Apr 1, 2026 · 1h 13m · big-technology
OpenAI President Greg Brockman: AI Self-Improvement, The Superapp Bet, Path To AGI, Scaling Compute
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
OpenAI President Greg Brockman breaks down OpenAI's strategic pivot toward a unified super app, the economics of scaling compute, autonomous agent workflows, and the near-term timeline for achieving artificial general intelligence.
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 21.4% of the talking time here. How this is scored →
Alex as informed peer 5.7 Guest teaching 4.3 Guest disagreement 2.3 Alex pushing back 4.8
speaking balance: gold is Alex, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 59:32 Rejection of Dario Amodei's compute risk critique
Greg directly and forcefully rejects Anthropic CEO Dario Amodei's accusation that OpenAI is recklessly 'YOLOing' infrastructure bets, arguing OpenAI planned ahead while competitors scramble for compute.
Hardest push from Alex ▶ 45:24 Holding Greg to his quote on losing pulse on problemsAlex refuses Greg's pivot to general accountability, interrupting to restate Greg's exact admission about losing granular touch with underlying work when using fleets of agents.
Biggest teaching moment ▶ 9:49 Architectural unification within GPT transformersGreg corrects the assumption that OpenAI's image generation relies on separate diffusion tech trees, explaining that their multimodal work runs natively inside the core GPT transformer architecture.
Alex holds their own ▶ 9:07 Deploying Hassabis world model counterargumentAlex challenges OpenAI's core bet on text reasoning by citing DeepMind CEO Demis Hassabis's view that image and video generation represent the true path to spatial world models.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Alex as informed peer | Guest teaching | Guest disagreement | Alex pushing back | Why |
|---|---|---|---|---|---|---|
| Episode Preview: AGI Timeline, Underdog Mindset, and Compute Scale | 5 | 4 | 2 | 4 | Alex opens the interview by questioning OpenAI's strategic shift away from video generation towards a super app despite winning in consumer markets. Greg clarifies that resource prioritization demands focusing on personal assistants and hard problem solving rather than spreading compute across too many tech branches. | |
| The Tech Tree Dilemma: Reasoning Models vs. Video Generation | 6 | 5 | 3 | 5 | Alex uses Greg's past Disney analogy to ask whether OpenAI can still leverage a single core model advantage across diverse products. Greg explains that Sora and GPT reasoning represent fundamentally distinct branches of the tech tree that cannot share compute easily. | |
| Betting on GPT Reasoning Over World Models | 5 | 5 | 2 | 4 | Alex asks why OpenAI favors text/reasoning models over video and world models that showed rapid generational progress. Greg explains that deep learning works across all domains, but GPT text intelligence has proven line of sight to AGI and scientific breakthroughs. | |
| Architectural Unification and the Breadth of General Intelligence | 7 | 5 | 3 | 6 | Alex cites Demis Hassabis arguing that image and video generation are closest to AGI because they must understand spatial physics. Greg acknowledges the trade-off but clarifies that OpenAI's image generation is integrated into the GPT transformer architecture rather than separate diffusion pipelines. | |
| Architecture and Vision of the OpenAI Super App | 4 | 4 | 1 | 3 | Alex asks for the definition and technical scope of OpenAI's upcoming super app. Greg explains it unifies ChatGPT, Codex, computer-use browsing, and agent harnesses into a single interactive layer. | |
| Personal Use Cases and Rollout Strategy for the Super App | 4 | 3 | 1 | 2 | Alex inquires about consumer utility and the expected shipping timeline. Greg outlines how memory and contextual awareness will transform personal tasks and explains the staged rollout beginning with Codex. | |
| Lowering Barriers for Non-Developer Codex Adoption | 6 | 4 | 2 | 5 | Alex shares a concrete anecdote of a non-developer building Adobe Premiere plugins with Codex and asks why Anthropic established an earlier lead in unified coding apps. Greg concedes that OpenAI historically neglected last-mile real-world developer usability in favor of benchmark competitions. | |
| Maintaining Underdog Discipline and Company Culture | 6 | 3 | 2 | 5 | Alex cites internal reporting about OpenAI ending exploratory 'side quests' in response to tightening competition. Greg shares that his scariest moment was post-ChatGPT complacency and emphasizes retaining an underdog discipline. | |
| The "Spud" Model and OpenAI's Training Pipeline | 6 | 5 | 2 | 4 | Alex asks about reports of OpenAI finishing pre-training on the next-generation model dubbed 'Spud'. Greg details the multi-stage training pipeline and explains how increasing model capabilities unlock open-ended domains like complex cancer research. | |
| Autonomous AI Researchers and the Takeoff Phase | 5 | 4 | 2 | 5 | Alex presses Greg to define 'takeoff' and clarify exactly what OpenAI's upcoming autonomous AI researcher will do. Greg explains it will execute the end-to-end duties of a research scientist under senior human supervision. | |
| AI Safety Risks, Prompt Injections, and Economic Impact | 5 | 4 | 2 | 4 | Alex asks whether rapid takeoff introduces catastrophic risks. Greg acknowledges safety concerns, highlighting technical mitigations against prompt injection attacks alongside broader macroeconomic considerations. | |
| Centralized Control vs. Ecosystem Resilience in AI Governance | 6 | 5 | 3 | 6 | Alex challenges Greg on whether the reward of AI racing is worth the asymmetric risk of open-source bad actors. Greg argues against centralization, comparing AI governance to the decentralized standards and resilience developed around the electric grid. | |
| Defining AGI, "Jagged" Intelligence, and the Near-Term Horizon | 6 | 4 | 2 | 5 | Alex cites Jensen Huang claiming AGI is already achieved and asks if Greg agrees. Greg characterizes current systems as 'jagged intelligence' and estimates we are 70 to 80 percent towards his definition of AGI. | |
| The Autonomous Coding Breakthrough | 5 | 4 | 1 | 3 | Alex asks what triggered the breakthrough in autonomous coding reliability in late 2025. Greg explains that jump from solving 20% to 80% of tasks came from improvements in base model pre-training. | |
| Repurposing Codex for Universal Knowledge Work | 6 | 4 | 2 | 5 | Alex calls out Greg's previous public assertion that Codex was strictly for developers. Greg explains that discovering general problem-solving and harness capabilities led them to reposition Codex for knowledge work. | |
| Managing Fleets of Autonomous Agents and Retaining Accountability | 7 | 4 | 3 | 7 | Alex quotes Greg on how managing fleets of agents makes one 'lose pulse on the problem' and pushes back when Greg conflates that with delegating responsibility. Greg clarifies that retaining oversight is necessary to build calibrated trust. | |
| Speech Interfaces, Tool Use, and the Universal "AlphaGo Moment" | 5 | 5 | 2 | 4 | Alex asks why breakthrough 'AlphaGo Move 37' moments haven't yet manifested widely across science or creative fields. Greg explains the historical constraint of verifiable reward functions in math versus subjective domains. | |
| Human Purpose, Connection, and the Value of Technical Skills | 6 | 5 | 2 | 5 | Alex brings up Peter Thiel's observation about math people being displaced faster than verbal thinkers, then probes whether frontier pre-training runs are still necessary. Greg explains that pre-training multipliers make downstream RL and inference vastly more efficient. | |
| The Indispensability of NVIDIA GPUs and Frontier Ambition | 6 | 5 | 2 | 6 | Alex questions whether OpenAI still needs massive NVIDIA GPU clusters if inference dominates, and demands the financial math behind $110B datacenter capital commitments. Greg explains compute acts as a revenue driver analogous to scaling a salesforce. | |
| Addressing Over-Investment Fears and Industry Compute Scarcity | 7 | 4 | 5 | 6 | Alex cites Dario Amodei's public critique that aggressive infrastructure bets are 'YOLOing' risk and risking bankruptcy. Greg bluntly disagrees, asserting OpenAI foresaw the severe global compute shortage that other labs are now scrambling to navigate. | |
| Custom Software Creation and Personalizing the Human-Computer Interface | 6 | 4 | 2 | 4 | Alex describes his own workflow creating custom production tools with AI, then cites YouGov polling showing strong public skepticism of AI. Greg shares examples of medical interventions and argues the positive impact narrative remains underreported. | |
| Environmental Myths, Energy Sourcing, and Grid Modernization | 7 | 5 | 4 | 7 | Alex cites Pew data showing community opposition to data centers regarding home energy costs and pollution. Greg counters by debunking water consumption myths and detailing how datacenter investment funds modernizing stranded electrical grid capacity. | |
| Political Contributions and Single-Issue Technology Advocacy | 6 | 3 | 2 | 5 | Alex asks Greg to account for donating $25 million to MAGA Inc. as a single-issue donor and asks whether overall national health should supersede single-issue technology advocacy. Greg defends supporting politicians who actively back American AI leadership. |
Statements from this episode (48)
Brockman: AI Progress Now Requires Real-World Feedback Over Benchmark Testing
“And that we're moving out of testing on benchmarks and sort of these almost cerebral demonstrations of capability to it actually being the case that for us to develop it further, we need to see it in the real world and get feedback from how people are using it…”
Brockman: OpenAI's Top Priorities Are Personal Assistants and Problem-Solving Agents
“And for us, it's very clear that they're, the stack rank includes two things at the top. One is the personal assistant. The other is the AI that can go and solve hard problems for you.”
Brockman: OpenAI Lacks Enough Compute to Fund Its Top Priorities
“And when we look at the compute we have, we are not even gonna have enough compute to fund those two things.”
Brockman: Sora Video Architecture Is a Separate Tech Branch From GPT
“The Sora models, which are incredible models, by the way, are a different branch of the tech tree than the core Reasoning GPT series. They're just built in a very different way, and to some extent we're really saying that pursuing both branches is very hard fo…”
Brockman: OpenAI Continues Sora Research Program Specifically for Robotics Applications
“We are actually continuing the SOAR research program in the context of robotics, right, which I think is very clearly going to be a transformative application, which is still a little bit in the research phase”
Brockman: AI knowledge work will see a real takeoff within the next year
“We're going to see this real takeoff Of this technology and knowledge work over the next year.”
Brockman: Speech-to-speech interface shares OpenAI's core model architecture
“Having a great speech to speech interface. That is something that also is going to make this technology very usable and very useful, but it's not a different branch of the tech tree. It's all kind of one model. And we just sort of tweak that in slightly differ…”
Brockman: OpenAI Has Definitive Line of Sight to AGI in 2026
“And I think that we have definitively answered that question of it's, it is going to go to AGI. Like we see line of sight and that it is at this point, we have line of sight to these much better models that are coming this year.”
Brockman: OpenAI Reasoning Model Solved an Unsolved Physics Problem in 12 Hours
“We had this result recently where a physicist had been working on a problem for some time. He gave it to our model. 12 hours later, we have a solution, and he said this is the first time he's seen a model where he felt like he was thinking, that it felt like t…”
Brockman: ChatGPT Image Generation Uses GPT Architecture, Not Diffusion Models
“The reason we're able to do that is because it's not actually on the world model, like diffusion model tech, tech branch. It's actually based on the GPT architecture, and so there, even though it's a different data distribution, the actual core technology at t…”
Brockman: AI Progress Is Now About Execution Harnesses Over Raw Models
“And we talked about it a little bit in the case of the underlying models, but the thing that's really changed over the past couple of years has been that it's no longer just about the model. It's about the harness. It's about how does the model get context? Ho…”
Brockman: OpenAI Is Unifying Agentic Interaction Harnesses Into a Single Super App
“All of that was something that we had multiple implementations of, or slightly different, and we're converging it. We're going to have one version of that, and almost end up with this AI layer that can be pointed at specific applications in a very thin way, so…”
Brockman: OpenAI Will Incrementally Ship Super App Vision Over Next Months
“We're taking incremental steps to get there over the next couple months. We should have shipped the complete vision of what we're talking about here, but it's gonna come in pieces”
Brockman: Codex app functions as a general tool-using agent harness
“The Codex app today is something which is a, it's really two things in one. It's a general agent harness that can use tools, and it's also a agent that knows how to write software.”
Brockman: OpenAI Will Expand Codex for General Knowledge Work Use
“We're going to make the Codex app just so much more usable for general knowledge work, because it already, what we've seen within OpenAI is all this organic adoption of people using it for that. So that'll be the first step, and there are many to come.”
Brockman: Codex usability for non-engineers remains quite low
“The Codex app itself was originally built for software engineers, and that it's almost like the current usability of it for non-software engineers is actually quite low, because there's a bunch of little things where, when you set things up, you run into some …”
Brockman: OpenAI Underinvested in Last-Mile Usability for Early Coding Models
“We always had the best numbers on different programming competitions, these very cerebral things, but the thing that we didn't invest in as much was that last mile of usability.”
Brockman Focused Primarily on OpenAI GPU Infrastructure Over Past 18 Months
“And that's where I've actually been spending most of my efforts over the past. 18 months has been really focused on our. GPU infrastructure on supporting the teams that do all of the training frameworks to scale up at these big runs”
Brockman: New 'Spud' Base Model Incorporates Two Years of OpenAI Research
“So I think of SPUD as a new base, as a new pre-train, and that we have had this I, I'd say it's like we have maybe two years worth of research that is coming to fruition in this model.”
Brockman: New Models Will Only Feel Different in Intelligence-Bottlenecked Domains
“I think that it will be A similar story where you, when you release it, there will be people who will try it and be like, this is a night and day different than anything I've seen. And then there will be some applications where we weren't necessarily intellige…”
Brockman: Friend Used ChatGPT to Find Treatment After Terminal Cancer Diagnosis
“I have a friend who used ChatGPT to understand different treatments for his cancer, and that he was told by doctors that he was terminal, that there was nothing they could do for him. He used ChatGPT to actually research a bunch of different ideas, and he was …”
Brockman: OpenAI Automating Its Own Research Scientists' End-to-End Workflows
“And at a practical level, I think I would view it as ex, is taking the full end to end of what one of our research scientists does and be able to do that in Silicon.”
Brockman: OpenAI Has Invested Heavily to Defend Against Prompt Injections
“If you're going to have an AI that is very smart, very capable, hooked up to lots of tools, you want to make sure that it can't be subverted by someone giving it a weird instruction. And that's something that we've invested in quite a lot, and I think have rea…”
Brockman: We Are Currently 70% to 80% of the Way to AGI
“I think that I'd say I'm, Basically, like, 70, 80% there, so I think we're quite close.”
Brockman: AGI Capable of Any Computer Task Will Arrive Within Years
“I think it's extremely clear that we are going to have AGI within the next couple years in a way that is still going to be jagged, but that the floor of task will just be almost for any intellectual task of how you use your computer. The AI will be able to do …”
Brockman: Late 2025 Models Boosted AI Task Completion from 20% to 80%
“So new model races really went from the AI being able to do like, 20% of your tasks to like, 80%.”
Brockman: AI capabilities working today will be incredibly reliable within a year
“But it is all indicated by what's kind of working right now, certainly within a year, sometimes much sooner is going to be incredibly reliable.”
Brockman: AI Models Now Have Raw Intelligence to Manage Spreadsheets and Presentations
“There's so much just very mechanical skill associated with Excel spreadsheets, with presentations, and if the AI has the context, it has the raw intelligence now to be able to do these things at a great level. So if we can just make it more accessible, suddenl…”
Brockman: Human Operators Hold Strict Accountability When Autonomous AI Agents Fail
“If you're Trying to build a website and your agent messes it up and your user is affected. It's not really the agent's fault. It's your fault. And so you need to care. And I think that for people to use these tools, right, you need to realize that human agency…”
Brockman: AI Agents Will Democratize Entrepreneurship by Autonomously Running Businesses
“The democratization of entrepreneurship is absolutely coming. And I'll say, here's these problems, there's this customer that's upset, you know, that they want to talk to a real human, like, you should go talk to them, like, all of that's going to happen.”
Brockman: AI Will Produce 'AlphaGo Moments' Across All Scientific and Creative Domains
“That is going to happen in every single domain. It will happen in science, in math, in physics, in chemistry, it's going to happen in material science, it's going to happen in biology, it's going to happen in healthcare, drug discovery, but it may also even ha…”
Brockman: AI will transform the economy even with zero further technical progress
“So I think that even with no further progress, there's still a massive shift that will happen. The economy being powered by compute and AI is still going to happen.”
Brockman: Pre-Training Capability Multiplies Through the Entire AI Model Pipeline
“Every single step of the model production pipeline multiplies. And so you want to improve all of them. And the thing that we see is we prove the pre-training. It makes all the other steps much easier. And it makes sense because it's a model is able to learn fa…”
Brockman: AI Scaling Now Balances Raw Intelligence With Inference Costs
“We used to really just focus on the raw pre-training capability, but not think as much about the inference ability. And that's been a big change over the past 24 months to realize that it's a balance between You can have this model that has all those great pro…”
Brockman: Massive Scale AI Training Requires Concentrated Compute Clusters
“Well, because the, there's multiple reasons but one is that even as the balance of how much inference versus training changes, that you cannot get massive scale training Through any other way besides this concentration of compute on one problem.”
Brockman: Treat AI Compute as a Revenue Center Like Hiring Salespeople
“You can think of compute not as a cost center, but as a revenue center. Think of it a little bit like hiring salespeople. How many salespeople do you want to hire? As long as you can sell your product, as long as you have a scalable way to sell that product, t…”
Brockman: Major AI Compute Purchases Require 18 to 24 Month Lead Times
“These compute purchases You have to lock them in 18 months, sometimes 24 months, sometimes longer, In advance of them actually being delivered, which means you really need to project forward.”
Brockman: Most OpenAI Revenue Has Come From Consumer Subscriptions
“To date, most of our revenue has come from consumer subscriptions, and that will always be very important.”
Brockman: AI and Compute Will Be the Primary Driver of Economic Growth
“The highest order bit on how this economy grows from here will be about AI, how well you can leverage AI and the computational power you have available to power it.”
Brockman: Everyone in AI Will Face Severe Compute Shortages This Year
“And I think that we will see even this year, how everyone who is participating is going to be compute strapped.”
Brockman: Competitors Scrambled Too Late for Compute When None Remained Available
“I think that what we have seen is that for other players, that they kind of realized that probably late last year and started scrambling to see what compute is available, and there really wasn't any”
Brockman: Top AI Models Will Face Compute-Driven Outages Within Six Months
“Six months from now, I think that everyone will feel it. And I think that we will all feel the pain of there's an awesome model. And there's just no availability because there's not enough compute.”
Brockman: AI Will Be the Primary Source of Future National Security
“I think this will be the source of economic and national security going forward. I think it's going to be about national competitiveness and that there are other countries like China where AI pulls in the exact opposite direction.”
Brockman: OpenAI Abilene Supercomputer Annually Uses Water Equivalent to One Household
“If you actually look at our Abilene facility, which is one of, if not the biggest supercomputers in the world, the amount of water it uses Is the same as a household over the course of a year, right? So it's really negligible water use.”
Brockman: OpenAI Commits to Funding Power Infrastructure to Protect Local Prices
“On power, we have a commitment that we are going to pay our own way to not drive up energy prices for people.”
Brockman: North Dakota Utility Rates Decreased Following Local Data Center Construction
“We've seen, for example, in North Dakota that people's rates have gone down because a data center has shown up and has helped with improving the utilities for everyone.”
Brockman Confirms Contribution to MAGA Inc. Alongside Bipartisan PAC Donations
“So the way I look at this, so my wife and I made that donation. We've donated to bipartisan super PACs as well.”
Brockman Identifies as Single-Issue Political Donor Focused on Advancing AI
“I, you know, I am a one issue donor. This is something where I feel like I have a unique contribution to make, but it's really about just expressing support for this technology is something that we should be leaning into as a country.”