Artificial General Software
For decades, we’ve been waiting for artificial general intelligence. AI that can think like humans across any task. But while we debate when AGI will arrive, something just as important is happening right now: the rise of artificial general software. As language models get more powerful, we’re approaching a point where AI agents could replace most of the apps we use today.
The Problem with Today’s Software
Think about your workday. You probably use dozens of different apps. Slack for messages, Google Docs for writing, Asana for tasks, Figma for design. Each one works well enough, but switching between them constantly breaks your flow. You lose time and focus just remembering how each one works.
This fragmentation limits what we can do. We spend more time managing our tools than actually creating things.
There’s a principle in machine learning called the “bitter lesson”: simple, flexible approaches eventually beat specialized solutions. The same might be true for software. Just as general AI promises to outperform narrow AI, general software could make specialized apps obsolete.
From Clicking to Talking
Traditional software forces you to learn its interface. Want to edit a photo? You need to know where Photoshop keeps its tools. Need to manage a project? You have to master your project management app’s workflows. We adapt to software instead of software adapting to us.
AI changes this. Instead of clicking through menus, you describe what you want: “Remove the background from this image” or “Set up a marketing project with our usual timeline.” The AI figures out how to do it.
This is a fundamental change. When you can accomplish tasks by describing them in plain language, complex interfaces become unnecessary. Why learn software when you can just talk to it?
What This Means for Software Companies
This shift threatens how software companies make money. Adobe and Microsoft stay dominant partly because their software takes years to master. Once you’ve learned Photoshop, switching to something else feels impossible.
But when AI can understand and execute any request, these advantages disappear. Why pay for complex software when a general AI agent does the same work through conversation?
The value in software will shift from building features to creating AI that understands what users want and delivers reliable results.
The Transition Period
This change will happen gradually. First, we’ll see AI assistants inside existing apps, like GitHub Copilot for coding. These assistants will handle more tasks over time, making traditional features less important. Eventually the AI becomes the main interface, with menus fading into the background. Finally, specialized apps disappear entirely, replaced by general AI agents.
We’re already seeing this start. ChatGPT began as a chatbot but now handles writing, coding, analysis, and more.
New Challenges
Natural language interfaces are easier to use, but they create new problems.
With traditional software, you can see all available options in menus. With AI, you don’t know what’s possible. When you ask an AI to “improve this document,” what will it actually do? The boundaries aren’t clear.
Users need to understand what AI can and can’t do without getting overwhelmed by technical details. And when AI can perform any action based on your words, preventing mistakes becomes crucial. What if you accidentally ask for something harmful?
What Developers Will Do
As software becomes more general, developers will focus on different problems. Instead of building specific features, they’ll work on teaching AI to understand different domains better. They’ll make AI responses more accurate and reliable. They’ll build safety measures to prevent harmful actions. They’ll create ways for AI to explain what it’s doing.
The work shifts from coding specific behaviors to shaping intelligence itself.
Key Questions for the Future
Several important questions remain. Will visual interfaces disappear entirely, or evolve into displays that adapt to each conversation? How do we describe what an AI can do when its abilities change based on context? How do we protect data when AI agents work across all your information? What will software companies sell when anyone can build anything with AI? How should we teach people to work with AI instead of traditional tools?
The Path Forward
The rise of artificial general software represents a fundamental shift in how humans and computers work together. We’re moving from specialized tools to intelligent partners that understand our intentions.
This transition will be difficult. Some jobs will change dramatically. Privacy and security need rethinking. We need ways to ensure AI acts according to human values.
But the potential is enormous. Imagine focusing entirely on what you want to accomplish, without worrying about how to use your tools. Imagine software that understands you and improves over time.
We’re still in the early stages of this transformation, but the direction is clear. Software that understands natural language will gradually replace the complex interfaces we’ve grown accustomed to. The question isn’t whether this will happen, but how quickly we can adapt to working with AI that truly understands what we’re trying to accomplish.