Welcome to Stewart Squared podcast with the two Stewart Alsops, where they explore the evolution of software from 1.0’s “magical incantations” to 3.0’s natural language interfaces, discuss operating systems and their hardware roots, and unpack the significance of vertical integration exemplified by Apple’s silicon and software unification. This episode touches on large language models as cognitive prosthetics and the intimate, sometimes emotional, relationships people are forming with them, while also questioning their potential as operating systems for an “Internet-as-a-computer” paradigm. Alongside reflections on curiosity versus intelligence and the risks of Skynet-like scenarios, the Alsops weave in insights from Andrej Karpathy and Stephen Wolfram.

Check out this GPT we trained on the conversation

Timestamps

00:00 Software 1.0 to 3.0 evolution, natural language as programming, operating systems history with Apple II and early hardware tinkering
05:00 Open versus closed systems, Apple’s vertical integration, hardware limitations, early networking with Ethernet and modems
10:00 Distributed computing, Internet as a computer, LLMs as potential operating systems, differences between real-time systems and batch processing
15:00 Cognitive prosthetics, LLMs enabling new forms of software creation, emotional relationships with AI, sycophancy and “glazing” effects
20:00 Skynet fears, military applications of AI, robotics as physical extensions of AI, IoT devices and infrastructure vulnerability
25:00 Device Authority, over-the-air updates, challenges of retrofitting legacy hardware, enterprise resistance to innovation, IT culture dynamics
30:00 Curiosity versus intelligence, human adaptability, LLMs lack of intrinsic curiosity, future of AI-human collaboration, ending reflections on staying engaged with technology

Key Insights

1. The transition from software 1.0 to 3.0 marks a profound shift in how humans interact with machines, moving from cryptic programming languages to natural language interfaces that let anyone issue commands without technical expertise. This evolution democratizes programming but also raises questions about how much control we’re surrendering to systems we no longer fully understand.
2. Operating systems once served as the invisible backbone of personal computing, managing resources and hardware interactions in machines like the Apple II. Today, as computing shifts into distributed networks and cloud systems, the concept of an OS is becoming more abstract, raising the possibility that LLMs could function as an “Internet-wide operating system” in the future.
3. Vertical integration, as exemplified by Apple’s control over both hardware (Apple Silicon) and software (macOS), creates performance and efficiency advantages that competitors like Microsoft struggle to match. However, it also limits user freedom and reinforces a “walled garden” model that frustrates programmers who crave open systems.
4. Large language models are increasingly viewed as cognitive prosthetics—tools that augment human thinking, accelerate research, and enable non-programmers to build software. Yet their growing intimacy with users sparks debates about the emotional bonds people are forming with AI and whether these relationships fulfill or erode our social and emotional needs.
5. The Skynet metaphor highlights fears that AI could one day control critical infrastructure, but for now, the more immediate issue may be subtle—how LLMs shape human cognition, amplify sycophancy (“glazing”), and replace real human interactions with simulated ones that feel authentic but lack depth.
6. Curiosity, not raw intelligence, emerges as the defining trait for effectively engaging with new technologies. Unlike AI, which lacks intrinsic curiosity, humans have the ability to wonder and explore, positioning us as perpetual learners even in an age of rapid technological advancement.
7. The integration of IoT devices into legacy systems, as seen with companies like Device Authority, underscores both the promise and complexity of connecting the physical world to the Internet. Real-time updates and over-the-air security patches hint at a future where all devices are online, but this also amplifies vulnerabilities in critical infrastructure if AI systems gain too much autonomous control.