Apr 8, 2026 · 56m · big-technology
OpenAI vs. Anthropic's Direct Faceoff + Future of Agents — With Aaron Levie
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
Box CEO Aaron Levie joins Alex Kantrowitz on the Big Technology Podcast to analyze the escalating competition between OpenAI and Anthropic, discussing the infrastructure requirements, security challenges, and economic potential of autonomous AI agents across enterprise knowledge work.
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 19.4% of the talking time here. How this is scored →
Alex as informed peer 5.4 Guest teaching 5.3 Guest disagreement 3.0 Alex pushing back 3.4
speaking balance: gold is Alex, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 17:33 Premise rejection on wasteful compute allocation
Aaron flatly refuses Alex's dystopian framing of algorithmically dominated creative production, arguing compute will naturally be directed toward high-ROI sciences rather than endless podcast variations.
Hardest push from Alex ▶ 16:18 Challenging the necessity of human editorsAlex directly disputes Aaron's claim that humans must remain in the editorial loop, proposing that multi-agent voting and automated audience testing will replace manual curation.
Biggest teaching moment ▶ 25:00 The enterprise context gap masterclassAaron uses the analogy of a brilliant new employee with zero institutional knowledge to educate Alex on why enterprise agent deployment is fundamentally an unsolved data and permissions problem.
Alex holds their own ▶ 53:28 Pushing back on cloud market analogies with compounding returnsAlex challenges Aaron's thesis that the market will simply lift all labs equally, arguing that compounding lead time in capital and compute creates winner-take-most dynamics.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Alex as informed peer | Guest teaching | Guest disagreement | Alex pushing back | Why |
|---|---|---|---|---|---|---|
| Enterprise LLM Adoption and Knowledge Work Agents | 6 | 5 | 4 | 3 | Alex provides a detailed framing of the rivalry between OpenAI and Anthropic across enterprise and consumer domains. Aaron gently reframes Alex's premise by noting that ChatGPT gained massive organic enterprise adoption independent of API sales before laying out the expansion from coding agents to general knowledge work. | |
| Economic ROI and Subjective Task Verification | 5 | 6 | 3 | 4 | Alex questions the practical utility of agentic tasks like video editing, invoking discussions with Greg Brockman. Aaron explains why coding tasks are far easier to automate due to instant verifiability compared to subjective, context-heavy knowledge work. | |
| Enterprise Adoption Hurdles and Bridge Platforms | 4 | 7 | 2 | 2 | Aaron delivers an in-depth breakdown of the friction enterprise knowledge workers face when adopting agents, contrasting Silicon Valley assumptions with enterprise data sprawl. Alex listens as Aaron outlines the market opportunity for bridge platforms. | |
| Multi-Agent Architectures and Human Editorial Roles | 6 | 4 | 4 | 6 | Alex challenges Aaron's timeline by pointing out existing automated video-switching heuristics and proposing multi-agent voting architectures. Aaron counters that human editors will shift from manual assembly to higher-level synthesis and directorial judgment across multiple agent drafts. | |
| Parallel Agent Compute in High-Impact Industries | 6 | 4 | 5 | 5 | Alex pushes a thought experiment about a hyper-optimized, algorithm-driven world across all knowledge work. Aaron explicitly rejects the premise for trivial media tasks, arguing compute economics will prioritize high-stakes fields like drug discovery and financial modeling. | |
| Mitigating Confirmation Bias and LLM Monoculture | 5 | 4 | 2 | 2 | Alex raises the issue of LLM monoculture and sycophancy, which Aaron illustrates with personal anecdotes regarding parenting prompts and Andrej Karpathy's findings on bidirectional LLM justifications. | |
| The Enterprise Context Gap and Data Infrastructure | 5 | 7 | 2 | 3 | Alex asks whether users will realistically cede control and access to agents. Aaron contrasts modern startups with legacy enterprises, explaining that AI performance is bottlenecked by fragmented corporate data architecture and missing tribal context. | |
| Agent Security, Prompt Injection, and Legal Liabilities | 5 | 5 | 2 | 2 | Alex expresses personal hesitation about granting autonomous inbox access to agents, citing prompt injection pranks. Aaron outlines sandboxed security practices before highlighting unresolved enterprise liability and regulatory precedents. | |
| Unified Agent Modalities and Box Agent Architecture | 5 | 5 | 3 | 3 | Alex asks whether lab investments in agents might be misplaced compared to basic chatbots. Aaron argues agentic workflows exist on a single compute-accuracy continuum, detailing Box Agent's speed versus precision trade-offs. | |
| Value Capture: Horizontal Labs vs. Vertical Applications | 6 | 6 | 3 | 4 | Alex presses Aaron on whether foundation model labs or applied vertical wrappers will capture the ultimate enterprise value. Aaron evaluates the 'bitter lesson' perspective against historical vertical SaaS defensibility. | |
| Next-Generation Frontier Models and Scaling Progress | 6 | 5 | 2 | 2 | Alex details upcoming frontier model releases based on his reporting with Greg Brockman. Aaron confirms Box's internal evaluation benchmarks are recording double-digit performance leaps, refuting the AI wall narrative. | |
| Cloud Infrastructure Parallels and Podcast Conclusion | 6 | 6 | 4 | 5 | Alex repeatedly presses Aaron to pick a winner between OpenAI and Anthropic. Aaron sidesteps using a historical parallel to the 2008-2010 cloud infrastructure wars, while Alex pushes back on the compounding advantages of early market leadership. |
Statements from this episode (23)
Levie: AI Labs Will Inevitably Compete Head-to-Head Across All Use Cases
“You know, I think it's, to some extent, it was sort of an inevitable outcome because if you think about it, like if you have this AI model that is super intelligence packed into a model, it eventually has to converge on On, you know, all of this, all the same …”
Levie: AI Is Shifting From Chatbots To Multi-Day Autonomous Agents
“And so I think we're clearly moving from a world where you will use AI as this thing you chat back and forth with. And that was kind of the first manifestation of the chat bot to now a paradigm where the agent is given a task. It has a set of resources it has …”
Levie: Knowledge work AI market is 30 to 50x larger than coding alone
“So this is kind of the big prize because it goes from the TAM The total addressable market being, you know, all of engineers to now the total addressable market is every knowledge worker, and that's probably about a 30 to 50 x larger market in terms of, you kn…”
Levie: Enterprise AI ROI will far surpass consumer use as tokens stay expensive
“The tokens are not going to be cheap anytime soon. And so the ROI on those tokens will just be much higher in the enterprise because it'll be, you know, generating something that is sort of, you know, impacts the GDP in some way.”
Levie: Automating video editing with AI is harder than coding due to evals
“Editing a video is like, you know, going to be actually in many cases, a harder task than coding. Because the, because again, the code right now is like, it has this great property of in the eval process, in the training process rather, you can instantly evalu…”
Levie: Enterprise AI agent adoption will take many years
“So all of those things add up to actually just meaning that, that the diffusion of these types of technologies will take many, many years as they go through the rest of the world.”
Levie: Most workplace employees will run AI agents in coming years
“But I think that I would expect most people have agents running in their daily life from a workplace standpoint over the coming years, just because the efficiency will just be too strong to kind of avoid.”
Levie: AI agents will turn video editors into senior review directors
“But I actually think that that you'll still have that ultimate person. Maybe what they'll review is five different cuts as options. And they are now playing the role of the, you know, the most senior editor in a, you know, TV show that, that, that, that, that …”
Levie: AI agents will increase life science experiments by 10x to 100x
“What we're gonna now be able to do is we will be able to run, you know, on the order of 10 to a hundred times more experiments across, you know, everything that we want to go detect.”
Levie: Society will learn over time how to prompt AI without bias
“And that, that'll be, I just think that'll be like a thing we generally learn over time in society, just as we eventually learned how to use search engines and other tools.”
Levie: Prompt AI for pros-and-cons tables to prevent sycophancy
“In general, what you really want is,
The, as much as possible, you want the agents to do things like, generate me a table of the pros and cons of this thing.
Right.
And make sure that you make arguments for both sides, and then you want to be really in the …”
Levie: Enterprises face years realizing AI problems are data infrastructure problems
“And so we are gonna be in for, again, years and years of enterprises realizing that an AI problem is really a data problem, and to get the AI the right data, They need to make sure they have infrastructure, software, tools, systems that all are in service of g…”
Levie: Having documents across 20 to 30 systems will not work with agents
“Where we come from in our industry of, you know, with enterprises managing enterprise content, companies have 20 or 30 different systems where their enterprise documents are, and that just simply won't work with agents.”
Levie: AI agents should have separate inboxes, not personal access
“The common practice and sort of state of the art is, is effectively don't give open claw or something access to your inbox. Create a separate inbox for the agent. And really treat that agent as a, another colleague that you're working with. And so it has its o…”
Levie: AI labs will not accept liability for enterprise use cases
“The labs are not going to, you know, take on the liability for every single use case that you do. They're gonna have very narrow liability that they have around copyright and IP protection and stuff like that, but they're not gonna, you know, they're not gonna…”
Levie: Extensive new case law for AI agents will emerge
“So in finance, in healthcare, in legal we have just incredible amounts of updated laws that will have to get written and case law that will be That will be generated over the coming years.”
Levie: AI systems have a strict trade-off between speed and accuracy
“There is one thing in AI that, that is is just like there's just no free lunch which is that you can have something fast, like insanely fast, but like moderately accurate or pretty accurate and insanely slow. Right. And like, you just get to choose.”
Levie: Vertical SaaS Proved $50B Companies Win Over Horizontal Tools
“Even in, you know, kind of Traditional SaaS software. We saw 30, 40, fifty billion dollar vertical software companies emerge in categories where there was already plenty of horizontal products that could have solved those problems, but just that relentless lev…”
Levie: AI Labs Win in Every Scenario as Foundational Intelligence Layer
“The good news is there's going to be value in, in both sides, because even the vertical domain specific players will be riding on top of the intelligence from the horizontal labs. And so in both, in all the scenarios, the labs win, you know, a very big prize l…”
Levie: Thousands of Successful Products Will Emerge at AI Application Layer
“I think there'll be hundreds of successful, thousands of successful products at that layer, simply because again, enterprises, they just want to, they want to wake up. They want to get their job done. They want to have some alpha relative to competitors. And t…”
Aaron Levie: AI Development Is Nowhere Close to Hitting a Wall
“I think certainly probably the biggest takeaway is just like we are nowhere close to hitting a wall. I remember it was probably only about a year ago where there was a lot of talk on like, oh, have we hit a wall? And these things are only kind of eking out, yo…”
Levie: Box Benchmark Shows Double-Digit LLM Gains in Four Months
“We have an eval that we give all of the new models. It's basically a complex knowledge work task, which is we give the, an agent a set of documents to work with, and then we ask it a series of very, very hard questions that we think correlate to pretty high en…”
Levie: Any AI Lab's Lead on Breakthrough Models Is Only 6 to 12 Months
“Unless there's some so kind of closed proprietary research event and breakthrough that happens that just simply nobody else knows about, and we have no evidence that we've ever had one of those in AI, like, you know, these things just eventually sort of emerge…”