OpenAI's Greg Brockman: Why Human Attention Is the New Bottleneck

Podcast: Training Data | Participants: Alfred Lin, Greg Brockman

1. Key Themes

The Compute Business Is Simpler Than It Looks — And Infinitely Scalable

OpenAI's core business model is deceptively straightforward: buy/rent/build compute, resell at a margin. The demand side is essentially unlimited because intelligence itself is an unlimited resource. Greg Brockman framed it plainly:

"We buy, rent, build, compute, and we resell it at a margin. That's it. As long as the margin is positive, then you want to scale it. Because the demand for solving problems, the demand for intelligence, that's unlimited." [00:01:18:500]

This is a profound reframing — OpenAI isn't a software company or an AI lab in the traditional sense; it's an intelligence utility. The limiting factor has never been demand, only supply.

Scaling Laws Are a Deep Physical Truth, Not a Temporary Trend

Rather than treating scaling laws as an engineering trick that might plateau, Brockman frames them as something closer to a law of nature — empirical, durable, and not yet fully theorized.

"They feel deeply fundamental. It's like the scientific truth that just like you think about physics and Newton's laws and things like that, there's somehow this truth of the universe... As you pour more compute into the models, they get correspondingly more capable. And it just keeps going, there's no wall." [00:03:08:540]

This has significant investment implications: the compute infrastructure buildout isn't hype-driven — it's responding to a genuine, persistent law of returns.

Human Attention Is the New Scarce Resource

As AI handles the execution layer, the critical bottleneck shifts entirely to human judgment and attention. This is perhaps the most strategically important insight in the conversation.

"Human attention is going to be this incredibly scarce resource, right? The doing of things now is easy. Is this a good thing? Is this what I wanted? Is this aligned with my values, with my desires? That is going to become the single most important bottleneck. And so I think building systems that take that into account and really think about the human factor, like, that's the most important thing to do now." [00:16:21:350]

This reframes where value accrues in the AI stack: not to automation tools, but to systems that intelligently route, triage, and escalate decisions to humans.

2. Contrarian Perspectives

We Are ~80% to AGI Right Now — Not Decades Away

Against the mainstream view that AGI is a distant, abstract milestone, Brockman makes a surprisingly concrete claim.

"According to my view of where we are, I think we're about 80% of the way there, in that we have models that are smart. They're very capable... I think that they are just so capable. It's really remarkable. Like, does anyone here feel better at writing software than GPT 5.4?" [00:05:17:720]

The remaining 20% presumably involves autonomous agency, sustained goal pursuit, and real-world messiness — but the framing suggests we are in the final stretch, not the middle.

The Organizational Structure of Companies Will Fundamentally Break — Not Just Evolve

Brockman isn't suggesting incremental team restructuring. He's suggesting the entire management hierarchy construct may become obsolete.

"A lot of how we run organizations right now, there's almost only one way to organize large groups of people where you have teams, you have management structures, and you have scopes, and you have these hierarchies and all these things. Maybe that can change. Maybe you can be much more flat, small teams that can really just do incredible things." [00:14:00:190]

The contrarian implication: investing in or building traditional enterprise software that assumes hierarchical org structures may be building on a shrinking foundation.

Solopreneurs Will Be Able to Build Businesses That Previously Required Entire Companies

This runs counter to the conventional wisdom that scale requires headcount.

"I think that we're going to have this ability for solopreneurs to build very incredible businesses. And so anyone who has a vision, I think we'll be able to realize it." [00:13:32:690]

This has massive implications for venture capital portfolio construction — the next billion-dollar company might be a team of 3, not 300.

Context Deprivation — Not Model Capability — Is the Real Current Bottleneck

While most debate focuses on model intelligence, Brockman identifies a simpler, more immediate problem: AIs are being asked to help without being included in the meetings, decisions, and context that define the work.

"You have all these meetings, you didn't include the AI. You know, that's not very nice to the AI. Like, you're asking it to help you with things and it has no information. So I think really leaning into how do you make sure the AI even has enough information in theory to solve the problem." [00:09:07:470]

This suggests the near-term competitive advantage isn't model quality — it's context architecture.

3. Companies Identified

OpenAI

AI research and deployment company. Mentioned as the central subject — highlighted for being at the frontier of model capability, deploying Codex internally across functions (engineering, finance, sales, IT), and building Chronicle (a persistent memory tool). Brockman notes they "live in the future" and co-design tools with their own usage.

"One of the amazing things about being at OpenAI is you do get to live in the future, right? You do get to really see the shape of what's emerging and we can co-design, right? We can really change the models, the harness, everything together in order to better serve the needs that we see." [00:09:35:430]

Stripe

Global payments infrastructure company. Mentioned as a proof point of compounding infrastructure value — Brockman was employee #4 and first CTO.

"Stripe as employee number four and then the first CTO. I just recently heard that they process 1.6% of the global GDP." [00:00:01:840]

4. People Identified

Greg Brockman

Co-founder and President of OpenAI, previously first CTO of Stripe. Identified as an elite builder-operator who thinks at the infrastructure, model, and application layers simultaneously. His framing of human attention as the scarce resource and his hands-on use of Codex internally signal genuine technical depth combined with strategic clarity.

"I'm not sure if there's ever an official title, but I've been called many things. Let's just say that." [00:00:55:060]

Matt Garman

CEO of AWS. Mentioned in passing but with a striking data point about compute scarcity.

"I was just with Matt Garman. He says the GPU compute availability in 2026 rounds to zero." [00:01:44:960]

5. Operating Insights

Maintain Human Accountability as a Non-Negotiable Checkpoint in Agentic Workflows

OpenAI's internal policy for Codex — requiring a human to sign off on every code merge — is a practical governance template any company can adopt immediately.

"We still want a human to be accountable for all code that gets merged, right? So at the end of the day, is it a good thing to merge this piece of code? Is it well structured? Is it going to make our code base more maintainable? We want to make sure there's a human who is signing off to say yes." [00:10:04:830]

Deploy AI Vertically With Domain Experts, Not Horizontally With Generalists

OpenAI's internal rollout strategy — putting a small dedicated team deeply inside each function (finance, sales, IT) to understand domain needs before building — is a model for enterprise AI adoption that avoids shallow, unused deployments.

"We are also going vertical by vertical within OpenAI to adopt these tools within finance, within sales, within IT. And there we have a small dedicated team who's really deeply understanding the domain, working with the people who are the experts in it, in order to build skills, in order to modify the codex UI, whatever it is that is needed in order to get it to be good." [00:10:34:130]

Build Data Provenance Into Your AI Architecture From Day One

As internal knowledge bases become AI-queryable, permission systems break down — derived artifacts can expose data that the source document was later restricted. This is an architectural problem that must be solved proactively, not retroactively.

"You need to make sure you have some way of tracking through the system to say, well, this output document came from the source one. The source one is no longer accessible to this audience. Let's go in and validate that as well." [00:13:03:070]

6. Overlooked Insights

OpenAI Is Quietly Building a Customer Co-Design Program for AI-Forward Enterprises

Buried at the end of the Codex discussion, Brockman reveals that OpenAI is actively onboarding select enterprise customers to co-build and co-define these agentic workflows — and that this is an open invitation. This is not a widely publicized program, but it represents early access to the most powerful internal tooling OpenAI has.

"We are starting to work with certain customers as well. So for people who want to be very AI forward and want to be part of defining this revolution, that there's a place for that. And I'd love to talk afterwards." [00:10:34:130]

For operators and investors in the room, this is a non-obvious competitive moat: co-designing with OpenAI gives you both preferential access and the ability to shape the tools your competitors will later use.

AI-Driven Scientific Discovery Is Closer Than the Market Prices — Quantum Gravity-Adjacent Results Already Exist

Brockman briefly mentions a physics result so significant that serious researchers had considered the problem unsolvable — and OpenAI's AI produced a formula that physicists view as a step toward quantum gravity. This was mentioned in passing and received almost no follow-up.

"We had a physics result where our AI came up with this very beautiful formula that physicists who have been working on this for quite some time thought was totally impossible. Thought it was like maybe an unsolvable problem... It's real serious physicists who really view this as a step towards really being able to get to some sort of answer for quantum gravity." [00:26:29:690]

If this trajectory holds, the companies and research institutions that embed AI into scientific discovery workflows in the next 12-18 months — particularly in physics, materials science, and drug discovery — may capture disproportionate value before the broader market recognizes what is happening.