Current Operating Systems are Dead in a Post AI World

The Operating System

The operating system evokes different reactions in people. Some view it as a work of art. A system that can self-repair and boot up instantly across millions of different permutations of hardware and software. Others see it as a gateway to possibilities they once only could be imagined. One thing is for sure, the modern operating system, whether on a laptop or phone is something that permeates every aspect of our lives, and is an exciting requirement or a necessary evil in today’s digital world.

The question is, how did the operating system saddle modern compute, and how will it change as it did from DOS in the 90’s, to GUI in the 2000’s and eventually to AI in the 2030’s?

Archaic Desktop Operating Systems

Modern compute was firmly solidified in many people’s lives starting in the 1980’s. The internet was a twinkle in most people’s eyes and was mainly used in research for large universities and in military uses.

Since then, the internet has become ubiquitous and continues to grow steadily, with no end in sight, adding several percentage points to its global user base each year relative to the world's population.

In contrast, the decline of modern laptop/desktop based computer was between 2010 and 2015 and has kept declining from there. Why? Mobile devices have taken hold over the method we connect to each other and to other tools and services.

As companies pivoted to capture the new market, companies focused more on mobile device operating systems than on desktops. Why would they invest millions in maintaining a fledgling operating system?

Desktop operating systems were built for an offline world, focused on local files and device independence. As mobile evolved for connectivity, desktops remained stuck, interacting with the world within a browser “operating system” atop an unconnected operating system, creating layers that don’t make sense.

Now that must change as AI enters the picture.

Archaic Mobile Device Operating Systems

In my mind, mobile device operating systems suffer from some of the same failures of desktop operating systems, focusing on an offline first device interaction. The difference here is that the apps are modularized to better take advantage and sit closer to the silicon than current desktop operating systems.

The main issue with most of these devices is in order to take advantage of being closer to the chip, you must pay the “app tax” of anywhere between 10-30% of your revenue to be accepted and distributed on the mobile devices. This tax is regularly challenged and avoided by making apps that are “dumb” wrappers, or forfeiting an app all together and forcing the user to interact with a web browser, once again, removing the advantage of the operating system, and using an inefficient operating system atop an operating system to access information.

Companies have focused on “apps” as the intermediary between how we work, and interact with devices. They focus their attention on making walls around their tools and products enhancing their bottom lines. They use heavily controlled APIs and create competing products with other companies that are basically the same product with the added benefit of being in their ecosystem, with a little bit of special sauce on how they handle the data, or have their sales pitch on improved security, or privacy or better interoperability, for a low price of $9.99/mo/user.

AI’s Pivot of the Operating System

For many, generative AI has opened up a whole new world on how interactions with the computer happen. What used to be an app or multiple apps is, in its current form, a chat window, with multimodal input and output. A user doesn’t need “an app” to transcribe the information to put into a Word document, to then be analyzed by yet another tool.

It is akin to when the iPhone came out, remember that segment, by Steve Jobs. “A phone, an iPod, an internet communicator”.

Artificial Intelligence represents that pivot point. Now the industry needs to create a modern operating system that can better take advantage of modern technologies, and create an outcome driven operating system rather than a process driven operating system that required intimate knowledge of how the system, tool or app worked to take full advantage of how it was designed.

To echo Jobs’ idea, an AI driven take might be “A assistant. A Brain. An Engine.”

Future of the Operating System

There are many leading technologies that will be the foundation of what a new operating system would look like. I want to focus on a few items that are already in the works, and some that I think will be foundational for an operating system driven by outcome, rather than by features.

Interconnectivity, or Model Context Protocol

In late 2024, Anthropic released the Model Context Protocol (MCP). MCP was supposed to solve an issue with the current use of AI. How do you get all your data into a massive model without copying and pasting into a chat window?

MCP is supposed to provide a shared interface or standard for an AI agent to connect to whatever format, whether its PDF, Excel, CSV, Word, Markdown, photos, voice, video etc. to provide a connection to all your information. Its goal is to unify access across data sources by interfacing with context rather than siloed applications.

MCP is, in my opinion, one of the building blocks of a new operating system. Instead of installing “the operating system” on a device to enable features that need to be installed, the MCP can interact with any part of the device or network securely without needing to set up “the environment”, allowing the operating system to act as an interface with the interconnected devices, tools, or products than with its own offline components. It will essentially become a network first operating system.

Unified Entity Context or Store Stuff in One Place

While MCP is the mechanism for retrieving the data, Unified Entity Context (UEC) is the content itself. Instead of thinking of data as what format its stored in, like CSV, XLSX, or markdown, we can focus on the content itself. I have an chart of all sales data over the last month.

You’re not going to care about individual parts of the data itself, you are focused on the outcome of the data. Are sales going up or down? How much is it down or up? What are the core products driving the sales? That information could (and should) be stored and accessed easily across tools.

Ideally, storage can now be managed in a sort of object storage, like MinIO, instead of formatted document, that is typically stored in a format readable by humans only because we have to evaluate it too.

With UEC, it could be in whatever information in whatever format, so you can focus on the outcomes rather than what and where the information is stored.

Designs for Outcomes, Not Skill Levels

I’ve talked about this one already here. But as interconnectivity on tools take over, the need for my task tool to also include an additional paid feature to store files, I could instead store information in whatever storage location I pay for, and have the task tool focus on being a task tool first, allowing developers to focus on ways to improve the tool, and more importantly get paid for features that create outcomes, rather than create features for problems that have already been solved, but need to now be implemented in their tool, so the tool feels “feature complete”. It also improves user adoption as they now don’t need to learn the way “that” tool saves data and the quirks it has, the user can focus on creating tasks with an AI knowing what it was for and what files I may need attached to it.

Outcome drives results, feature creep frustrates both advanced and basic users with “features they don’t need” or “features they wish they had” and focuses on the core issues. Getting stuff done.

Final Thoughts

The benefit with focusing on these items (and others), is it will change the way we interact with computers. Rather than using a mouse and keyboard to complete steps we have to remember, we could create outcomes on a phone, laptop, AR glasses, watch, whatever we have available to us.

No longer do we need such an archaic interface as an offline operating system to interact with windows and browsers to interact with information we use every day. A screen is now a screen that can do things, whether its a laptop or watch, or AR or VR headset. The AI, with you at the wheel is driving the outcomes, without the walls of self-proclaimed “best app for..” or “best tool for”… Just inputs, outputs and results.