When AI Products All Look the Same, Is That Actually the Point?

Here's a pattern I've seen before: when everyone in tech starts shipping the same product, people assume it's panic. No vision. Throwing spaghetti at the wall. But sometimes—not always, but sometimes—convergence happens because everyone independently discovered the same underlying truth.

Right now, we're watching OpenAI, Google, Lovable, Replit, and Anthropic all move toward what looks suspiciously like the same product. The everything app. The super app. The kitchen sink with an API. And yes, the reflexive take is that this represents strategic confusion, that these companies are diluting their focus in a desperate grab for market share.

I think something else is happening. Something that matters more than any individual product launch.

The Code Thesis

The AI Daily Brief's latest episode walks through a week's worth of announcements that, on the surface, seem unrelated. Google upgrades AI Studio with multiplayer and persistent builds. Lovable pivots from app-building to "general tasks." OpenAI reportedly plans to merge ChatGPT, Codex, and their browser into one desktop app. Anthropic adds Telegram integration to Claude Code.

Different companies, different announcements, identical trajectory. As tech analyst Peter Yang put it: "Code is the foundation of all knowledge work. If an agent can write code, it can also generate apps, presentations, animations, and more."

This isn't a new observation—we've heard versions of it since the early days of computing. But the AI coding models changed the equation. Once your AI can competently write code, the distinction between "coding tool" and "knowledge work tool" starts to dissolve. A presentation isn't fundamentally different from a web app if they're both just code rendered differently.

Google's announcement demonstrates this clearly. They didn't just add coding features to AI Studio—they added database integration, authentication, deployment pipelines. The boring infrastructure that sits between prototype and production. And their flagship demo? Real-time multiplayer games. Not because they think everyone's dying to build laser tag simulators, but because it showcases multimodal capabilities that leverage their YouTube corpus.

They're not building a coding tool. They're building a creation tool where code happens to be the mechanism.

The Lovable Controversy

When Lovable announced they were expanding beyond app-building into data science, pitch decks, and marketing—basically positioning themselves as a "general-purpose co-founder"—the response was swift and skeptical.

"First sign that Lovable is dead," wrote one observer. "Pivoting to general assistant is the most investor-pleasing move you could do." Another called it "complete strategic dilution."

Except Lovable had just reported ARR jumping from $300M to $400M in a single month. That doesn't sound like a company whose core business is "going nowhere."

The criticism assumes that app-building and pitch-deck-creation are fundamentally different products requiring different expertise. But if both are just different output formats from the same underlying code-generation capability, then Lovable isn't pivoting—they're just being honest about what their tool already does.

Replit made essentially the same move weeks earlier with Agent 4, promising to "ship working apps, sites, slides, and more." The pattern appears in Gamma, where you can create documents, presentations, mobile experiences, or webpages from the same starting point. GenSpark's general agent uses code behind the scenes to deliver whatever output format you request.

The abstraction is happening. We're moving from "I need a tool that codes" to "I need a tool that creates," and code is becoming the invisible infrastructure.

What Moats?

But here's where it gets uncomfortable: if everyone has access to similar AI models, and those models can quickly replicate any feature another company ships, what exactly stops users from switching?

As investor Ed Sim notes: "When shipping new features cost near zero, every company becomes every company. And when switching costs are also near zero, who wins?"

This is genuinely new territory. I've watched tech cycles since the PC era, and we've always had some kind of moat—network effects, data advantages, proprietary technology, switching costs. The AI era appears to be collapsing all of them simultaneously.

No barriers to entry, but also no moats. You can spin up a competitor in weeks. Users can port their workflows in minutes. The traditional playbook—build a moat, defend it, extract value—doesn't seem to apply.

Some see OpenAI consolidating everything into one super app while Anthropic makes Claude Code infinitely extensible through integrations, and conclude these are radically different strategies. Maybe. Or maybe they're both responding to the same fundamental problem: when features are trivially copyable, what matters is either total integration (OpenAI) or total flexibility (Anthropic). Different paths to the same destination—making yourself impossible to leave.

The Everything App That Isn't

What makes this particularly interesting is that "super app" might be the wrong framing entirely. OpenAI's Fidji Simo said the company is "done with side quests," suggesting this isn't about building an everything app—it's about recognizing that Codex already is the everything app, and organizing the rest of the product suite around that reality.

Google spent four months rebuilding AI Studio from scratch just to properly integrate vibe coding. That's not a side feature. That's a foundational shift in how they think about their entire product.

The convergence isn't companies cynically chasing the same market. It's companies independently arriving at the same conclusion about what AI actually unlocked. Not better chatbots. Not smarter autocomplete. The ability to turn intent into executable code, and therefore turn code into the universal interface for knowledge work.

Which raises the uncomfortable question: if code really is the foundation of all knowledge work, and AI can now write competent code, how much of the traditional knowledge work stack collapses?

I don't know the answer. But I know we're about to find out in real time, with a viciousness that comes from having no moats and infinite competition. The next few months won't be interesting—the next few years will be interesting. And possibly brutal.

For now, though, we have a bunch of fun new toys to play with. Given what's coming, I'm enjoying them while I can.

—Mike Sullivan