Jul 11, 2026 · 1h 0m · news
Why OpenAI and Anthropic Won't Win the App Layer | Glean Founder · 20VC with Harry Stebbings
Clips from this episode (5)
Short vertical cuts produced from the tape, captions burned in. Where a cut lands on a statement from the ledger, its card says so.
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
In this episode of 20VC, Glean founder Arvind Jain sits down with host Harry Stebbings to discuss why specialized application platforms will outpace general frontier models, the major corporate shift toward open-source AI, and the surprising economic realities and high hidden costs of deploying AI agents in the enterprise.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Harry holds 25.2% of the talking time here. How this is scored →
Harry as informed peer 5.6 Guest teaching 5.0 Guest disagreement 2.8 Harry pushing back 5.1
speaking balance: gold is Harry, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 29:45 Arvin rejecting headcount reduction premise
Arvin forcefully rejects Harry's assertion that CEOs should cut headcount, stating 'I absolutely don't believe in it' and arguing that larger teams using AI will outperform smaller ones.
Hardest push from Harry ▶ 19:14 Harry refusing consumption breaking bundling premiseHarry directly rejects Arvin's claim that consumption pricing negates Microsoft's bundling power, citing complex enterprise vendor sign-off dynamics across major legacy corporations.
Biggest teaching moment ▶ 33:26 Arvin revealing $1M/month AI agent triage billArvin stuns Harry by revealing Glean spent $1,000,000 per month on a single internal engineering triage agent, demonstrating the extreme and questionable economics of current AI agents.
Harry holds his own ▶ 19:14 Harry demonstrating deep enterprise procurement knowledgeHarry displays sharp domain knowledge of enterprise IT procurement, naming specific giant corporations to prove why vendor consolidation beats consumption flexibility.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Harry as informed peer | Guest teaching | Guest disagreement | Harry pushing back | Why |
|---|---|---|---|---|---|---|
| Opening of the Interview: The Paranoid Founder | 2 | 2 | 1 | 1 | Harry introduces Arvin Jain and asks introductory questions about founder mindset and a summary of Glean. The conversation is warm and collaborative. | |
| Enterprise Skepticism & Operational Dependence | 4 | 5 | 2 | 4 | Harry cites Alex Karp's CNBC comments regarding enterprise skepticism. When Arvin starts to drift into AI efficacy, Harry interrupts to steer him back to whether enterprises should fear frontier model dependence. | |
| The Shift to Open Source & Cost Controls | 5 | 5 | 3 | 5 | Harry asks directly about investor fears that Anthropic will cannibalize Glean just like Figma or legal tech. Arvin reframes Anthropic's vertical expansion as shallow and market-expanding rather than cannibalistic. | |
| The Advantage of Being First to Market | 5 | 4 | 2 | 4 | Harry asks about first-mover advantage and probes whether model layer commoditization is accelerating due to Chinese open-source options. Arvin agrees on 90%+ use case commoditization and details how Glean manages model costs for customers. | |
| Geopolitical Paranoia: Chinese Models in Western Enterprises | 5 | 5 | 2 | 6 | Harry presses Arvin on why enterprises would hesitate to use Chinese open source models on-premise if data stays private. He also questions whether frontier model valuations are completely mispriced. | |
| Microsoft Copilot & the Bundling Threat | 7 | 4 | 3 | 8 | Harry explicitly rejects Arvin's claim that consumption pricing breaks Microsoft's bundling power, citing complex enterprise vendor approval processes and listing specific mega-corporations like VW, Ford, and Tyson Foods. | |
| The Quest for ROI in Enterprise AI Spend | 5 | 6 | 2 | 4 | Harry prompts a discussion on enterprise AI ROI citing Alex Karp. Arvin educates Harry on the distinction between increased coding speed and actual feature shipping speed. | |
| Stringent Code Reviews vs AI Speed | 6 | 6 | 3 | 6 | Harry challenges the logic of strict code reviews, arguing it negates the speed gained from AI code generation. Arvin explains the hidden costs of context assembly and long-term code maintainability. | |
| Replacing Labor with AI & the Future of Teams | 8 | 6 | 6 | 8 | A heated debate occurs over team headcount. When Arvin states Glean plans to grow from 1,000 to 5,000 employees, Harry strongly dissents, citing conversations with major CEOs who are shrinking headcount to focus capital on elite tech and engineers. | |
| The True Cost of AI Agents: A $1 Million Example | 8 | 7 | 5 | 8 | Arvin drops a shocking figure that an internal triage agent cost Glean $1M per month, leading Harry to joke about hiring Cristiano Ronaldo. Harry forcefully counters Arvin's belief that tech costs shouldn't replace labor costs, arguing that software replacing human labor is the entire premise of tech. | |
| Token Budgeting & encouraging AI Adoption | 5 | 5 | 1 | 3 | Harry asks how Glean manages internal token budgets and references Nikesh Arora's leadership practices at Palo Alto Networks. Arvin outlines the power-law distribution of internal token usage. | |
| Recruiting Challenges in the Age of High AI Salaries | 6 | 5 | 2 | 4 | Harry breaks down startup seed math given elevated developer salaries ($500k/year requiring $6M rounds). Arvin reflects on valuation optics from Glean's Series C and the talent market. | |
| Shifting Mindsets: Discipline vs Aggressive Land Grab | 6 | 5 | 3 | 6 | Arvin admits his natural financial discipline might be holding Glean back during an aggressive market land grab. Harry catches a logical contradiction, pointing out that Arvin's advocacy for generalized composite roles naturally implies smaller team sizes. | |
| AI Model Sovereignty and Open Source Ecosystem | 8 | 6 | 5 | 7 | Harry cites OpenRouter usage statistics showing Chinese models outperforming US models and speculates whether Sam Altman would offer equity/promises to Trump for regulatory protection against open source. Arvin repeatedly rejects Harry's cynical political framing. | |
| Quick Fire Round and Founder Realities | 4 | 4 | 2 | 3 | In the quick fire round, Harry and Arvin discuss CS degrees, top investors, startup capital overabundance, and founder stress. Harry asserts that CEOs must accept being perpetually unhappy. |
Statements from this episode (45)
Jain: Glean combines ChatGPT, Claude, and Gemini into one product
“Today the way to think about Glean is that first, it's a super set of JetGPD, Cloud, Gemini, All of those combined into one product experience.”
Jain: Institutional enterprise learning will accumulate inside AI agents
“In the future, all of that institutional learning is actually going to accumulate in that agent that is doing that work.”
Jain: Not owning AI agent learnings creates total vendor dependence
“If you don't have any control on running that agent yourself, if you don't own the learning that it actually that actually, like, you know, gains over the years, then you're basically fully dependent on these AI companies, you know, to do, you know, to get you…”
Jain: Enterprises are actively routing away from frontier models to open source
“That's something that's happening now. So I think we are at a real inflection point with open source.”
Jain: Enterprise open-source AI adoption is driven by cost, not data privacy
“I think right now the open source drive is coming from the cost point of view.”
Jain: Enterprise fears over AI models training on company data have vanished
“When AI just came, there was a, companies were a lot more afraid of getting their data outside of, you know, their own control and model companies training with, you know, with their data. But that sort of is a fear that it's no longer there. Like, you know, l…”
Jain: Anthropic's vertical AI feature packs are shallow
“They are launching these sort of vertical packs, but I think they're quite quite shallow in my opinion.”
Jain: Information seeking is the world's largest AI application today
“That's like the largest application or use case for AI in the world today. Is in fact information seeking and question answering.”
Jain: Glean was the first enterprise AI company to deploy RAG
“We actually get a lot of credit for being the first enterprise AI company in the world. The first ones to actually bring RAG into the enterprise. The first ones to build conceptual semantic search.”
Jain: the model layer is definitely commoditizing for most enterprise use cases
“90% or greater of use cases cannot be fully handled by many, many different models, including open source models. So there's definitely there's definitely commoditization from that perspective. In fact, like, you know, we are clean that's actually one of our c…”
Jain: Enterprise AI adoption hurdle is Chinese model origin, not open source
“From a point of view of open source and using the model, everybody's going to be fine. The question is going to be, are they okay with the Chinese model or not? That's the only question here. It's not open source versus closed source.”
Jain: Enterprise adoption of Chinese AI models will become normal
“And the early movers will make the move first and then it will become a more normal thing.”
Jain: Standalone AI model business is not as lucrative as expected
“I think the model business, you know, on its own is, is actually probably not as lucrative as everybody believes, but these companies now have a lot more things.”
Jain: Anthropic is an application-level platform, not just a model vendor
“So they very much, you should consider them an application level company, not just a model company.”
Jain: Majority of enterprise AI workloads will be open source within three years
“Well, we've been telling customers, I believe that majority, majority of enterprise workloads will actually be on open source models in three years for sure.”
Jain: Glean loses more sales to Microsoft Copilot than to frontier labs
“If we look at our experience with as we go and prospect the, I think we would hear this answer more often that, well, like, you know, we are a Microsoft customer and we already have, you know, we're getting co-pilot. And so, therefore, like, it doesn't make se…”
Jain: Majority of current enterprise AI spend is focused on coding
“I think the majority of the AI spend right now is on coding.”
Jain: AI coding tools have not accelerated enterprise product shipping speed
“The actual shipping speed of products has not increased, even though coding, you know, increased, coding speed increased significantly.”
Jain: Nearly 100% of initial code at Glean is AI-generated
“Now it's probably about almost a hundred percent. Like, nobody's actually, like, writing the initial code, you know, by hand anymore.”
Jain: Engineering bottleneck has shifted from writing code to reviewing it
“The real bottleneck has shifted from The person writes the code to personalize the review.”
Jain: Massive AI-generated codebases become unmaintainable over time
“When you write code, for example, with AI you know, we, you know, you can write like, you know, a million lines of code but it becomes incredibly hard to actually maintain it and understand it and manage it over time.”
Jain: AI cannot fully replace a single job role today
“I think it's a wrong goal in my opinion to say that, hey, like replace yourself with AI. First of all, I think you're giving too much credit to AI in, you know, when you say that. It's just not ready right now. There's, you give me a name of one job that you c…”
Jain: Human plus AI teams will always outcompete pure AI automation
“I think the, you want to be performing the best in whatever you do. And I don't think you're going to take a 90% solution. Well, I mean, like, look, the they're kind of like, you know, like, it's a, it's not about you not being cost-constrained. It's about you…”
Jain: Glean aims to grow headcount to 5,000 within five years
“Well, hopefully, you know, 5000. We're gonna grow.”
Jain: AI-amplified companies producing 10x output will beat those shrinking headcount
“If you were trying to do the same amount of work and you believe you can do it with fewer people and therefore you shrink your competition can also do the same, but they chose actually not to do the same amount of work. They chose to actually, you know, elevat…”
Jain: The world's greatest companies will not operate with only 100 employees
“And the, I don't think the world's greatest companies are going to be companies with a hundred people”
Stebbings: AI developer tools like Cursor remain dramatically underpriced
“So no, I don't at all for Cursor or for any of the dev tools. I think it's still dramatically underpriced. When you look at Mark Benioff spending three hundred million on Anthropic, it's 3.7% of developer salaries. I think that's relatively small.”
Jain: Glean's triage agent handles 95% of issues at an absurd $1M a month
“I would say it's absurdly expensive. I'll give you an example. We had this, we have a really cool triage agent for engineering. You know, and we have 15 people team on call team that was supposed to, that their work was to actually triage Every single producti…”
Jain: Open-source AI achieves equal results at one-tenth the cost
“But I also know that already you see with open source that you can do the same amount of same work for a 10th of the cost, right?”
Jain: AI inferencing costs will drop by orders of magnitude
“I think good technologies figure out how to make technology really, really cheap, and it's gonna happen here too. That's my belief. Like, you know, you're going to see it. You're going to see inferencing costs come down by orders of magnitude.”
Jain: Enterprise AI token spend follows a power law
“There's a power law, like, you know, in our company, and also at all of our customers, you will see some people who spend 10,000 dollars or 15,000 dollars in tokens every month, and then you have others who are spending 20 dollars.”
Jain: AI talent pay scales have dramatically surged across startups and labs
“But if you now start to talk about, okay, AI talent, ML talent top people are sought after, you know, way more than ever before. And also the pay scales have completely changed and not just from the model companies, but even from startups, because, you know, s…”
Jain: Glean never proactively raised funds after its first round
“So first, actually, we never actually went out to race you know, except for our first round of the company, you know, we always had somebody come in, and it was a relationship that got built over some time, and kind of became the de facto, like, you know, that…”
Jain: Glean raised Series C at $1B+ valuation on sub-$5M revenue
“I would say, like, our series C probably felt the most I guess you could say the most expensive, because we barely had any business, like, definitely, like, you know, sub two or three million dollars, maybe five million dollars, I don't remember exactly, but t…”
Jain: Shifting Glean from capital discipline to aggressive spending
“I've always personally, like my style has been a little bit too disciplined to be the right strategy anymore. I get that feedback from my team that, you know, we are trying to be conservative. We're trying to make sure that our capital goes a long way. And in …”
Jain: Securing enterprise AI clients later will be 10x harder
“We are absolutely in a land grab, like, you know, no question. Like every single company in the world wants a product like ours today. And either we get in today or it's going to be like 10 times harder to actually get in the future.”
Jain: Composite job roles will replace specialized roles within five years
“Well, the comp, comp composite roles will be will be very common. So like, as for example, you know, somebody who can build a product I don't know what to call them, but they can, they act like engineers, product managers, designers similarly in go to market Y…”
Jain: Tactical data analyst and recruiting sourcer roles will disappear
“A lot of analyst roles, like, or like the data analyst roles, which are not business thinkers, you know, they were given a task that, hey, like, I need to see this data, and then they produced, like, they sort of go and build those specific dashboards, configu…”
Jain: Only China and France have produced AI models outside the US
“The only Country in the world, you know, that has produced models outside of US is China. And then of course, maybe a little bit in, like, you know, France with Mistral.”
Jain: Open-source developers cannot build AI models without massive upfront funding
“They require a lot of upfront investment, which is not open source friendly in many ways. A lot of open source software has been skunk works, like developers is getting no funding associated with them and they still get something built. They couldn't build mod…”
Jain: Nvidia is heavily investing in US open-source AI model development
“There's a lot of motivated parties that actually want to promote, like, including NVIDIA, for example, You know, they're putting a lot of investment in promoting, like, you know, development of great open source models in the US.”
Jain: Excess startup capital creates failure paths and bad habits
“I actually do think that, you know, there is too much capital available today in the for startups. And it's actually sometimes creating failure paths for people. I think they're not getting what it takes to build a great company.”
Jain: too much capital pushes startups into unsustainable structures
“I actually do think that, you know, there is too much capital available today in the for startups. I like, I'll give you an example, like a startup that has raised a seed round, decides to pay half a million dollars to an engineer, like you were saying before …”