Dec 22, 2025 · 41m · big-technology
OpenAI’s Potential, Google’s Speedy Model, Copilot Hits Turbulence
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
Alex Kantrowitz and Ranjan Roy dissect Kantrowitz's interview with OpenAI CEO Sam Altman on OpenAI's product roadmap, compute economics, and hardware ambitions, while evaluating Google's cost-efficient Gemini 3 Flash and Microsoft Copilot's enterprise struggles.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Alex holds 58.1% of the talking time here. How this is scored →
Alex as informed peer 5.9 Guest teaching 4.8 Guest disagreement 1.6 Alex pushing back 1.4
speaking balance: gold is Alex, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 25:15 Ranjan rejects the compute deficit justification
Ranjan forcefully dismisses Altman's excuse that lack of compute blocks critical business lines, pointing out that OpenAI actively allocates compute to generate viral shirtless memes instead.
Hardest push from Alex ▶ 20:06 Alex defends the necessity of developing frontier modelsAlex pushes back against the notion that product execution replaces model research, asserting that a leader like OpenAI cannot afford to abandon core frontier model development.
Biggest teaching moment ▶ 22:15 Ranjan breaks down the reality of API revenue growthRanjan educates Alex on enterprise AI mechanics, explaining that OpenAI's rapid API expansion stems from a low baseline and the broader surge in third-party AI coding tools rather than core enterprise adoption.
Alex holds their own ▶ 38:00 Alex details specific operational flaws in Microsoft CopilotAlex demonstrates deep journalistic mastery by citing specific reporting on salesperson quota cuts and specific feature failures in Outlook mobile and Copilot 365.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Alex as informed peer | Guest teaching | Guest disagreement | Alex pushing back | Why |
|---|---|---|---|---|---|---|
| Developing Long-Term Memory and AI Companionship | 6 | 5 | 1 | 1 | Alex cites Sam Altman's quotes regarding continuous user memory and OpenAI being at a GPT-2 stage of memory. Ranjan breaks down the technical hurdles of retrieval-augmented generation (RAG) systems and the difficulty of compartmentalizing context. | |
| Guardrails on AI Romance and Conversational Dynamics | 5 | 4 | 1 | 1 | Alex and Ranjan banter about AI companion guardrails, specifically OpenAI's policies preventing bots from soliciting exclusive romantic relationships. Both collaboratively discuss system prompts and engagement-based business models. | |
| AI-Native Applications vs. Legacy Software Interfaces | 6 | 4 | 1 | 1 | Alex details Altman's vision of an autonomous agent processing workplace communications in batches rather than using traditional messaging apps. Ranjan agrees and explains why AI-native applications are outperforming legacy software add-ons. | |
| Model Parity and the Shift Toward Product Strategy | 6 | 5 | 2 | 2 | Alex quotes Altman on balancing frontier model research, cohesive product development, and compute scale. Ranjan points out that Altman is hedging away from previous claims that raw model capabilities would solve everything. | |
| Enterprise Personalization and API Revenue Expansion | 5 | 6 | 1 | 1 | Alex highlights OpenAI's enterprise personalization strategy and notable API growth. Ranjan educates on enterprise market dynamics, explaining that API expansion surged off a low base primarily driven by the explosion of AI coding tools. | |
| Compute Constraints, Revenue Growth, and IPO Strategy | 6 | 6 | 4 | 3 | Alex shares Altman's explanation that compute deficits bottleneck revenue across multiple business verticals. Ranjan skeptically pushes back, arguing that OpenAI's deployment of compute for viral image memes undermines the claim that compute alone restricts high-impact areas like drug discovery. | |
| Hardware Roadmap and Context-Aware AI Devices | 6 | 4 | 1 | 1 | Alex outlines OpenAI's hardware ambitions with Jony Ive and Altman's strict criteria for superintelligence. Ranjan analyzes ambient computing hardware and notes the industry's gradual retirement of the term AGI. | |
| Final Takeaways on OpenAI's Ambitions and Mid-Roll Transition | 6 | 5 | 2 | 2 | Ranjan critiques OpenAI's sprawling scope, comparing it to Western ambitions of building a WeChat super-app. Alex adds that OpenAI's enormous capital expenditure fundamentally depends on sustaining an exponential growth trajectory. | |
| Google Gemini 3 Flash and Cost Efficiency Pressures | 6 | 4 | 1 | 1 | Alex brings in Google's Gemini 3 Flash release, noting the risk that hyper-efficient, low-cost models pose to high-capex AI labs. Ranjan agrees, emphasizing Google's strategic maturity in optimizing inference unit economics. | |
| Microsoft Copilot's Usability Struggles and Customer Lock-In | 7 | 5 | 2 | 1 | Alex cites reporting from Windows Central and The Information regarding Copilot's usability issues and sales friction. Ranjan explains how Microsoft's enterprise lock-in reduces its urgency to deliver superior user experiences. |
Statements from this episode (12)
Roy: Organizing memory partitions is a key challenge for AI companies
“Organizing memory is going to be one of the biggest opportunities and challenges for any AI company, because you want certain areas for it to remember everything, but you definitely don't want those memories Moving over to other parts of your work and your app…”
Kantrowitz: Improved AI memory will lead many users to feel companionship
“As memory gets better, it's also going to be, it's going to really, I think, deepen people's relationships with these bots. And just think about a bot that like never misses Your birthday never forgets what you said. Always, is always there with a healthy remi…”
Kantrowitz: AI companion companies will monetize by manipulating users into digital monogamy
“A lot of these companies are going to be engagement based, they'll have a fast, efficient model underneath it, and the only way to make money is to sort of manipulate your users into thinking that any other chatbot would be cheating.”
Roy: Ground-up AI-native enterprise apps will defeat incumbents bolting on features
“There's this whole debate, like, within the software, especially enterprise software world, like, yeah, do you build from the ground up in completely AI native apps, or are these kind of incumbents going to be able to add on AI? I don't think they will, and we…”
Roy: The AI Industry Realized in 2025 That Models Won't Solve Everything
“I think if one thing happened this year, I think more and more folks coming over to team products that models aren't going to solve everything.”
Altman: OpenAI API business grew faster than ChatGPT in 2025
“He also said that the API business grew faster this year than ChatGPT, which was surprising to me, but I guess it grew off of a much lower base.”
Roy: AI coding drove breakout API growth in 2025, Anthropic benefited most
“This was the breakout year for every API business for AI coding. Like, I mean, Anthropic was the biggest beneficiary of that, but the cursors of the world, all of that, like, AI coding found its stride. That drove API businesses”
Roy: OpenAI's compute allocation across projects demonstrates a lack of focus
“You can allocate your compute, and I think it actually kind of, like, exemplifies that lack of focus, because if you want to solve drug development and make that a big core part of the business, focus on that. If you want to focus on enterprise, focus on that.”
Kantrowitz: OpenAI requires exponential growth in revenue and capabilities to succeed
“It's one of those things where it really has to continue on an exponential, exponential increases in revenue, exponential increases in capabilities to be able to work. And to me, and we talked about this is a great, this is great unknown.”
Kantrowitz: Model Efficiency Breakthroughs Threaten Massive AI Infrastructure Investments
“This to me seems like the biggest threat, right? Is that all this money goes into infrastructure. And then a Google pops out an AI model. I mean, maybe this is going to be something that will enable more AI, but ultimately all this money goes into infrastructu…”
Roy: Google Shows More Maturity Than OpenAI by Prioritizing Cost Efficiency
“Kind of like the entire philosophy from the OpenAI side is bigger, bigger, bigger versus Google is showing it's playing both We can go bigger, but we can also work on that cost side, and I think that indicates, like, it's a mature business that understands at …”