Don't Subscribe So Casually

Or, why your AI subscription is worth analyzing

Most people pick subscriptions the same way they pick snacks. Subscriptions, however, are more like roommates than Oreos. Anything you subscribe to gets a small, recurring vote on who you become.

Being conscious of this has been worthwhile throughout the modern era of subscriptions, but the rise of chatbots — which can be customized, tailored, and packaged to become even more addictive and to amplify the negative effects of subscription models — makes revisiting the issue especially important. I don’t think people think of ChatGPT as a normal subscription: it’s shiny, new, and powerful, but from a financial perspective — and from the perspective of the company’s incentives — it’s just another Netflix.

If someone offered you a magic button that gave you ten dollars now, but carried a high probability of altering your tastes, your routines, and the way you think, would you press it?

A subscription is that button.

Subscriptions versus goods

Buying a thing—like a kettlebell, a tablet, or a Chihuahua-sized raincoat—is relatively easy because you can just weigh the pros and cons of that specific item. Granted, advertising exerts a major psychological influence, and clever store layouts prove our purchasing whims are, to a large extent, manufactured. But there are straightforward strategies to overcome these psychological foibles: don’t shop when you’re hungry, wait a week before hitting “buy,” and return the items you never use.

A subscription is an access pass to some sort of good, like a selection of TV shows or warranty plan. It is inherently future oriented: you are purchasing a set of possibilities for a future time period, thereby giving your future self more choice. This gives you additional options, yes--- which is the favorable part of the equation--- but it also changes your future behavior, sometimes substantially.

There’s no clear delineation here between because all goods change your psychological makeup. The presence of a physical object doesn’t immunize you against subscription-like effects. Many physical goods are simply hardware anchors for ongoing subscriptions. My chief concern here is with the entire class of products—whether digital services or plastic and silicon—that operate as significant, continuous influences on your future behavior.

Even good subscriptions are hard to evaluate rationally

Some subscriptions are just bad deals. For instance, gym memberships are subsidized by people who barely go to the gym and would be better served by buying day passes. Those people would be economically better off dropping their gym membership if they know they aren’t going to use it.

However, even worthwhile subscriptions have consequences that are fundamentally complex and difficult to analyze. Take my Uber One membership. Is it helpful? Well, it makes ordering food significantly cheaper, but I’ve also ordered more takeout. Is this worth it? Did I really consider this when I subscribed?

There are great, worthwhile subscriptions out there. Insurance is a lifesaver and prevents millions of people each year from losing their homes while giving them peace of mind. Costco inspires a loyalty that borders on the theological. All of these seem to have positive utility for the people who use them.

My key point is that matter how good the deal is, a subscription shifts the subscriber’s preferences. People who subscribed to Costco have been morphed, changed, transformed from their pre-Costco selves into people who shop at Costco. They buy more at Costco because it is cheaper, and prefer it to alternatives. They may buy different things and have slightly different preferences.

This psychological nudging is also a factor for goods themselves. “If I buy this running shoe,” one might think, “I am more likely to run in the future and thus buy many additional pairs of running shoes.” Yes, this is true, but it’s a much weaker compulsion. Substantial research has proven that subscriptions have high magnitude effects on consumer behavior. Most products (with some notable exceptions, such as Apple products) don’t function this way.

You can only limit the psychological effects by limiting your benefits

Buying a subscription will, in the vast majority of circumstances, change who you are.

You can, of course, try to discipline your way out of this trap. If you already order takeout three times a month, vow to never increase that frequency, and find that Uber One makes those exact orders cheaper—by all means, partake. You can treat an HBO subscription as a purchase of one month of Game of Thrones, and no other shows. Then the purchase becomes like a good, and you can once again manage the pros and cons.

However, playing it that safe means leaving a lot of value on the table. You’ll miss out on Succession and a host of other phenomenal shows. At Costco, there are a plethora of great deals that can genuinely enhance your life and reduce your yearly expenses. And credit cards! Don’t even get me started on credit card rewards.

Embracing these benefits is inextricably tied with surrendering to outside psychological manipulation. It takes a lot of willpower to analyze your streaming choices and purchases on any regular basis; most people just act. And can you really, really predict you won’t watch more TV if you have HBO or that you won’t buy ANYTHING extra, ever, on Uber Eats?

Now think of how limitless the domain is with Claude or ChatGPT...

As a consumer you’re at a big disadvantage here

As a consumer, you are at a structural disadvantage. Companies are money making machines. You’re just a person who wants to have a good life.

All incentives, for The Company, point to them offering stuff that makes them money. This doesn’t preclude the consumer from benefiting, but user benefit is a secondary concern.The more direct concern is that consumers feel like they are getting a good deal and continue purchasing the stuff.

When the product is a physical good, The Company’s incentives still largely align with yours. They might employ marketing tricks or or make a product less long-lasting to encourage more purchases,, but the consumer still has to actively decide to buy the object.

Subscriptions remove this friction, allowing companies to optimize for behavioral modification over direct product value. Armed with massive financial resources, a company will always structure a subscription for long-term profitability. This is not a Dr. Doofenshmirtz-like evil plot, and it usually takes the form of optimizing aggregate metrics like “profit per subscriber” or “monthly retention.” (I’ve looked at the financial reports, as boring as they are.)

I promise you that every company with a subscription is actively doing this. ChatGPT is A/B testing which responses make you stay longer. Netflix does this with the movies they show you. Spotify determines which “2010s Morning Sad-girl Era Mix” will engage you the most (it’s been pretty accurate for me).

The Company may not know how you specifically will react to a tweak, but they understand the macro-level behavior of millions of users. You cannot match this asymmetric warfare.

Subscribe to companies that align with your values

Your subscription decisions should be heavily influenced by the company’s stated goals, and if they align with your own.

For Costco, the reason they offer a subscription is to make you shop more there. The company explicitly describes its model as low prices producing high volume and rapid turnover. The membership fee gives Costco a reason to retain members, while the warehouse model gives members a reason to consolidate ordinary purchases there. If you want to start buying more things in bulk for a cheaper price per-unit, great! It’s a good subscription for you.

Amazon, on the other hand, has decided to make Prime the marketplace of convenience. They want to be the default place you go without price-comparison or complex searches; they relentlessly optimize the shopping process to reduce the number of clicks you need to make a purchase. Of course, this means they rely on lazy purchases and impulse buys to stay in business. If you’re a subscriber, can you think of a time you bought something just because it was easy?

AppleCare, iCloud, and Apple products in general, are a good example of something with clear upsides and downsides for different consumers. Apple wants to lock you in to their ecosystem and make product integration, like between iPhones and Macs, as seamless and convenient as possible. Of course, their goal here is to sell you more iCloud, more iPhones, more Apple products. No doubt these products are more expensive than the competition. However, if this cross-device ease of use is what you seek, the Apple ecosystem is like no other.

Admittedly, a lot of this is vibes-based. Companies lie, and not infrequently; tobacco companies did not exactly lead with “our business model depends on making a deadly product and getting half the country addicted to it.” But it’s fairly straightforward in most cases to get a rough picture of the product that company leadership envisions. Some giveaway words I found in reports were “convenience,” “engagement,” “retention,” “ecosystem,” “frequency,” and “loyalty,” to name a few.

On the preservation of self

Since time-immemorial, people have bragged about “self-made,” in both the financial and philosophical senses of the phrase. I will leave it to the philosophers and armchair historians to answer if there ever truly was such a person. However, the online era has made it incredibly difficult to figure out where corporate influences end and our own thoughts begin. Subscriptions are one particularly explicit manifestation of this.

AI subscriptions are particularly insidious because they carry the double whammy of a highly manipulative product that you can literally talk to and are of course subject to all other characteristics of normal subscriptions. Furthermore, it’s pretty unclear what the business model of these companies are. To paint a grim picture: would they really be so unhappy if chatbot usage were to cause mass deskilling and force widespread reliance on AI? What kind of change does this exert on the human psyche?

I don’t really have a good answer here, because it is becoming increasingly difficult to compete without using these chatbots to their fullest extent. Still, it’s worth continually reevaluating the incentives of companies like Anthropic, OpenAI, and Google, and weighing those incentives carefully alongside the raw capabilities of the models.

If you don’t actively choose who you want to be, one of these corporations will happily make that choice for you.

Thanks for reading The Best Worst Case! Subscribe for free to receive new posts and support my work.