The Rise of AI Operating Systems - Building Blocks for AI-Native Startups

A new concept is emerging that promises to change how businesses operate: the AI Operating System (AIOS).

This blog post will explore what an AIOS is, where this idea originated, and how it could reshape the future of startups and enterprises alike.

The Current Startup Operating System

Let’s acknowledge the messy reality most startups face today. Running a company often means navigating a chaotic maze of SaaS tools, with founders staring at bloated spreadsheets packed with subscription fees. Often, no one even knows who’s using what, leading to overspending.

The more tools you add, the harder it gets to manage. Some businesses try consolidating with all-in-one platforms like HubSpot or Atlassian, but there’s still no single solution for everything. Instead, they end up piecing things together using tools like Zapier or Slack connectors—building an integration nightmare that only adds to the chaos.

The Vision of an AI Operating System

The AI Operating System: This idea took off last year when Andrej Karpathy, known for his work with Tesla and OpenAI, shared a bold concept on Twitter.

Karpathy’s vision looked like a literal operating system—large language models as the CPU, a context window as RAM, and peripherals handling inputs and outputs. In essence, AI becomes the core computing unit. But he didn’t stop there. Karpathy pushed this idea further, imagining an entire company run by AI systems, each playing a distinct role within the organization.

This idea echoes Sam Altman’s thoughts on a “10-person unicorn,” where AI amplifies human potential. He even speculated about a “one-person unicorn,” where a CEO is the only human, and AI takes care of the rest.

Initial Examples from Enterprise World

In a recent conference call, Klarna’s CEO, Sebastian Siemiatkowski, revealed significant changes in the company’s approach to software as a service (SaaS) providers. Klarna has already discontinued its use of Salesforce and plans to phase out Workday.

Siemiatkowski highlighted that these changes are part of larger internal initiatives focused on leveraging artificial intelligence, standardization, and simplification. “We are shutting down a lot of our SaaS providers as we are able to consolidate,” he stated.

This strategic move aims to streamline operations and enhance efficiency, positioning Klarna for future growth and innovation.

AI-Native Applications Paradigm

We're already seeing the emergence of AI-native applications. These are not just existing tools with AI tacked on, but entirely new ways of approaching tasks and processes.

For example, we're seeing AI-enhanced developer co-pilots, synthetic users for marketing, and AI-generated content for sales. These point solutions are becoming more interconnected, creating suites of AI-powered tools that work seamlessly together.

A concrete example of what AI-native applications might look like comes from a recent demonstration using Gemini 1.5 Pro. A user recorded a 21-minute workout video on their phone and asked the AI to analyze it. The AI was able to track exercises, identify weights used, and even provide form critiques for each exercise. This gives us a glimpse into a future where AI could serve as a personal trainer, analyzing your workout in real-time and providing instant feedback.

Just like software scaled to become a near-zero cost for giants like Airbnb and Uber, we’re on the brink of something even bigger: intelligence becoming a marginal cost. Instead of investing 18+ years and hundreds of thousands of dollars to develop a human mind, companies will soon tap into AI that rivals or surpasses human intellect—on demand and dirt cheap. Think about that. Endless copies of human-level smarts at a fraction of the cost. The game’s about to change.

Components of an AI Operating System

To understand how an AIOS might function, let's break down its key components:

• APIs and Data Sources: These form the foundation, connecting various data inputs and outputs.
• General AI Systems: Large language models like GPT serve as the central "brain" of the system.
• Narrow AI Systems: Traditional machine learning and deep learning models handle specialized tasks.
• Tools and External Software: These augment the AI's capabilities for specific functions.

In this ecosystem, humans interact with these components at a higher level of abstraction, focusing on tasks, insights, and decision-making while the AI handles the underlying complexities.

Building an AI-Native Startup: A Framework

So, how can startups and enterprises prepare for and capitalize on this AI revolution? Here's a framework based on experience working with AI startups and enterprises undergoing AI transformations:

1. Research: Stay informed about AI advancements, especially those relevant to your domain. Even if it feels like you're "wasting" resources, R&D in AI can lead to valuable IP.
2. Discovery: Conduct workshops to understand the intersection of human work and AI capabilities in your business. Map out key roles, processes, and tasks, then explore how AI could transform them.
3. Development: Start building AI-enhanced products and features. Begin with off-the-shelf solutions where possible, but focus on developing custom AI applications for your core business domains.
4. Scaling: Integrate AI across your business operations, continuously refining and expanding its use based on results and new capabilities.

The AI Stack for Startups

As you build your AI-native startup or transform your existing business, consider this AI stack:

1. Infrastructure and Compute: Leverage cloud credits and services rather than building your own GPU clusters.
2. AI Models: Use a mix of commercial and open-source models, both general (like GPT) and narrow (specialized for specific tasks).
3. AI Ops and Development Tools: Build reusable code pieces and leverage existing platforms and frameworks.
4. Safe AI Practices: Invest early in tools and frameworks for responsible AI use, considering data protection and user safety.
5. AI Assistants and Agents: Develop AI systems that can act as assistants or even autonomous agents within your business processes.

The rise of AI Operating Systems represents a paradigm shift in how we conceive of and run businesses. From transforming startup operations to potentially enabling one-person unicorns, the impact of AI on the business world is just beginning to be felt.