There are many big questions about AI, but as a tech industry person, the one that is most interesting to me is “will AI eat the tech industry, or will the tech industry eat AI?”

What would it mean if AI ate the tech industry? OpenAI or Anthropic wouldn’t just be providers of chatbots or video generators—they’d become operating systems and ecosystems, the primary interface for accessing all information. Microsoft did that with Windows, Google with search (and advertising, and Google Drive, and Android), and Apple with the iPhone. 

Some people really do spend hours a day on ChatGPT already, and use it to do many of their daily tasks—so in some cases, yes, AI is eating the tech industry. Sam Altman wants his own blockchain currency, his own social network, his own developer tools, his own hardware platform (with Jony Ive), and more. He clearly wants to remake the world around this technology and be the main driver (so does Elon Musk, who wants “X” to become the “everything app”, which, thank God, is not happening). 

By hyper-scaling tech-world standards, the statement “AI-powered interfaces could be the dominant paradigm for every interaction with a computer” is more or less reasonable. Could it come to be? Probably not, but it’s not a zero-percent chance, either, so everyone is going to keep swinging in order to see if they are the ones who can take over the world.

And then, by contrast, there’s MCP. MCP stands for “Model Context Protocol.” It was created by Anthropic, the makers of Claude, and it’s open-sourced. It’s supposed to provide a standard way for chat apps like Claude Desktop to talk to old-school APIs—databases or web platforms like Shopify or GitHub. 

This means that you can use AI as the interface to those services. Which is very “webby” and old-school. You could, for example, ask a particular store on Shopify to find T-shirts with funny slogans, or GitHub to summarize all the changes in a given codebase since the last major release—things that were possible before, but not in plain language in a couple of minutes. You used to need engineering support, but now you can connect to a database and say “show me the trends in fidget-spinner sales” and Claude could convert that into a big nasty SQL query and then summarize the results for you.

MCP is kind of old world—it’s a standard for interchange between one kind of system (LLM-powered) and another (often database-powered). It’s a compromise that lets the new world, which embraces ambiguity and can process natural language, interact with the old world, which is driven more by classic data management, and is more focused on accurate results. And recently, OpenAI announced that they, too, would offer very full support for MCP.

If both OpenAI and Anthropic decide to support something, that’s that. Anil Dash likens MCP to “Web 2.0, 2.0.” Perhaps that’s optimistic—there isn’t, for example, a standards body that collaborates across vendors, driven by a public standards process—but it’s also progress, too. Things slowing down, talking to one another, behaving in a more neighborly fashion—that’s how everyone gets to play, instead of just a few vendors. Standards aren’t just nice-to-have; a good standard, like HTML, or MIDI, or ECMAScript, or GAAP, or even ISO 20700, provides a lot of opportunity for smaller players to innovate and build good practices. You can build a career around a standard. It’s less exciting than taking over the world and owning everything, but it’s less stressful, too.

To me, MCP doesn’t represent the second coming of Web 2.0 as much as a quiet acknowledgement, in some corners of the AI world, that AI might not eat the tech industry. People might just have to content themselves with market caps of hundreds of billions of dollars, not tens of trillions. AGI might turn out to be a bust, or at least much farther away. And in the meantime, computers are going to keep doing computer things—browsing the web, checking in code commits, removing redeye, helping you order pants. LLMs bring a lot of main-character energy to the tech industry, but they might find themselves as supporting actors after all.