AI Operating Systems

●10 min read●●

What will be the new entry point to the internet? We have all these substrates built up over 40+ years of internet. Which will survive, and which will be replaced?

In Pixar's film Ratatouille, Chef Auguste Gusteau championed that "Anyone can cook"—even a lowly street rat. Apple's vision was to enable anyone to create, and yet for the longest time, we've been forced to get advanced degrees and wield the laws of physics like magicians to make any technology a reality. Regardless of how far we've come, we haven't yet crossed the chasm.

Those tides are changing. For the first time we're on the edge of "Anyone can build" being true in both the digital and physical world, and who knows maybe we'll even extend this ability to animals so they can one day build what they need without needing opposable thumbs or an advanced cortex.

But, what will we need to get there?

Before We Build, We Must Rebuild

When thinking about the future, a lot of assumptions get made that we can simply extend what currently exists until eventually the old way becomes the new way. The problem is this just hasn't been historically the case, and it's not apparent that AI will be any different.

We need to rebuild — burn down the old way and tease out what could be once we're truly free of what is.

Even OpenAI and Anthropic, with all their advances on Canvas or Artifacts or computer use or chain-of-thought, fail to recognize the need to rethink their initial interaction design: instant responses. Instead of pausing, they forge ahead and double-down on immediacy.

It took courage—and a bit of theft—for the transition from MS-DOS to Windows. We didn't just keep extending a black and white terminal to inject a bit of user interface here and there while holding true to the core of what a terminal was meant to be. No, we blew away the terminal and vowed to eliminate the need for most people to ever know terminals even existed.

And yet here we are with everyone from indiehackers to Salesforce simply extending the software everyone is accustomed to—with the same old user interaction designs and patterns—hoping that we'll evolve enough to create a new 10x shift.

If there's a reason AI will take time to be an inflection point for the world, it's because we wasted valuable time extending instead of having the courage and creativity to rebuild.

The Silent Wars

Platform wars are often won before companies realize they've even begun. Even once they realize their mistake, most companies lose because they fight for the past while jaded and tired consumers opt for the future — a better alternative to their existing reality.

Instead of all-hands-on-deck urgency, companies drag their feet thinking there will be more time to capture mindshare, new interfaces, and new use cases. There was some truth to that when the only available compute was under your desk or on your lap but future operating systems will be cloud-based, leaving even Apple and Microsoft vulnerable.

Just like black swan events, the key mistake existing players make is to think future platforms will resemble current ones. In some way, they feel duty-bound to continue supporting the old way and their old customers' habits rather than freeing themselves to imagine what could be.

And yet during my short lifetime, we've already had many platform shifts that completely upended overconfident incumbents:

  • personal computers
  • operating systems
  • internet providers
  • bulletin boards
  • web browsers
  • forums
  • ecommerce
  • gaming consoles
  • music players
  • mobile phones
  • media, news & TV
  • app stores
  • social networks
  • transportation
  • food & groceries

The list goes on and on, always with a strong smell of hubris and hindsight that apparently large organizations never learn from. Rarely does this happen as a form of internal product cannibalism — it's always new players supplanting tired and lazy existing ones.

Very few people look at IBM or General Motors with the same desire and respect as they do Apple or Tesla, and one day that will be true of Apple and Tesla when they lose the will to reinvent.

The hidden truth most are blind to: all existing software—and most companies—will be replaced.

Operating Systems as Orchestrators

Existing operating systems & software applications operate completely in the blind, with a human guiding them to accomplish everything. Existing software is so inherently broken that we even sandbox applications from others such that virtually zero knowledge is shared amongst software on the same machine. Humans built the rails for all of this in ways that they think give them competitive advantage and security over other builders.

This will all fall away and be replaced.

Future operating systems will be virtual environments that start from templates but shape and evolve themselves as their individual users evolve. Even within the same team, different employees will use completely different interfaces into the same datasets & outcomes, allowing each person to operate in a way that is most optimized to them as individuals.

Each human will be backed by countless agents who themselves manage teams of agents & humans. Yes, agents managing humans.

These new interfaces won't be built exclusively by other humans; they'll be summoned by the end user themselves based upon what they need and how they need it. The operating system will assemble existing tools and even build entirely new ones as it scaffolds out from its day-one interaction with a new person and learns their preferences and needs.

It'll slowly learn to orchestrate their life, and do so in a way that frees them from waiting for an immediate (and often incorrect and short-lived) answer. Instead, it'll do what humans do: grok the world, reason about it, create a plan, get buy in from others, test and evolve its plan in small ways, and then scale up as it starts to see traction around progress and truth. Even once it's found traction, it'll continue to evolve solutions as the world shifts and people's needs change.

Instead of only building for people, these systems will also build for other agents' needs such that the total addressable market will grow far beyond what we can imagine today.

It'll also build trust in this way, living up to the old adage of "set it and forget it", because people will realize the future was never about immediacy, it was about correctness and quality over an extended time horizon. Current AI models regurgitate a guess; even modern search engines like Perplexity do very little to actually seek & verify for truth while trying to convince people they are somehow better than a human manually navigating Google for truth. What is Perplexity searching for exactly, if not the truth? Immediacy, it seems... and at what cost.

This is all because they lack a key element: an orchestrator.

Very simply an orchestrator helps you get from where you're stuck today to where you need to go tomorrow; and once tomorrow arrives, it adapts yet again. It deeply understands you and others—beyond the simple intent unlock that LLMs provide from human language. This understanding will stem from a collective context that expands even beyond your current job.

An orchestrator will be something you take with you from place to place, with each of us trying to get access to the very best orchestrators we can afford for both our professional and personal lives. It will be your personal chief of staff.

Freedom will mean something different in the future: the time to think at a higher register and the privilege to not have to think about the old monotonous way of work all the time, because that can be deferred now—for a price.

Just as we aren't preoccupied with being hunter-gatherers, we soon won't be preoccupied with being pattern workers. If it can be repeated, it can be learned and deferred to an orchestrator.

Legos All the Way Down, and Up

At the heart of every future operating system & agent will be a primary orchestrator. Beyond serverless, we'll be computerless and defer all actions to the cloud to orchestrate on our behalf.

But to get there, we'll need to enable AI systems to build themselves so they can build & evolve software for end users in a way that's never been done before—far beyond v0 or bolt or Cursor.

The way to do that is to teach them how to use common patterns & tools to accomplish a myriad of tasks in standardized ways while having the flexibility to personalize each to a given person's or team's needs.

The same way open source has enabled humans to collectively move faster, we'll soon decompose major platforms into recomposable parts that can be mixed and matched to solve both localized and global needs.

The advantage will shift from interface lock-in on tools like Slack or Gmail to be more about data access and APIs as the world increasingly and temporarily becomes more siloed off to fight off the impending raiders.

Instead of platforms, we'll have piece parts. All of those datasets and all the world's physical services and inventories will come with off-the-shelf components that can be used and evolved, often times with associated puzzle pieces that easily plug and play to accomplish standardized workflows and tasks.

If you play this out, entire companies will be replaced by composable systems, all backed by orchestrators (and humans), as we learn to do to software what we've already learned to do with electronics & hardware: make software truly composable.

Beyond making software truly composable, we'll find ways to make entire service providers and their systems composable in ways that will allow us to easily reimagine the world around our needs.

Instead of accepting the world as static and immovable, we'll utter our needs into the ether and watch as it shifts and learns and evolves before our eyes. We'll watch what used to take years be solved in days. Instead of crisis, we'll feel hope... because we'll all finally feel empowered to change the world how we collectively wish it to change.

And sitting in the middle of that will be an orchestrator. A new operating system built from the ground up to be the matchmaker between what a group of people need and whoever or whatever can deliver upon those needs.

Once we have orchestrators and composable legos, the world will start to change very, very quickly. All because...

Anyone Can Build

The best ideas are a sharp point that pierce the world and solve a problem for a niche audience before expanding to solve more and more needs for a larger audience.

All it takes is one person with the audacity to question the status quo and the courage to see change through to the end. A person with true belief.

And now, I believe it'll soon be your turn to build. All it will take is rethinking & rebuilding everything first.

-- Rob