Used by humans, built for AI

It’s the start of a new year, and I've been thinking about something that's been quietly changing in the tech world. It's a pattern I've started noticing everywhere, even though it was hidden in plain sight at first. Once you see it, it's hard to ignore. The pattern you ask? we're increasingly building apps for AI, not just for humans.

For years, we've built software with us in mind. Think about how websites are designed or how we use programs on our computers. Everything was made so we could easily understand and interact with it.

But things are shifting. Now, we're seeing a new trend where applications are being designed so Large Language Models (LLMs) can use them effectively. This means the app is designed to "speak" the language that AI understands.

Think of it this way: Instead of just designing things for human use, we're designing them to be used by AI, which will then interact with us.

Two examples of this shift are gaining popularity:

• The Model Context Protocol (MCP) by Anthropic: This is a new open-source standard designed to help AI connect to the systems where data lives. As Anthropic puts it, it's about helping "frontier models produce better, more relevant responses" by connecting AI assistants to "content repositories, business tools, and development environments." Essentially, it's a way to give AI the context it needs to provide truly useful answers. Whether you're building an AI-powered IDE, enhancing a chat interface, or creating custom AI workflows, MCP provides a standardized way to connect LLMs with the context they need.
• The "llm.txt" proposal: Instead of a tool or library, this is a proposal to create a standard file called /llms.txt on websites. Think of it as a special information sheet just for AI. This file provides key details about how LLMs can best use the website's content during the inference process – making it easier for AI to extract and utilize information correctly.

It used to be that we lived in a world where software gave us the same answers for everybody. We all had the same "view of the cave", like in Plato's allegory. But now, AI acts as a translator and an intermediary. The answers you get in the future will depend on the context, and on how well the LLM understands your intent and persona.

The world we’re building isn't just for us anymore, it’s for AI too. It’s designed to be understood by these powerful models which then provide an interface between us and the applications we use every day. This subtle, but important shift is transforming the tech landscape.