Infinite choice is the curse of the AI era, and you already hate it

AI companies are building products that overwhelm users with infinite choices, leading to decision paralysis and dissatisfaction. Drawing parallels to Netflix, Costco, and Apple, the article argues that consumers prefer curated, limited options that reduce cognitive load, as evidenced by the success of authoritarian product models. The author warns that the AI industry's focus on abundance is a fundamental design flaw that users already hate.

Infinite choice is the curse of the AI era, and you already hate it Why every AI product feels like work AI companies have built their entire product empire on the misconception that infinite choice is a Good Thing ™️ They’ve convinced themselves that abundance is a gift - that we somehow want more options, more drafts, more noise. Ask for a logo, get fifty. Ask for an email, get ten. Ask for a strategy, and drown in a tidal wave of vomited and re-vomited decks, memos, and over-engineered frameworks. But nobody actually wants this. Nobody wants infinite choice and infinite options. Consumers have proven they hate it for thirty years with the products they love, return to, and buy. Every successful product, from the iPhone to the Spotify algorithm, relies on one simple premise: we steal your decisions, we tell you what to do, and you love it. We Literally Pay Not to Have Choice Netflix launched on the selling point of an infinite catalog: every movie at your fingertips forever. And then they discovered an infinite catalog destroys engagement, and the company has spent north of $20 billion in original content trying to fix the disaster. Infinite catalogs drive paralysis and churn. Subscribers stared at the interface, couldn’t pick anything, and felt depressed about their evening. So Netflix built algorithms to strangle the options. The company created categories, buckets, personalized queues, and autoplay. It invested billions in original content to produce a single answer to “what should I watch tonight?” Netflix pivoted its whole product into a machine that killed the choice it promised to deliver. And engagement skyrocketed. Costco sells around 4,000 SKUs. The average supermarket carries 30,000. Costco’s model is commercially authoritarian: we decided for you. You get one mayonnaise, one ketchup, and one olive oil. And customers experience this as pure relief. Costco posts customer loyalty scores that dwarf most retailers’ ambitions. The constraint is an actual value proposition, and buyers worship it. When Steve Jobs returned to Apple, he gutted the product line to four devices and murdered the options that had infested Apple throughout the 90s. Every few years since, Apple removes something a port, a headphone jack, a feature users swear is structural , and users adapt and go back to buying more Apple products at higher prices. Analysts might call it arrogance, but over and over again, customers prove they would rather pay a premium to be bossed around than face an open shelf. What Restaurants Know That AI Doesn’t I’m the co-founder of a bar here in Sydney https://www.bar2061.com/ , and we’re incredibly proud to have a bare-bones menu. We have a handful of dishes we love, that our regulars love - late-night toasties and so on - and a limited number of bespoke cocktails. People find something they like, and they trust that we’ll make it well, because we’re not distracted by trying to stay on top of a 200-item Bible of canapés and mixed drinks. A 200-item menu dumps the kitchen’s failure onto your cognitive load instead of offering you freedom. Every item past a certain threshold means the staff refuses to offer an opinion or say, “This is what we’re good at, and this is what you should eat.” In omakase, the Japanese tradition, the chef decides what you eat based on what’s good that day. Omakase commands prices that would make a steakhouse blush because the chef strips away every choice. People pay more, sometimes dramatically so, to have fewer options. Silicon Valley executives apparently cannot comprehend this. But deciding is work that depletes the same cognitive resources as manual labor, and people with finite resources will pay a fortune to get that work done for them. Early dating apps competed on volume: more matches, more profiles, more swipes. Users discovered more matches correlated with worse outcomes. More matches meant perpetual browsing and permanent background anxiety that the right person was in the next batch. It created a learned helplessness about commitment that made entire cohorts functionally unable to go on a second date… Re: Executive Cowardice Senior executives have plenty of information. They know what their customers want. Deep down, they know . Their bottleneck is cowardice. The consultancies and chiefs of staff who command the highest fees walk in and say: you should do this. They provide the guts the executive lacks. McKinsey etc charge premium fees for cover, not analysis. Current AI can’t provide this. It does research well, but it can’t draw definitive conclusions and stand by them in an argument. It generates options compulsively, hedges reflexively, and keeps spewing multiple perspectives with the evenhandedness of someone terrified of being wrong. Engineers optimized it for the appearance of comprehensiveness rather than the delivery of guidance. It’s the world’s most capable intern: brilliant at gathering, completely gutless at deciding. The Paradox Is About to Get Worse Barry Schwartz documented this dynamic in 2004 in The Paradox of Choice. More options reliably produce worse outcomes across purchase decisions, medical choices, romantic partnerships, and career moves. This is one of the most robust findings in behavioral economics. But tech executives spent twenty years building products that deliberately violated it, in the name of engagement metrics that turned out to measure anxiety as much as satisfaction. AI will violate it harder, at higher intensity, across every product category. Every AI writing tool gives you ten versions. Every image generator gives you a grid of four. Every AI search interface now returns a summary with six angles, three caveats, and a dozen links for further reading. The volume of options hitting knowledge workers has increased by an order of magnitude over the past two years and shows no sign of stabilizing. Pre-AI information overload looks quaint from here. Decision fatigue has moved from psychology journals into mainstream conversation. Consumers buy capsule wardrobes, meal prep systems, and morning routines with zero variables. People pay coaches and consultants specifically to reduce the number of choices available to them. The demand signal is deafening. But the industry is building in the exact opposite direction. What the Winning Products Will Have to Do Generation is becoming a commodity; every foundation model can produce competent output on demand. The moat belongs to products that use generation internally and deliver conclusions externally, doing the synthesis behind the curtain and handing you an answer, not a menu. The product will say “here’s your onboarding email,” instead of “here are six approaches; which direction feels right?” The product will say “here’s the angle; here’s why it wins,” instead of “here are eight potential angles for this pitch deck.” The AI does the work of having a point of view. The user gets the output of that work. This sounds simple, but it’s the hardest thing to build, because builders must hold the product accountable when it is visibly wrong. That’s why most AI products won’t do it. Taking a position means being accountable for the position. Generating options is safe; every direction is defensible because the generator didn’t actually pick one. The option-generation interface is the AI industry’s version of the 200-item menu: it looks like capability, but it’s the product and its creators hedging their bets. The highest-value human services have always understood this. The doctor who says “here’s what you need to do if you want to live to see your kid graduate” is worth a whole lot more than a doctor who says “here are your options, let me know what you decide.” The editor who cuts the bad paragraph is worth more than the editor who leaves a comment musing about it. The premium has always gone to people willing to have a view and defend it. AI is no different.