Sep 22, 2025 · 54m · big-technology
Revealed: OpenAI Tells Us ChatGPT's Major Use Cases
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
On the Big Technology Podcast, Alex Kantrowitz and Ranjan Roy analyze newly released empirical usage data from OpenAI and Anthropic to uncover how consumers and enterprises actually use generative AI. They examine key trends across chatbot use cases, corporate research marketing, Meta's smart glasses rollout, and the broader decline of traditional web and broadcast media.
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 57% of the talking time here. How this is scored →
Alex as informed peer 5.6 Guest teaching 4.3 Guest disagreement 3.1 Alex pushing back 3.2
speaking balance: gold is Alex, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 3:06 Roy rejects Kantrowitz's distinction between guidance and search
Ranjan directly interrupts Alex to dispute his claim that practical guidance represents a novel use case separate from legacy search engines, citing Howcast and early 2000s SEO strategies.
Hardest push from Alex ▶ 3:46 Kantrowitz refutes Roy's media-aggregation comparisonAlex firmly counters Ranjan's counterexample, arguing that legacy content sites were superficial SEO creations whereas generative AI delivers emergent, tailored real-time guidance.
Biggest teaching moment ▶ 35:40 Roy breaks down per-capita statistical distortion in D.C.Ranjan demonstrates how Anthropic's formula for per-capita usage is distorted by a few concentrated federal enterprise contracts in a geographically small jurisdiction.
Alex holds their own ▶ 8:18 Kantrowitz reveals the strategic motive behind OpenAI's agent definitionsAlex demonstrates sharp domain expertise by connecting OpenAI's inflated 'doing' categorization directly to their need to validate the agent-centric release of GPT-5 despite user preference for thought partnership.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Alex as informed peer | Guest teaching | Guest disagreement | Alex pushing back | Why |
|---|---|---|---|---|---|---|
| Practical Guidance vs. Search: Decoding ChatGPT's Core Use Cases | 6 | 5 | 6 | 6 | Ranjan explicitly challenges Alex's assertion that practical guidance is distinct from traditional search by citing historical content farms like Howcast and eHow. Alex pushes back vigorously, arguing that generative AI provides an emergent, tailored technological solution rather than SEO-driven aggregated articles. | |
| Asking, Doing, and the Disconnect in Agentic AI | 7 | 4 | 3 | 4 | Ranjan critiques OpenAI's definition of 'doing' as mere drafting rather than true autonomous agentic execution. Alex extends this observation with deep analysis, suggesting OpenAI broadened its definition to justify pivoting GPT models toward agentic capabilities despite user demand favoring thought partnership. | |
| Research as Content Marketing and Demographic Trends | 4 | 6 | 4 | 2 | Ranjan deconstructs the OpenAI white paper's format, explaining how academic styling serves as covert content marketing to boost business narratives around demographics and democratization. Alex readily agrees and contributes supporting context on corporate messaging strategies. | |
| AI Companionship and the Reality of Romantic Roleplay | 6 | 5 | 2 | 3 | Alex and Ranjan calculate the true volume of romantic roleplay messages behind OpenAI's seemingly small 1.9% statistic, translating it to tens of millions of interactions weekly. The exchange is highly collaborative and analytical as both hosts joke about token burn on flirty responses. | |
| Coding Saturation and the Shift Toward Non-Work Usage | 5 | 3 | 2 | 2 | The hosts discuss Claude's high coding concentration compared to ChatGPT's shift toward 70% non-work usage. Both align on market positioning without significant conflict, agreeing that consumer dominance remains OpenAI's primary narrative objective. | |
| Decision Support, User Trust, and the Death of the Open Web | 6 | 4 | 3 | 3 | Alex discusses the unmeasured economic value of decision support and recalls Cloudflare CEO Matthew Prince's insights regarding user trust and the bypass of web links. Ranjan supplements with observations on hallucination limits and the secular decline of the open web. | |
| ChatGPT's User Growth Trajectory and Mid-Show Break | 5 | 2 | 1 | 1 | Alex outlines ChatGPT's explosive growth curve from 200 million to nearly 800 million weekly active users over 2024. Ranjan compares this adoption curve to the early iPhone rollout before they transition into a mid-episode commercial break. | |
| Anthropic's Economic Index and Historical Tech Adoption Rates | 6 | 4 | 3 | 3 | Alex reviews Anthropic's Economic Index comparing AI adoption velocity to historical breakthroughs like electricity and personal computing, recounting Sundar Pichai's 'fire' remark. Ranjan questions the validity of comparing lightweight software diffusion to capital-heavy physical infrastructure. | |
| Claude's Strategic Focus: Autonomy, Education, and Public Sector Adoption | 6 | 5 | 3 | 3 | Alex notes the surge in autonomous tasks and high per-capita Claude usage in Washington D.C. Ranjan contextualizes the metric by explaining that concentrated federal contracts in a smaller population area naturally distort per-capita figures. | |
| Automation vs. Augmentation in Enterprise Workflows | 5 | 4 | 2 | 3 | Alex scrutinizes the distinction between automation and augmentation in Anthropic's report, noting that automating isolated subtasks often simply augments worker capacity. Ranjan agrees, highlighting the technical limitations of Anthropic's current enterprise connectors. | |
| Meta's AI Glasses Rollout and the Live Demo Dilemma | 5 | 4 | 5 | 5 | A friendly disagreement emerges over Meta's new display smart glasses. Alex expresses hesitation about intrusive head-mounted screens and compares the hype to early Apple Watch promises, while Ranjan defends the heads-up display paradigm based on his testing of Snap Spectacles. | |
| Jimmy Kimmel's Hiatus and the Death of Traditional Broadcast Media | 6 | 5 | 3 | 3 | The hosts analyze Jimmy Kimmel's broadcast hiatus within the framework of modern media distribution. Alex cites Jeff Jarvis to emphasize the structural death of traditional broadcast television, while Ranjan details how network affiliate economic models have collapsed in favor of viral clip distribution. |
Statements from this episode (21)
OpenAI: 75% of ChatGPT Use Focuses on Guidance, Info, and Writing
“Three quarters of conversations focus on practical guidance, seeking information, and writing, with writing being the most common work task, while coding and self-expression remain, remain niche activities.”
Kantrowitz: ChatGPT's Practical Guidance Is a Brand New Category, Not Search
“I think there was a misconception that AI was going to be just like search redone, a refactoring of search, and it's probably led by publishers who are like, where's our traffic going? But the fact that practical guidance is number one, where you can go to, an…”
Roy: OpenAI categorizes ChatGPT messages as 49% asking, 40% doing, 11% expressing
“They had three patterns within messages, asking, doing, and expressing. About half of messages, 49% are asking and then expressing, doing is 40%, expressing is 11%.”
Kantrowitz: OpenAI inflated 'doing' stats to conceal lack of agent usage
“I think what we're seeing in the numbers here is the vast majority of activity within ChatGPT is really thought partner. It is a thought partner tool, and the agentic stuff, of course, is super early, but they didn't want maybe the numbers, and this is maybe s…”
Kantrowitz: AI companies cannot serve companion, thought partner, and agent concurrently
“And that to me is, I think it's a problem, and that's why I think these companies are gonna have to have clarity over which one they're going for. You cannot really do all three. Maybe you can do it in the same product, but you definitely need a switcher.”
Kantrowitz: OpenAI reports only 1.9% of ChatGPT messages involve relationships
“Only 1.9% of Chachi BT messages. Are in the topics of relationship, of relationships and personal reflection.”
Roy: ChatGPT processes 18 billion messages per week
“Oh, actually, if you look at it, it's eighteen billion messages per week. So I just did, that's thirty-five million relationship messages right there per week.”
Kantrowitz: Coding is 4.2% of ChatGPT messages versus 33% for Claude
“Another interesting stat, they say four, 4.2% of chat GPT messages are related to computer programming compared to 33% of work-related Claude conversations.”
Kantrowitz: ChatGPT consumer usage is 70% non-work and 30% work-related
“30% of consumer usage is work-related, and approximately 70% is non-work.”
Roy: Web degradation and SEO spam drive users toward AI chatbots
“I think just so much of the web actually became less trustworthy or just overly SEO optimized or just unusable that people are actually just excited that there's something that cleanly gives you an answer that's at least seemingly or is fairly accurate, so.”
Kantrowitz: ChatGPT Pacing to Hit 800M to 900M Weekly Users in 2025
“This year, it's on track to hit eight hundred million weekly users by the end of the year, maybe even more, maybe nine hundred million.”
Anthropic Report: 40% of US Employers Now Use AI at Work
“In the US alone, Anthropik says, 40% of employers report using AI at work, up from 20% in 2023, two years ago.”
Kantrowitz: AI Reached Internet's Five-Year Adoption Rate in Two Years
“Even the rapidly adopted internet took around five years to hit adoption rate that AI has reached in just two years.”
Kantrowitz: Coding Is Claude's Top Use Case at 36%
“Coding is still number one use case at 36%. But educational tasks have surged from 9.3% to 12.4%, and scientific tasks are now six point, they're, they went from 6.3% to 7.2%.”
Anthropic Report: Autonomous Tasks Delegated to Claude Jumped to 39%
“Directive conversations where users delegate complex, delegate complete tasks to Claude have jumped from 29, 27% to 39%.”
Anthropic Report: Washington D.C. Leads Per-Capita Claude Adoption
“The city that, or region that had the most clod adoption per capita is Washington, D C.”
Roy: Claude's Connectors Are Not Good Enough for Scaled Agentic Workflows
“Claude has some connectors where that allow you to do stuff with other systems. I've tried them. They're not. Great right now. So I cannot imagine that at any kind of scale usage, like people are building these like complex agentic workflows using it.”
Roy: Subtle Augmented Computing Layer in Smart Glasses Will Be Massive
“So to me, this is gonna happen. This is the direction. I think the screen, again, if you're watching reels while crossing the street, that's not good. If you have, like, a light bit of information in the twenty-degree field of view in the far right of your len…”
Roy: Meta Is Winning the Wearable Computing Interface Race
“That's what, that's exactly how I've seen this whole interface of computing for a long time, and they are, they're winning. They're winning right now, and they're moving in the right direction as well.”
Kantrowitz: Zuckerberg's voice prompt crashed Meta's live demo servers
“Apparently Zuck said, hey, Meta, start live AI. It activated everybody's AI in the room, and effectively DDoSed the servers. All of the servers were all, all the traffic was also routed to the company's servers in Iceland, apparently. So, that's why it broke.”