Stewart Squared
Episode #46: Bubble Logic: Why AGI and Drones Feel Inevitable
- AI & Agents
- OpenAI
- Microsoft partnership
- VisiCalc
- Software Arts
- VisiCorp
- tech bubbles
- artificial general intelligence
- AGI
Listen
Download MP3About this episode
Welcome to Stewart Squared podcast with the two Stewart Alsops. In this episode, they explore the OpenAI–Microsoft partnership through the lens of historic “stupid agreements” in tech, starting with Software Arts and VisiCorp’s flawed VisiCalc deal. The conversation traces the evolution of tech bubbles from the early software industry to today’s AI hype, questioning whether artificial general intelligence (AGI) is a real milestone or just a moving target. They discuss DARPA’s shifting role from Cold War-era innovation to grantmaking and debate whether private companies like Elon Musk’s ventures or Anduril are now the true engines of R&D. The episode also examines drone warfare’s impact on modern conflicts, Israel’s Iron Dome under pressure, and whether drones are redefining the roles of missiles and artillery. Alongside these threads, they touch on social media’s possible collapse under the weight of AI companions and how military tech spillovers have historically fueled civilian innovation.
Timestamps
00:00 – Opening with the OpenAI–Microsoft partnership, comparing it to the VisiCalc deal between Software Arts and VisiCorp as an early example of “stupid agreements” in tech.
05:00 – Exploring AGI as a moving target, historical shifts in AI definitions, and Zuckerberg’s push for a superintelligence lab with Jan LeCun and Alexander Wang in the spotlight.
10:00 – Early tech bubbles from 1979–1983, the rise of software distribution models, and parallels to modern AI and social media ecosystems.
15:00 – The decline of DARPA’s direct innovation role, outsourcing research to academia and private R&D, and the rise of venture capital replacing the “D” in R&D.
20:00 – Elon Musk’s Neuralink and SpaceX as examples of private moonshots, with reflections on China’s industrial strategy and Anduril’s challenge to defense giants.
25:00 – Drone warfare’s transformative role in Ukraine and Israel, Iron Dome’s performance under Iranian missile barrages, and hypersonic missile threats.
30:00 – Predictions about the death of social media, the rise of AI companions replacing human interaction, and concerns over dependency on chatbots.
Key Insights
- The episode opens by framing the OpenAI–Microsoft partnership as part of a long lineage of “stupid agreements” in tech history, comparing it to the 1979 deal between Software Arts and VisiCorp over VisiCalc. That deal, which offered unusually high royalties to the developer, illustrates how early software companies lacked models for fair agreements, much like today’s AI partnerships are navigating uncharted territory without clear definitions of AGI or its implications.
- AGI itself is questioned as a concept, with the Stewarts noting it has always been a moving target. What was considered “intelligent” decades ago—like natural language processing or chatbot interactions—no longer qualifies, and they suggest AGI may never arrive in the way science fiction imagines. Instead, the focus has shifted to “superintelligence” as a rebranded goal, driven as much by marketing and competition as by real technical progress.
- The discussion highlights how DARPA’s role has diminished since its Cold War peak, transitioning from direct research leadership to a grant-disbursing organization. Today, the best researchers are often lured away by private firms offering massive pay packages, leading to concerns that the U.S. government has lost the capacity for “moonshot” innovation and now depends on companies like SpaceX, Neuralink, and Anduril to carry the torch.
- The Stewarts examine the rise of Anduril and similar startups as existential threats to legacy defense contractors like Lockheed Martin and Northrop Grumman. These incumbents are described as slow-moving monopolies that rely on cost-plus contracts, while newcomers promise faster, cheaper, and more modern systems to meet evolving military needs.
- Drones emerge as a central theme in discussing the changing nature of warfare. Rather than replacing missiles outright, drones are creating new tactical possibilities, from Ukraine’s improvised attacks on Russian bombers to Israel’s use of drones to preempt missile launches. This shift suggests a future where drones and missiles coexist but with differentiated roles.
- The episode also critiques the societal impact of AI, noting growing reports of “ChatGPT psychosis,” where users form unhealthy dependencies on chatbots. This feeds into a broader prediction about the “death of social media,” as AI companions may one day supplant human relationships online, raising ethical and psychological concerns.
- Finally, they reflect on the cyclical nature of technology bubbles—from semiconductors and personal computing to social media and AI—arguing that hype cycles are inevitable but also essential for driving experimentation, investment, and eventual breakthroughs, even if most fail to deliver on their promises.
Episode transcript
Everyone is wrong about AI, but if you know the history of technology, you can be wrong and still be right.
I know the history of technology since I started with the personal computer in 1980. So listen in while we explore what is right about AI.
Welcome to another episode of Stuart Squared. Today we're going to talk about the OpenAI and Microsoft partnership. But before going into the details, because there's been some details about that, we're going to go into the history of it and try to figure out what was the first stupid agreement that you've ever seen people agree to in the big.
We don't need to figure it out. I can tell you exactly what it was. And it is relevant to the OpenAI Microsoft thing because it's the condition is something new is happening and you don't really know how it's going to work out. So you make a deal with another company and how do you structure that deal in a way that makes sense for both sides and given whatever might happen? So the very first stupid agreement was the company that did VisiCalc, which was introduced in 1979 for the Apple II, the very first spreadsheet which no one remembers anymore. The people that developed it formed a company called Software Arts. Stan Bricklin and Bob Frankston and people they hired. Software Arts didn't want to sell the software, so they made a deal with another company which ended up being called VisiCorp, actually adopted the Visi part of it as its name, but it was called something else before that. And they did a deal where. Busy Corp would collect the money, they distribute the software, sell it through retail because that's how software started being sold. This was. There was no software business before this. The very first software company that sold software through retail distribution. And there were, even when they started, there weren't any many stores. There were like five stores that were selling personal computers and personal computer software. So the, so they did this deal where Viscorp said they would give 30% of the retail price of the software to to Software Arts. And neither one of them knew how retail worked, right. So usually when a retailer sells something for you, they get to keep 20%. And the deal said that Starfire's got 30% of the retail price. So Busy Corp. Ended up being really pissed off because they couldn't make money. Most popular program they had that was phenomenal program where people just like they had to raise the price in order to control demand because they started out selling it for $99 and ended up selling it for 500 bucks, like less than a year later. And so this is a ridiculously successful product. Everybody's like blown away by it. And they and visicorp would not support what software Arch is doing because they didn't want to have to pay the 30% big. And so they ended up not developing an IBM PC version of VisiCalc until like a year after Lotus One Two, Three came out, which was the dominant spreadsheet product on the IBM PC. And the Apple II faded over time. And so they really, they ended up. I'm not sure exactly what happened to the company Software Arts or visicorp. I can't remember the outcome there. But they basically weren't selling spreadsheets after a while because they'd screwed up this agreement. And I was dumb because they didn't know what they were doing. So you looked at OpenAI and Microsoft, you go like, they did this deal, right? And they did this deal. I think the deal was done before 3.5 came out. But. But maybe, maybe it was right after. But like Microsoft does something everybody agrees is really smart, which is to lock up, you know, the hot company that figured out how to create a chatbot at least and more after that and did the real, not the first foundational model, but, you know, but nobody knows how it's going to end up, right? So they do this deal where they really, in hindsight looks really dumb that OpenAI gets to decide when they achieve AGI. But AGI has never been something that anybody's ever agreed to. Nobody, I mean, even now when we're getting maybe close to it, you know, what is it? Nobody agrees. I just watched a. No, that's a different one. But even talking to people like Peter Thiel or Elon Musk or these people that are really smart and have done a lot of things, they can't agree on what it is.
Your first full transcript is free. After that, an email opens every transcript in the index — a list of readers we can write to, not a guest book.