Introducing the Dead Software Theory

Sam Altman’s recent comment on the dead internet theory got a lot of attention (not in a good way), and it’s what prompted me to consider how this idea could be taken even further.

The dead internet theory suggests that much of the web is no longer produced by humans but by bots, creating an empty, artificial experience. It’s a compelling idea that captures a feeling many of us have while scrolling through endless streams of generated content.

But what if this trend extends beyond content and begins to reshape the very tools we use every day? This brings us to my proposal of the dead software theory.

The theory builds on a simple observation.

Our primary way of interacting with software is about to change, big time.

The long traditional approach of logging into distinct applications, clicking through menus, and learning complex interfaces is drawing to a close. Instead, we will speak to conversational AI agents that operate these programs for us, in the background.

From Dead Internet to Dead Software

The dead internet theory gained traction because it felt true. We see it in the spam comments, the auto-generated articles, and the bot accounts that plague social media.

The data is starting to really support what isn’t a “conspiracy theory” anymore.

• Over 51% of internet traffic was automated in 2024, with malicious bots hitting a record high that has continued into 2025.

• The growth from AI crawlers is even more dramatic, with PerplexityBot traffic surging over 157,000%.

• OpenAI crawlers alone are generating more than a billion requests per month.

All of those facts were covered by Ignacio de Gregorio in a great Medium article here.

This trend is only accelerating with the rise of agentic AI, where autonomous systems can browse websites, fill out forms, and interact with digital services on our behalf.

This is the bridge to the dead software theory, where the same logic of background automation moves from content on the internet to the software tools we use to run our businesses and lives.

Software as we know it (discrete programs like Excel, PowerPoint, or Photoshop that require manual operation and know-how) is going to be a backend service for AI.

You won't need to open PowerPoint to create a presentation. You'll describe the presentation you want, and an AI will generate the [.PPTX] file for you, complete with slides, speaker notes, and data visualizations.

The software becomes a powerful engine that you no longer need to know how to drive directly. The interface is simply a conversation.

Conversational Interfaces Are the New UI

This isn't a distant future. It's happening now. Ethan Mollick recently pointed out a new capability in Claude that I hadn't realized was already live. Claude can now generate entire PowerPoint presentations and Excel spreadsheets from a simple prompt. Tucked away in Claude's settings is a beta feature that allows it to produce these complex files.

To test this, I gave Claude a slightly absurd prompt, building off of Dr. Mollick’s Hamlet prompt.

Create an investment banking-style merger proposal between the Montagues and the Capulets from Romeo and Juliet.

Claude generated a complete PowerPoint file, including a deal rationale, synergy analysis, and a pro-forma capitalization table. It created a plausible business case for a fictional merger from a 400-year-old play.

My first job was as an investment banking analyst back in the 90s, and believe me, this is not a bad pitch deck. I particularly appreciate the currency adjustments into Florins.

I’m only showing 6 of the slides, but it actually created twice as many.

I could direct the AI to make edits, refine the financial models, and adjust the tone, all through conversation. I didn't need to know how to format a single slide in PowerPoint.

I just had to know what I wanted. The software did the work, guided by the AI.

This capability is not an isolated trick. We see this pattern repeating across the industry. Model improvements often arrive in sudden, unexpected bursts of new functionality. Google's recent "Nano Banana" update in Gemini is another example of these step-change advancements. The entire software industry is moving in this direction, integrating conversational AI at a fundamental level.

More Powerful Than You Can Possibly Imagine

The dead software theory doesn't mean Excel is going away.

It means Excel is becoming a service that an AI can call upon. This is a profound shift in our relationship with technology. It’s reminiscent of Obi-Wan Kenobi's famous line to Darth Vader:

By "killing" the traditional software interface, we are unleashing its underlying power and completely removing the barrier of user expertise. You no longer need to be a spreadsheet guru or a presentation designer to get the full value from these tools.

You only need to be able to articulate your goal clearly.

This dramatically increases human efficiency by abstracting away the tedious mechanics of software operation.

In a time when many people are concerned about AI causing job losses, the dead software theory offers a more optimistic perspective.

AI can help bridge the gap for many people to leverage software systems efficiently. I hope that this leads to unprecedented entrepreneurship and efficiency gains.

Prepare for the Efficiency Boom

The transition to dead software will unlock immense productivity. When the friction of using complex tools is removed, people can focus entirely on strategy, creativity, and execution. The whole of human enterprise moves forward faster.

Start thinking of software as a capability, not a destination. Instead of asking "How do I use Excel to do this?" start asking "What is the best way to analyze this data?" The AI will handle the "how," selecting and operating the right tool for the job.

Your role is to define the "what" and the "why."

Master conversational delegation. Don't just experiment. Actively delegate one repetitive software task to an AI each week. Instead of building the weekly sales report in Excel, describe the report you need to Claude or Gemini and refine the output.

Build the core skills for the next generation of knowledge work and focus on using your voice effectively. I have often recommended Monologue.to for this, but you can start with simple Speech to Text on your devices.

Focus on outcomes, not processes. The value you provide will no longer be tied to your proficiency with a specific software suite. It will be measured by the quality of the outcomes you produce.

The dead software theory means that everyone will operate at a higher level, turning conversational ideas into reality.