine. What about Lazy Guy?

Frog Guy is easy to defend. Look at him. He’s learning ecology! He’s helping his neighbors! He’s making his little pond! What a wholesome unemployed person.

Fine. Let’s make the argument harder.

Lazy Guy does absolutely nothing impressive. He plays Xbox, watches television, sleeps late, and eats cereal at 2 p.m. He has no plans to start a nonprofit, learn Mandarin, cure cancer, or build Frogopolis.

Okay. Now what?

If Lazy Guy is consuming resources other people desperately need, we have an allocation problem. If his labor is actually needed for society to function, we have a labor problem. Those are real problems.

But imagine neither is true. His consumption isn’t meaningfully depriving someone else, and his labor isn’t needed to make what society consumes.

What economic problem does forcing him into employment solve?

It may solve a moral discomfort: I don’t like that he gets things without working for them.

Okay. But that’s different.

How much inefficiency are we willing to create just to make sure nobody gets something they didn’t earn?

Imagine we build an entire system to make sure Lazy Guy proves he’s productive enough before receiving basic support:

• applications
• eligibility checks
• work requirements
• case workers
• compliance systems
• appeals

Maybe eventually we force him into a job nobody particularly needs done. At some point, wouldn’t it be easier to just let the dude play Xbox?

I’m not saying every welfare program works this way. That’s a thought experiment, not a budget analysis. I’m asking what our actual goal is.

Freedom includes the freedom to occasionally accomplish absolutely nothing. Some people might use that freedom badly.

That’s what makes it freedom.

Okay, but who takes out the garbage?

This is the classic objection to any post-work idea. If everyone can do whatever they want, who takes out the garbage?

I don’t have a perfect answer, but I have a few:

• Automate as much unpleasant work as we reasonably can. If robots can safely eliminate dangerous, repetitive, physically destructive work, that seems like exactly what robots are for.
• Eliminate unnecessary work. If we produce less garbage, there is less garbage to handle.
• Take out your own damn garbage. Some work is just part of being alive around other humans. I don’t expect an advanced civilization to eliminate my responsibility to clean my own toilet, although if somebody invents a robot that does it, I will not object on philosophical grounds.
• Pay people. For unpleasant specialized work that society really needs and can’t automate, pay enough that somebody willingly does it.

That last one gets interesting when people have a real ability to say no. If nobody will do the job for $15 an hour once refusing it no longer means starvation, maybe the job isn’t worth $15 an hour. Try $25. Try $40. Improve the conditions. Automate part of it. Split the work differently. Maybe we shouldn’t be doing it at all.

What if people won’t do the things we want them to do once they have the genuine freedom to say no?

Maybe that’s information.

The theoretical freedom to quit a job isn’t particularly powerful if quitting means losing your home, healthcare, and ability to feed your kids. Economic security changes bargaining power. Now “no” actually means no.

That seems pretty pro-freedom to me.

Who owns the robots?

This may be the most important question in the whole thing.

Let’s say automation works beautifully. The robots are incredible. AI makes everyone more productive. Factories need fewer people. Logistics become mostly autonomous. The economy can make much more stuff with much less human labor.

Awesome. Who owns it?

It’s easy to turn this into Bezos gets all the robot money and refuses to share, except that’s not actually how Amazon works. Jeff Bezos owns about 8.8% of Amazon. Vanguard owns about 7.2%. BlackRock owns about 5.9%.

Amazon is publicly traded. Millions of regular people own Amazon indirectly through mutual funds, retirement accounts, pensions, and index funds. I’m one of them.

So…don’t we already own the robots?

Kind of…

…And I think that’s actually pointing us toward something useful.

The problem is that we own them very unevenly. According to the Federal Reserve’s Distributional Financial Accounts, in the first quarter of 2026 the top 10% of American households by wealth held roughly 87% of the corporate equities and mutual fund shares represented in those wealth groups. The bottom 50% held roughly 1%.

So yes, the robots are publicly owned. Some of us just own a lot more robot than others (I am not one of them).

That makes me much more interested in ownership than I was when I started researching Amazon. Maybe the problem isn’t primarily that Bezos gets the robot profits and refuses to spend them. It’s that ownership of the productive technology is heavily concentrated.

Who owns the robots may matter more than whether the robots exist.

We already have other ownership models

This is where I remembered ESOPs.

Employee ownership isn’t hypothetical. Publix describes itself as the largest employee-owned company in the United States, and eligible associates can become owners through employee stock ownership and stock purchase plans.

Publix isn’t a utopia. That’s not my point. My point is that this creates a different relationship between the worker and the capital.

Imagine two otherwise identical companies automate 30% of their work. At Company A, the productive capital is owned almost entirely by outside investors. At Company B, employees own a meaningful chunk of the company. Same robots, same productivity improvement, different people participating in the upside.

There are bigger versions of this idea. Alaska has paid residents an annual dividend tied to the state’s resource wealth since the 1980s. In 2025, the Permanent Fund Dividend was $1,000, with about 619,000 payments made.

Oil isn’t AI, and a natural resource isn’t the same thing as privately created productive capital. I’m not saying “Alaska did it, therefore give everyone Robot Checks.” I’m interested in the mechanism: a valuable asset produces economic returns, and some of those returns go broadly to citizens.

Norway takes the idea much further through its sovereign wealth fund. At the end of 2025, the Government Pension Fund Global held about 21.3 trillion Norwegian kroner, with more than 70% invested in equities.

A modern market economy can collectively own an enormous portfolio of productive capital. That’s not science fiction.

I also recently learned the word predistribution, which I find really interesting. Instead of letting wealth and ownership become extremely concentrated and then relying entirely on taxes and transfers to fix it afterward, the idea is to structure things so more people share in ownership and economic power in the first place.

Employee ownership is interesting through that lens. So are broadly owned investment funds and stronger worker bargaining power. I don’t know what the correct mixture is, and I don’t particularly care what ideological bucket any of this puts me in.

I care whether it works.

UBI capitalism?

I have no idea what the economic system of 2150 should look like. Nobody else does either.

In my lifetime, though, I increasingly suspect the transition looks less like “capitalism ends Tuesday” and more like capitalism with some kind of increasingly strong floor under it. Universal basic income is one possibility, and there are legitimate objections.

Let’s start with the easy one: would some people work less?

Probably.

One large U.S. unconditional-cash experiment found recipients worked about 1.3 fewer hours per week on average and were about two percentage points less likely to be employed.

Okay?

Part of my argument is that people being able to work somewhat less isn’t automatically a failure. People deserve rest. If somebody uses increased economic security to spend Friday with their kid instead of at work, I don’t know why I’m supposed to record that as a loss.

The Alaska dividend gives us another useful data point. A published study found no significant reduction in overall employment from the dividend, while part-time work increased.

That doesn’t prove UBI works. Alaska’s dividend isn’t enough to live on, and different UBI plans would have different effects. But the cartoon argument that giving people unconditional money means everyone stops working doesn’t seem especially convincing.

Inflation is a harder problem. If you give everyone more buying power while the number of houses, doctors, cars, or cheeseburgers stays exactly the same, prices can rise. Housing is an especially obvious problem because supply is already tight in a lot of places. How a UBI is funded matters, how large it is matters, and what the economy can actually produce matters.

This is why I don’t think UBI is the automation dividend. It’s one possible way of distributing part of it. Maybe the future is some combination of:

• cash
• employee ownership
• public investment funds
• universal services
• lower prices
• shorter working hours

I don’t need one grand unified economic theory before I’m willing to make something 1% better.

A 1% improvement in people’s lives is still an improvement.

The goal isn’t survival

There’s another problem with saying nobody should have to work “to survive.”

Survival is a pretty shitty standard. Here’s your 2,000 calories. Here’s your concrete box. We successfully prevented your death. Congratulations.

That’s not what I’m talking about.

The goal is human flourishing, not merely preventing corpses.

And what counts as a reasonable baseline changes with technology and society. A smartphone is a great example. People love complaining about a homeless person having an iPhone.

If he’s really poor, why does he have an iPhone?

Okay. You want him to get a job, right? How does he:

• find one?
• apply?
• get an email from the employer?
• do a video interview?
• receive his work schedule?
• navigate to work?
• access his bank account?
• handle two-factor authentication?

What’s the alternative? Does the hiring manager call the nearest pinecone?

A smartphone can be an astonishing piece of technology that would have been an absurd luxury not that long ago and a basic tool for taking part in modern society.

That doesn’t mean everybody gets every luxury imaginable. Scarcity still exists:

• land
• energy
• copper
• lithium
• water
• ecological capacity

There are only so many waterfront houses in Jupiter, Florida. AI doesn’t manufacture more coastline. Some things will remain scarce even in an extremely automated economy, and markets may still be useful for deciding who gets a lot of those things.

That’s okay.

I still want cool shit

This was actually one of my concerns when I started following this idea to its logical conclusion. If we ever reach some kind of highly automated post-work society, do we stop getting cool technology? Does everybody get Government Phone Model 7?

Because I don’t want that.

I like markets. I like competition. I like weird products. I like the idea that somebody can start a company because they think the existing options suck. I don’t see why a strong economic floor requires getting rid of that.

A high floor doesn’t require a low ceiling.

You can guarantee people a reasonable baseline and still let somebody start GarrettPhone Inc. because he has a deeply held ideological objection to removing the headphone jack. People can still make money, compete, build companies, and get rich. Maybe some of those incentives are useful.

Money isn’t the only reason people make things anyway. Look at:

• open-source software
• Wikipedia
• game mods
• music almost nobody listens to
• people restoring cars that will never be worth what they put into them
• incredibly detailed wiki pages about fictional spaceships

Humans are freaking weird.

We make things because making things is fun. We compete because competition is fun. We learn because knowing things is fun. We solve problems because unsolved problems are fun annoying.

AI plus automation actually make it easier to start a business, too. If software, design, customer service, logistics, engineering, manufacturing, and marketing all become cheaper to access, smaller companies may be able to compete in markets that currently take enormous amounts of money to enter.

Maybe instead of mass production giving us the same twenty products, highly automated manufacturing gives us mass customization. GarrettPhone doesn’t need ten million customers. Maybe it needs 10,000. Eventually 1,000. Maybe someday I design exactly the phone I want and an automated factory builds one.

There is an opposite possibility too. AI may increase the advantage of companies that already control enormous amounts of capital, data, infrastructure, and compute. That’s a whole other article.

But I don’t buy the idea that removing desperation automatically removes innovation. People already create things when nobody is threatening to make them homeless if they stop.

Spreadsheet Guy gets laid off

Okay. Back to Spreadsheet Guy.

The company finally figures it out. The AI can make the reports. It can gather the data, build the spreadsheet, make the charts, and even email the report to the executive who probably wasn’t reading it anyway.

Spreadsheet Guy’s job is gone.

This is where the story normally becomes tragic, and today it probably is. Spreadsheet Guy has a mortgage. He needs health insurance. He needs groceries and electricity. His ability to consume the output of the economy is still tied to his ability to sell his labor to somebody.

But imagine we’ve started capturing an automation dividend. Maybe there’s some UBI. Maybe Spreadsheet Guy owns stock. Maybe his company has an ESOP. Maybe healthcare isn’t tied to his employer. Maybe increased productivity has made a bunch of things cheaper.

Maybe it’s all of those.

Spreadsheet Guy doesn’t immediately lose access to society just because society no longer needs his spreadsheet.

For six months, he does absolutely nothing. He’s exhausted. Twenty years of Spreadsheeting will do that to a person. He plays video games, sleeps late, takes walks, and generally doesn’t accomplish much of anything.

Good. People deserve a break, even if the job they needed a break from was bullshit.

Eventually Spreadsheet Guy gets bored. He starts gardening and notices some frogs. He reads something about native plants and learns that the stuff he planted isn’t particularly useful to local wildlife, so he starts replacing it. Then he digs a little pond. The pond needs more plants. He starts reading about wetland ecology.

Then…

Spreadsheet Guy becomes Frog Guy.

They’re the same person.

The “unproductive people” we’re worried about creating through automation are the same human beings we’re currently calling productive. We didn’t remove Spreadsheet Guy’s work ethic. We removed our ability to force him to use it on spreadsheets.

Maybe he becomes Frog Guy. Maybe he starts a business. Maybe he raises his kids, makes an album, takes care of his dad, or spends the next twenty years playing Xbox.

I don’t know.

For the first time, he gets to find out.

This is where AI comes back into the story

There’s a lot of anti-AI sentiment right now, and plenty of it is deserved. Copyright questions are real. Environmental costs are real. Bad AI-generated slop is real. Labor disruption is real. Market concentration is real. There are AI uses I don’t think should happen.

But I increasingly think fighting the existence of AI and automation is the wrong primary battle.

AI is already producing measurable productivity gains in some settings. One well-known study of more than 5,000 customer-support workers found an AI assistant increased issues resolved per hour by about 14% on average, with much larger gains among less experienced workers.

That doesn’t mean the AI economic revolution is guaranteed. Daron Acemoglu has produced a much more conservative estimate, arguing that AI’s economy-wide productivity impact over the next decade may be fairly modest.

Fine. My argument doesn’t require AI Jesus.

If AI increases productivity 1%, there’s a dividend. If it’s 10%, there’s a bigger dividend. If robots eventually do almost everything, then we have a very different civilization to design.

Instead of spending all of our energy trying to make sure machines never become capable of doing our work, I’d rather spend a lot more of it making sure their ability to work actually benefits us.

The IMF has made a much more sober version of this point, arguing that policy should broaden the gains from AI while dealing with disruption and inequality instead of simply trying to suppress productive investment.

I don’t want to preserve Spreadsheet Guy’s spreadsheet because Spreadsheet Guy needs a paycheck. I want Spreadsheet Guy to have a claim on the productivity that made his spreadsheet unnecessary.

Who owns the robots may matter more than whether the robots exist.

Also, we have to not destroy the planet

There is one enormous assumption underneath all of this: we actually have to be able to sustain the abundance we’re creating. Fully automated luxury anything doesn’t work particularly well if we cook the planet first.

This is another place where Amazon is almost a perfect example. Amazon can make each delivery more efficient, use less packaging, deploy electric vans, and make data centers more efficient. Those are real improvements.

Then Amazon grows so quickly that its total emissions increase anyway.

Efficiency isn’t enough if every efficiency gain just becomes permission to use more resources. The automation dividend can’t just be, Cool, now we can manufacture ten times as much shit.

Maybe some of the dividend should be that we can make what we actually need with less energy, less material, less waste, and less human time.

Again: optimized toward what?

I think the scarce resource is time

This is where I ended up after spending way too much time asking whether Amazon Basics USB cables were ethical.

Technology is extraordinarily powerful. We’ve gotten very good at measuring what it lets us produce:

• More units per hour.
• More packages per route.
• More queries per server.
• More reports per employee.
• More GDP.
• More output.
• More more more more more.

But maybe that’s not the only thing technological progress is for.

Technology should expand the range of lives humans are free to choose.

The washing machine took something humans had to do and made it easier. Farm machinery took huge amounts of necessary human labor and replaced it. Computers eliminated mountains of tedious calculation. If AI can eliminate another pile of necessary labor, good. That should be the dream.

We don’t eliminate work. We eliminate the requirement that work take the form of a job.

People will still work. They’ll build businesses, raise kids, make art, compete, study things, grow food, write terrible novels, write great novels, and build Frogopolis. And yes, some of them will do absolutely nothing.

Fine.

If their consumption isn’t meaningfully depriving anyone else and their labor isn’t needed to make what society consumes, forcing them into employment doesn’t solve an economic problem. It solves our discomfort with somebody getting something they didn’t “earn.”

I don’t think that’s enough.

Because maybe the ultimate scarce resource technology should help us reclaim isn’t lithium.

It’s time.

You get somewhere around 80 years if you’re lucky. No technology is manufacturing another Tuesday you spent at work. Once it’s gone, it’s gone.

The automation dividend isn’t ultimately some abstract economic policy.

It’s your life back.

That’s where this Amazon rabbit hole finally landed me.

I started by asking whether Amazon was evil. I don’t think that’s the right question anymore. Amazon is an incredibly powerful machine for optimization, and AI is about to make machines like it even more powerful.

The question is what we optimize them for.

More packages? More products? More profit? More GDP? Sure. I like stuff. I like technology. I like capitalism when it works. I want the weird phones and video games and businesses nobody has thought of yet.

But I also want some of the gain back as life.

Less time doing work that doesn’t need a human anymore. More freedom to make things, learn things, start companies, raise families, help people, grow gardens, watch frogs, play Xbox, or do absolutely nothing for a while.

Technology is extraordinarily powerful. I don’t think we should measure its success only by how much more shit it allows us to produce.

Measure it by whether it actually improves human life and expands human freedom.

If that eventually leads us to fully automated luxury gay space communism, well…

Bezos is already working on the rockets.