Sam Altman on AI Adoption, Non-Consensus Bets, and OpenAI Strategy

Organized & Compiled by Shenchao TechFlow

Guest: Sam Altman, CEO of OpenAI

Host: David Senra, Founder / Host of the David Senra Podcast

Podcast source: David Senra

Sam Altman on Building OpenAI and Betting on the Impossible

Broadcast date: August 23, 2026

Disclosure: Sam Altman is the CEO of OpenAI. This episode focuses on OpenAI’s strategy, products, technology roadmap, and industry vision, and his views are closely aligned with the company’s interests. The content below faithfully presents the guest’s opinions and does not constitute an independent third-party assessment.

The most interesting part of this conversation is that Sam Altman, as the creator of AI, openly admitted he hasn’t yet crossed the “iPhone moment.” He invented Codex years ago, yet still uses a 20-year-old computer workflow: copying and pasting between messaging apps, checking emails only to reply to the easiest ones, and maintaining his to-do list exactly as before. He says the reason is simple and common: people psychologically equate inefficient methods with “work” itself.

This is precisely the most valuable insight for ordinary people throughout the entire issue. Over the past decade, from venture capital to OpenAI, what Altman has truly distilled has little to do with technical details: a set of universal principles for how to judge, how to bet, and how to make trade-offs. For instance, he believes that "research also follows a power law"—the single best idea matters more than all others combined; when hiring, he looks for whether someone is willing to stand by deeply unpopular beliefs; he learns more from success than from failure, because success is easier to replicate. He also unusually discussed his personal life: after his eldest son was born, he wrote him a letter every Sunday, but stopped after just eight letters—yet that framework of "not hiding behind anything" led him to reflect more honestly.

He also discussed OpenAI’s platform strategy and AI risks, but all of this is predicated on one assumption: that humans are the entire point. He is more concerned about loss of control and excessive concentration of power than about AI taking jobs.

On adoption and inertia

• Toby Lütke is the most forward-thinking CEO I’ve ever seen—he’s six to eight months ahead of other CEOs. But I’m not entirely sure 2026 will be the year when all businesses are completely reshuffled; I think it’ll happen a bit more slowly.
• Economy and society have tremendous inertia. People continue doing the same things, buying the same tools, and working in the same ways. This will make the upcoming major transition smoother, but also means we’ve been overly optimistic about the timeline.
• The way I use my computer today is no different from how I did 20 years ago. Better tools are available, yet deep down, I’ve convinced myself that this inefficient method is what “work” really means.

About Non-Consensus Staking

• In 2015, we said we were going to build AGI and were mocked by intellectual giants in the field. Later, when we bet on large language models, we were mocked again.
• The research also follows a power law: the best idea is more important than all the others combined.
• You’re looking for people whose way of thinking is clearly different from others and who are willing to hold onto beliefs that are deeply unpopular. If they’re wrong, they’re wrong; if they’re right, they’re incredibly right.

On Success and Failure: You Learn More from Success

• I learn more from success. Failure certainly offers lessons, but there are often many reasons why things fail, making it hard to identify the correct cause and effect.
• When something truly works—such as what strategies have been effective at YC or OpenAI—apply those lessons to the future to create greater impact.

About the letter to my son: An honest framework

• When my first child was born, I would tell him about my day and what I was worried about when I came home at night to soothe him to sleep.
• Writing a letter to your child has one advantage: you can’t hide behind anything. You truly care about how your child will view you in the future, so you’re more likely to honestly reflect on what didn’t go well this week and how to improve next week.
• This is a very interesting mental framework, but I stopped after writing eight letters.

On platform strategy and trade-offs: Sacrificing a good product for a great one

• We should be more of a platform company than a product company—a direct interface and an API that enable people to build everything on top.
• Last year, we discontinued Sora and the Atlas browser. Both were excellent products, but in a world with limited compute power, talent, and resources, we had to sacrifice good products for great ones.
• One of the hardest lessons for entrepreneurs is: killing good ideas.

On Risk and Human Agency

• The two risks from AI that concern me most are loss of control and excessive concentration of power. Both outcomes are fundamentally anti-human.
• The right path is not to entrust the future to a few companies, a few individuals, or a single model. People are the entire meaning of it all.
• It’s a very poor sales pitch to trade away people’s influence over the future for the promise that “we will cure all diseases and make the world cheaper.”

On OpenAI’s early days: Four and a half years without a product

• We founded the company at the end of 2015 and didn’t release our first product until mid-2020—four and a half years. This is completely opposite to YC’s “ship early” principle.
• In January 2016, a dozen people gathered in Greg Brockman’s apartment. Everyone was excited, but soon they looked at each other and thought: What do we do now? They didn’t even have a whiteboard.
• We went through a long period of chaotic stumbling before we finally stumbled upon the GPT research path.

David Senra: Why do you think Toby Lütke of Shopify is one of the most interesting CEOs today?

At every early stage of AI, Toby was the most forward-thinking CEO. He wrote his own software, conducted his own experiments, and sent us incredibly detailed product feedback. He was the first to say we weren’t an NPC company—we had to embrace agents, or we’d fail. He took action, had a deep understanding of technological boundaries, and was about six to eight months ahead of other CEOs.

David Senra says 2026 will be the year when every business is reshuffled, and he even plans to rebuild a native AI version of Shopify at night. Do you think this timeline is realistic?

I agree with Spirit, but I’m less convinced about the timeline. I think it will be slower than he expects. What I doubt isn’t the possibility, but the inertia of economic and social systems. People will continue doing the same things, buying the same products, and using the same tools. AI is one of the most remarkable technologies humans have ever created, but societal and economic adaptation will be much slower.

This is actually a good thing in many ways—it will make the upcoming major transition smoother. But we’ve all been overly optimistic about the timeline.

David Senra: This reminds me of what Larry Ellison said in the 1980s: it’s not a software problem, it’s a people problem. The technology is already there; what’s needed is to change human behavior.

Yes, changing behavior is much harder than tech enthusiasts imagine. I often cite the example of people still renting movies from Blockbuster even when Netflix was still mailing out DVDs. The power of that inertia is far greater than the technology itself.

David Senra: While inventing these products, have you found yourself still working according to old habits, even though you have better tools available?

Absolutely. I’ve been waiting for this question for a long time. The biggest inconsistency in my mind is that I’ve been using computers the same way for 20 years. Now, with a magical tool called Codex, I should theoretically be using computers in a completely different way. I shouldn’t copy and paste between different messaging apps, shouldn’t mindlessly scroll through emails and pick the easiest ones to reply to, and shouldn’t maintain my to-do lists using old methods.

But if I look at my revealed preference, it seems I still prefer the old way. This only shows one thing: somewhere inside me, I feel that this inefficient method looks like "work" and "being productive."

David Senra: What would make you truly dive deep into using your own product?

I think it will be a gradual process. Completely changing someone’s habits and workflows is difficult. We can still build better products to make the transition more seamless. Right now, it feels like the smartphone era before the iPhone. I used a Palm Treo back in 2003—the technology was already there; what was missing were the product ideas that made the iPhone the iPhone. Today, we have all the technological pieces, but we haven’t yet had that iPhone moment that completely transforms human-computer interaction.

David Senra: You were interested in AI as a child—why did you choose this path in 2015 when everyone thought it was impossible?

I was a nerdy kid from a young age, spending Friday nights on the computer, watching sci-fi, and reading sci-fi. I always thought AI would be the craziest thing. In college, I even worked in an AI lab for a summer, but nothing worked back then. My professor clearly told me: deep learning would never yield good results—it was the surest way to ruin your career. That was around 2005, and I believed him.

Later, I accidentally stumbled into entrepreneurship and fell in love with it. Going from founder to investor is actually the opposite of the conventional path. I strongly recommend this direction because, as an investor, you get to observe the most critical crux moments across a vast number of companies; while you may not have the daily hands-on practice, your dataset is extremely large.

David Senra: How does investing directly help you in managing OpenAI now?

A huge help is power-law thinking. You must reprogram your brain: your best investment will outperform the sum of all others, and your second-best will outperform the sum of all the rest. This is also true for AI research.

When we founded OpenAI at the end of 2015, there were almost no efforts related to AGI, aside from a handful like DeepMind. When we said we were going to build AGI, we were mocked by intellectual giants in the field. Later, when we bet on large language models, we were mocked again. Coming from a venture capital background, I understand that high-risk bets are fine—as long as they deliver immense value if they succeed. The greatest researchers are often those who hold non-consensus views, pursue novel approaches, possess high energy, and operate outside the standard norms.

David Senra: What exactly do you mean by "non-standard"?

You're not looking for someone who makes minor adjustments to existing ideas and then tries hard to convince you they're different. Such people are merely imitating Peter Thiel’s “be different,” but are still fundamentally following the crowd.

The truly valuable people are those whose way of thinking clearly differs from others and who are willing to hold onto beliefs that are deeply unpopular. They may be wrong, but if they’re right, they’re profoundly right.

David Senra: Where are you primarily focusing your efforts now?

Research and computing power. I’d love to spend more time on the product—we have excellent people working on it—but the most critical priority right now is building smarter models and making them accessible to many people in an efficient and scalable way. This is the high-leverage challenge of sustaining exponential growth, and it’s precisely the kind of complex problem I excel at.

Expanding computing power requires coordination among chips, semiconductor foundries, data centers, power systems, finance, policy, supply chains, and logistics. This may already be, or is becoming, the most expensive infrastructure project in human history. It combines technical, commercial, policy, and supply chain challenges—all layered together.

David Senra: You were previously a startup investor and now manage research projects—are they similar?

Very similar. On the surface, researchers and founders may seem different, but there are many parallels in how to identify non-consensus bets, how to assess conviction, how to understand exponential growth, and how to manage outlier talent. The building I’m sitting in right now is the research building, and my office is here.

David Senra: How many product lines do you currently have? Where do your revenue sources come from?

We just merged ChatGPT and Codex. Previously, there were ChatGPT, Codex, and API; the name Codex led many to mistakenly believe it could only write code, when in fact it can handle any task.

I believe OpenAI should be more of a platform company than a product company. We will certainly build some products, but what most people truly want is a unified interface to access personal or enterprise-grade AGI, along with an API that enables anyone to build anything on top of it. We will optimize AI performance at every point along the cost-performance curve. If you need high-end AI for scientific discovery, that’s available; if you need low-cost AI for large-scale repetitive tasks, that’s available too.

David Senra: What mistakes did you make and what did you sacrifice to learn this?

The hardest lesson is sacrificing good products for great ones. Last year, we shut down Sora—it was fun, cool, and had users—but it consumed massive compute resources, so we redirected that compute toward Codex. We also discontinued the Atlas browser; I think it might have been the best browser, but we needed to allocate our talent elsewhere. In a world where compute, talent, and resources are limited, we must be extremely focused: on general intelligence, knowledge work, scientific discovery, and the upstream efforts that enable them—such as our own chips, data centers, and infrastructure.

David Senra: Are you focusing on this? Did Peter Thiel give you any advice?

When ChatGPT was first released, it grew rapidly but felt unstable and lacked perceived value; many people were discussing whether they should pivot. I went to Peter, and he listed a bunch of potential directions we could shift toward. He said: “The more obvious mistake, beyond the fact that it’s growing, is to do something else. Its power lies in that Google-style blank text box—whatever you type into it, it can handle. It doesn’t fit the Silicon Valley wisdom of feeds, network effects, or user lock-in, but if the blank box worked for Google, why wouldn’t it work here?”

I listened, went all in, and it worked great.

David Senra: What are you concerned about regarding AI?

Two primary risks. One is loss of control—AI becomes too powerful for us to ensure the desired level of control. The other is excessive concentration of power, where a single company, model, or individual holds too much authority. Both are fundamentally anti-human.

The correct path is: humans must deeply control the future, and humans must be deeply empowered. Humans are the entire meaning of it all. We cannot, out of distrust in people, gradually hand over control to AI models. That is a deeply cynical stance.

Another equally dangerous version is that, out of fear of negative outcomes, we restrict who can use this technology and how it can be used, concentrating power in the hands of a few companies. I give you medicine to cure diseases and affordable goods, but you give up your autonomy and influence over the future, while accepting extreme inequality. This is a very poor sales pitch.

David Senra: Why would anyone want to use such phrasing?

I think it’s fear and power. Many people are frightened by the magnitude of AI risks and feel they must trade away significant freedoms for security. But this often becomes an excuse for rent-seeking behavior.

I agree with some of the doomers: this is powerful technology, and we should proceed cautiously, prioritizing safety. But I don’t agree that they view it as an unsolvable problem. When OpenAI was founded, two broad consensus views were held: first, that it would be impossible to create something very AGI-like within a decade; second, that even if we did, we couldn’t guarantee its safety. Now we’ve created something that most people at the time would have considered very AGI-like, and the world hasn’t ended—alignment hasn’t collapsed. Our predictions should be updated accordingly.

David Senra: Everyone is using AI right now, yet many people also dislike it—how do you explain that?

People have always feared rapid socioeconomic change. Not everyone felt good during the Industrial Revolution either. I think this inertia in human society is even a beneficial feature—it provides stability during times of upheaval.

On the other hand, the AI field—including ourselves—has done a very poor job of explaining to the public the benefits and how risks are being mitigated. We’ve repeatedly said things like “There’s a 25% chance of destroying the world” or “50% of jobs will disappear next year,” which is naturally frightening. What we’ve failed to communicate even better is that AI can empower people with more authority and personal freedom, not less.

I believe we’re about to witness the largest wave of small business entrepreneurship in history. AI is empowering this movement. In the past, starting a small company required significant privilege, luck, and resources—but AI is lowering the barriers. This space, including us, is talked about far too little.

David Senra: This reminds me of how Intel’s big three once taught more classes to explain microprocessors to customers and investors than community colleges did. Why isn’t anyone in the AI industry doing the same?

We have no excuses; we should be doing more. We’ve been trying, but haven’t yet found the most effective approach.

David Senra: Why has YC had such a profound impact on you and OpenAI?

YC didn’t just change how I run my company—it transformed the entire startup ecosystem. It instilled principles such as iterative deployment, empowering technologists, being willing to bet on young but energetic and ambitious individuals, shipping embarrassingly early versions, and learning from real user feedback.

OpenAI is, in many ways, a product of YC’s philosophy. But one thing is completely opposite: we released our first product four and a half years after founding the company. YC would say you should ship as quickly as possible, yet we had no external customer signals and had to find our own ways to substitute feedback on whether customers liked the product.

During the Dota 2 era, we created a leaderboard to allow different ideas to publicly compete for rankings, giving researchers an objective and authentic signal of progress. We also learned the power of external demos, encouraging researchers to showcase their results to key individuals they wanted to impress. We also tried some ineffective methods, such as fake deadlines.

David Senra: What was the scene on the first day of 2016?

In January 2016, a dozen people gathered in Greg Brockman’s apartment. Everyone was excited, like the first day of school. But soon, we looked at each other and thought: What do we do now? There wasn’t even a whiteboard. Greg had someone go get a whiteboard; when it arrived, silence fell again. That energy quickly faded. We all knew we wanted to build AGI, but had no idea where to start.

We did what we knew how to do: wrote papers, came up with ideas, and experimented. Many things didn’t work, but over time we found the rhythm for evaluating research bets and figured out how to prevent talented people from being wasted. After a long period of chaotic stumbling, we gradually stumbled upon the path to GPT—from the unsupervised sentiment neuron to GPT-1, and then to scaling laws, which gave us the confidence to invest in more compute.

David Senra: Which is more useful—what you learn from success or from failure?

I learn more from success. Failure certainly offers lessons, but there are often many reasons why things fail, making it hard to identify the correct cause-and-effect relationships. When something truly works—like which practices made YC successful or which approaches drove OpenAI’s success—applying those lessons to the future has a greater impact.

David Senra: Do you keep a journal?

When my first child was born, I would come home at night and tell him about what I had done and what I was worried about that day to help him fall asleep. Later, I thought these stories were interesting and maybe he’d want to read them when he grows up, so I started writing him a letter every Sunday. But I didn’t keep it up for long—only wrote about eight letters.

David Senra: You should keep writing. Founders don’t write autobiographies at 40 and wait until 70, by which time too much has been lost to time. Even writing a book for internal use would be valuable.

I’ll consider it. One benefit of writing to your child is that you can’t hide behind anything. You truly care about how your child will view you in the future, so you’re more likely to honestly reflect on what didn’t go well this week and how to improve next week. It’s a very interesting mental framework.