The Four Horsemen of the AIpocalypse
AI is fracturing into at least four different visions of the future.
The biggest technology companies in the world are making different bets on how AI should work. This isn’t just competitive positioning; it reflects genuinely different theories about where value will accrue and how people will want to interact with AI.
It’s worth understanding these bets, because we’re going to be living inside all of them, everywhere, all at once.
The Invisible Future: AI as Glue
The first future is one where AI disappears into your existing tools. You don’t open an AI app, your apps just get smarter. Your calendar understands your priorities. Your email surfaces what matters. Your file system anticipates what you’re looking for.
Apple made this bet explicit last year. Their Foundation Models framework gives any developer access to on-device AI: free inferences, offline-capable, privacy-preserved. The AI isn’t something you buy; it’s something that makes everything else work better. In this view, the most successful AI features are the ones you forget are there.
Google and Microsoft are making similar moves. The AI becomes infrastructure: invisible, ambient, always on.
The Specialized Future: AI that Knows Your Work
The second future is vertical. Instead of one AI that’s moderately good at everything, you get AIs that are exceptional at specific things.
The market is betting heavily on this future. For example, Harvey is building legal AI that understands citation formats and contract pitfalls. It’s not a chatbot with legal prompts, but a system trained on how lawyers actually work.
General-purpose AI is impressive precisely because it handles everything adequately. But “adequately” isn’t good enough when the stakes are high. A legal AI that’s 95% accurate is malpractice. A medical AI that misses edge cases is worse than no AI at all.
The specialists are building for the 5% that matters. They’re integrated into specific workflows, trained on specific data, evaluated against specific success metrics.
If you work in a domain with expensive mistakes and hard-to-articulate expertise, this is probably your future.
The Everything Future: AI as OS
The third future is the most ambitious: AI doesn’t enhance your apps or specialize in your field; it becomes the operating system. One place to go. One intelligence to talk to. Everything flows through the conversation. It’s not an app or part of an app. Apps are in it (sort of): either third-party apps that are called when necessary or apps that are created on the fly by the AI itself.
This future includes AI-first hardware: devices where the AI isn’t a feature but the foundation. The app paradigm dissolves. You describe what you want; the AI figures out how to deliver it.
The Mediated Future: AI as Filter
The fourth future is probably the most radical — and the least discussed. This isn’t AI as a tool you use or a place you visit. It’s AI as the universal intermediary. Every digital interaction, filtered through AI. Every piece of information, filtered and curated by an agent before it reaches you.
In this future, you don’t browse the web, an AI browses it for you and reports back. You don’t scroll feeds, an AI decides what’s worth your attention. You don’t comparison shop, an AI negotiates on your behalf. The raw internet, the unfiltered firehose of information and content, becomes something most people never see.
We’re already partway there. Search results are increasingly AI-generated summaries rather than links to sources. Email apps draft replies you can send with one tap. Social feeds are algorithmically sorted, and now AI is starting to generate the content that fills them.
This future raises some serious questions. Who decides what the AI filters out or lets in? What biases get baked into the mediation layer? When everything is pre-digested, what happens to serendipity: the accidental discovery, the stumbled-upon insight? Who pays for all this content? Why would places like Amazon or The New York Times allow something like this? If my AI is Gemini is it going to show me sites and products that are favorable to Google—even if they aren’t favorable to me?
The convenience is obvious. The cost is harder to see.
Living in The Overlap
These futures aren’t mutually exclusive. They’re going to coexist, and you’re going to navigate all four.
Your phone’s AI layer might handle the mundane: scheduling, triage, coordination. Your specialized tools will handle professional work where domain expertise matters. Your conversations with general-purpose AI will handle the open-ended thinking, the “I’m not sure what I need yet” exploration. And an AI filter will bring or hide things from you without your even knowing it.
The monolith of “AI” was always an illusion, a temporary artifact. What we’re watching now is differentiation, the technology maturing into its other shapes.
Get comfortable with the complexity. It’s not going away.
