Dialogue with Sam Altman: Not taking any equity in OpenAI was the biggest mistake I made

Compiled & Translated by: TechFlow

TechFlow Introduction: As the evolution of Artificial General Intelligence (AGI) and Artificial Superintelligence (ASI) becomes the foremost frontier of global technological competition, OpenAI CEO Sam Altman appeared on the renowned Indian podcast "Figuring Out" for an in-depth conversation with host Raj Shamani spanning nearly 70 minutes. The dialogue not only systematically reviewed how the Scaling Law evolved from an intuitive hypothesis into a scientific consensus, but also offered a rare and candid response to external misunderstandings and public controversy surrounding OpenAI's massive compute bet, dramatic executive departures, and the fact that he holds no direct equity in the company.

From betting the company's fate on amassing compute during the chaotic period when GPT-4 training was complete but not yet unveiled, to withstanding internal resistance and pushing forward the first-generation ChatGPT despite a mere 5% retention rate in internal testing, Sam Altman deeply deconstructed the underlying principles of non-consensus innovation and high-leverage decision-making. He also delved into the ambition-cultivation methodology honed during his Y Combinator era, the exponential compounding logic of personal career development, and the physical and mental costs behind the ultimate human endeavor of AI-driven scientific discovery—from the Navier-Stokes equations to fundamental mathematics.

Guest: Sam Altman (CEO of OpenAI, former President of Y Combinator)

Host: Raj Shamani (Renowned Indian business podcast host, entrepreneur)

Podcast Source: Figuring Out with Raj Shamani (https://www.youtube.com/watch?v=TjPPIf-M6KE)

Original Title: Sam Altman On AI Revolution, Superintelligence & Future Of Humanity | FO583 Raj Shamani

Air Date: November 3, 2026

Key Takeaways

• Reflections on Forgoing Equity and Social Misunderstanding: Sam Altman candidly admitted that choosing not to hold any equity in OpenAI was a practical decision-making mistake. Although the original intent was driven by a pure sense of mission—viewing it as an artistic pursuit—the inherent mindset of the business world could not comprehend such counterintuitive behavior, instead breeding unwarranted suspicion and confusion among the public about hidden motives for power.
• The Superlinear Returns of Compute: The most core insight running through Sam Altman's investment and entrepreneurial career is the astonishing magic of Scale. From the network of YC-incubated companies to early neural networks, anything that works at a small scale often exhibits outsized returns far beyond linear when dramatically amplified—this is also the first-principles reason behind OpenAI's firm bet on large models.
• The Life-or-Death Bet After GPT-4: In August 2022, during the long vacuum period when GPT-4 had just finished training and the world had yet to witness its capabilities, the team realized the model would open the door to reasoning and Chain of Thought. At a time when the broader world widely believed compute expansion was nearing its twilight, OpenAI withstood enormous uncertainty and pushed all of the company's chips onto a massive supercomputing cluster.
• The Counterintuitive Decision to Push ChatGPT: The first-generation ChatGPT internal test version had a retention rate of only 5%. Management and the research team were broadly opposed to releasing it, believing the product was crude and the timing was not right. Sam Altman insisted on launching it based on YC philosophy, firmly believing that a product that a tiny minority is fanatically passionate about—even if rough—must be subjected to real-world testing and feedback as early as possible.
• The Trust Dilemma of Core Team Departures: In the face of successive departures and the formation of rival camps by top figures such as Ilya Sutskever, Mira Murati, Dario Amodei, and Elon Musk, Sam Altman pointed out that this was not purely a conflict of interest, but rather that when touching technology capable of reshaping humanity's destiny—such as superintelligence—top geniuses universally develop a massive messianic paranoia and divergence over direction, believing that "only if I am at the helm can it be safe enough."
• From Engineering Tool to Scientific Discovery: OpenAI's most exciting next strategic goal is to make AI an engine for discovering entirely new scientific laws. From breakthroughs in Navier-Stokes fluid dynamics to proving frontier mathematical propositions, AI is transitioning from merely processing humanity's existing experience to exploring the underlying laws of the unknown universe.
• Ambition Can Be Learned: The most delightful discovery during his tenure at the helm of YC was that entrepreneurs' ambition and aspiration can be entirely reshaped through environment and peer groups. Place a person among peers who dare to challenge absurdly grand goals, and their originally narrow cognitive ceiling will be completely shattered.
• Exponential Compounding in Careers: The vast majority of people in the workplace merely repeat the same year of experience twenty times, rather than accumulating twenty years of deep compounding. The true rule of success lies in continuously breaking out of one's comfort zone, using non-consensus decisions at each stage to drive exponential personal leaps.
• The Physical and Mental Cost of Being Sam Altman: Shaping humanity's most critical technology inevitably means enduring the crushing weight of an extraordinarily frenetic life and serving as the projection target of global anxiety. Having lost the freedom and ease of working casually in coffee shops, he relies on intensive schedule management and the meditative focus cultivated in his early years to maintain an extremely high activation threshold and emotional stability at the eye of the storm.

2. Key Insights

• On Public Reflections About Forgoing Equity: "Looking back, not holding equity in OpenAI was a mistake I made. Although my original intent was very pure—like an artist fully devoted to creation, or a top athlete training hard without caring about getting rich—the business world simply isn't accustomed to such outliers. When you don't play by the conventional rules, the public starts wildly speculating about what unspeakable power you're really scheming for behind the scenes."
• On Trading Short-Term Low Status for Long-Term High Status: "The essence of non-consensus innovation is this: you must be willing to trade short-term low social status for long-term truly elevated status. When OpenAI was first founded, we could even be described as low-status—academia saw us as a bunch of unruly upstarts, asserting that AGI would take at least a hundred years or might never be achievable. But as long as the facts prove you right, being misunderstood by the whole world is your most powerful moat."
• On the Courage to Release the First-Generation ChatGPT: "If you're not somewhat embarrassed by the first version of your product, you've launched too late. At the time, ChatGPT's internal test retention rate was a dismal 5%, and all the data said don't ship it. But I firmly believed in YC's golden rule: rather than having a million people kind of like you, it's better to have five thousand people hopelessly and fanatically love you."
• On the Shakespearean Drama Behind Superintelligence: "Building superintelligence is impossible without wild twists and turns. When the stakes are so high that they concern the future of human civilization, many top minds develop a psychological fixation: they dare not hand the steering wheel to anyone but themselves. This distrust of power concentration and technological direction has given rise to the entire industry's epic saga of partings and conflicts."
• On AI's Ultimate Quest to Explore Unknown Science: "What excites me most isn't just building a chat assistant, but letting machines help us understand the universe. The astonishing capabilities AI has demonstrated in mathematics and problems like the Navier-Stokes equations show that we are far closer than anyone expected to the singularity where machines autonomously discover new physics insights, crack intractable diseases, and unlock clean energy."
• On the Compounding Rule That Breaks Cognitive Mediocrity: "The tragedy for many people is that they claim to have twenty years of work experience, when in reality they've merely mechanically repeated their first year's experience twenty times. You must force yourself to stay on an exponential growth curve, even if every stage comes with tremendous discomfort and completely new unknowns."

3. Main Text

1. Reflections on the Misunderstanding of Forgoing Equity: Why Is Selfless Dedication Seen as an Anomaly in the Business World?

Raj Shamani: When I read reports about you in the US and India, I noticed a very polarized phenomenon: you've sparked the greatest technological revolution of this era, yet you didn't take even a tiny bit of equity in the company you founded. But somehow, many people have suddenly started casting you as a villain. How did this happen? Why are you facing this situation?

Sam Altman: I think this largely stems from American society's natural vigilance toward power, especially the deep skepticism toward the enormous power now held by tech giants. In some ways, having this oversight and vigilance is a good thing, but sometimes it does feel like it's gone a bit too far.

Raj Shamani: So what do you think is the biggest misunderstanding the outside world has about you?

Sam Altman: Frankly, the very point you just mentioned is indeed one of the biggest misunderstandings. Looking back now, deciding not to hold direct equity in OpenAI was actually a very real mistake. When I made that decision, I believed I had extremely noble and sufficient reasons, but it created enormous confusion and chaos for the outside world.

Raj Shamani: Why do people misunderstand this decision? What is it that they can't grasp?

Sam Altman: Because it completely departs from normal human thinking. In people's traditional understanding, when someone founds a company and fights tooth and nail as a business leader, the only reasonable motive is to obtain massive wealth returns. This behavior pattern is too incompatible with conventional business logic, and the public is naturally uneasy about outliers they cannot understand. If an artist pours their heart into pure art, or a sports star trains day and night for competitive glory rather than making big money, everyone finds it perfectly natural and deeply respectable. But in the business realm, people start to wonder: since he doesn't want money, what does he actually want? Is he secretly scheming for greater power and control?

Raj Shamani: So what is your true driving force?

Sam Altman: Because this is the most fascinating, most important, and most beneficial thing for human society that I can imagine. Being able to personally participate in and drive such an exploration is itself an unparalleled thrill and privilege. For me, this is the most worthwhile endeavor to dedicate my life to.

Raj Shamani: You wrote a famous passage many years ago: you should be willing to trade short-term low status for long-term elevated status. As long as you are right, being misunderstood by the public is a tremendous advantage rather than a disadvantage. But the key question is: before the final results are clear, how do you know whether you're merely being misunderstood or simply wrong?

Sam Altman: The truth is: before the dust settles, you can never be one hundred percent certain. You can only have increasingly strong conviction—as you continuously gather new evidence, your certainty gradually increases. But the unknown risk is precisely the most exciting part of this adventure. When we first founded OpenAI, it was a very low-status endeavor. Everyone treated us as a joke, saying AGI would be a hundred years away even if achievable, thinking we were just a bunch of upstarts without proper academic credentials harboring unrealistic fantasies. But we saw physical laws that others didn't see. As we solved more and more puzzles, our confidence grew stronger. Although the probability of ultimate success was minuscule at the time, its expected value was high enough to justify betting everything.

2. The Superlinear Returns of Compute: From YC Experience to the Intuitive Breakthrough of the Scaling Law

Raj Shamani: When the whole world was bearish on AGI, what underlying signal made you so certain this would work, and willing to bet your entire career on it?

Sam Altman: I believe a person can only obtain a very small number of truly critical great insights in their lifetime, and all you can do is apply them across different dimensions again and again. For me, one of the most core insights is: Scale has an incredible superlinear emergent property.

Raj Shamani: Where did this understanding first take root?

Sam Altman: I've witnessed this pattern in many things—including the development of companies, the rise of nations, the evolution of cities, personal careers, and the investor and alumni network I built at Y Combinator. Anything that works well at a small scale, once you go all out to amplify it several times or even dozens of times, often erupts with astonishing outsized returns. Because I'm fundamentally a pattern recognizer, when we saw early on that neural networks not only worked but became smarter as we expanded parameters and data, I developed an extremely strong intuition: we must push scale to its limits at all costs. Although we hadn't yet derived the rigorous mathematical formula of the Scaling Law at the time, the simple rule that making networks bigger leads to capability leaps was already evident as early as the AlexNet era.

Raj Shamani: But this signal should have been visible to top scholars worldwide. Why didn't other major companies and institutions go all out to push it at the time?

Sam Altman: This is one of the most puzzling phenomena of my entire life. Going back to 2015, Google internally had almost the world's top deep learning talent and the most abundant compute reserves, but traditional academia had a deep-rooted disdain for the brute-force approach of simply piling on compute. They believed true intelligence must rely on elegant new algorithms, symbolic logic, or biomimetic structures—brute-forcing compute by a thousandfold was seen as unsophisticated grunt work. This intellectual arrogance gave us a precious four-to-five-year window in an unclaimed wilderness.

3. The Life-or-Death Bet After GPT-4's Arrival: Pushing All Chips In During the Vacuum of No One Understanding

Raj Shamani: Can you reconstruct for us one of the most critical life-or-death moments? A decision where the outside world was warning you it was absolutely impossible to succeed, and you—armed with half certainty and half what ordinary people would call obsession—made a choice that completely changed OpenAI's fate?

Sam Altman: The most breathtaking decision came at the moment GPT-4 training had just been completed, around August 2022. At that point, there were still about seven to eight months before we would officially release GPT-4 to the world. We were testing this model internally, and everyone knew the intelligence it demonstrated was unprecedented. But more critically, it took us some time to deeply realize the entirely new research dimensions it opened up: the model would continue to evolve at unimaginable speed after GPT-4, and we could fully build upon it to develop Chain of Thought and entirely new architectures with deep reasoning capabilities.

Raj Shamani: What strategic choice did that mean you needed to make at the time?

Sam Altman: We decided to bet the entire company's future on a super-scale compute procurement of unimaginable magnitude. At that point in time, the outside world hadn't even seen a glimpse of GPT-4, let alone understood what it could do. Mainstream opinion was even declaring that large language models had hit a wall. Everyone thought we were crazy—spending billions of dollars to pre-order unproven ultra-large-scale compute clusters was seen as suicidal. But because we had personally used GPT-4 and had thoroughly internalized the certainty of the Scaling Law, we believed we were most likely right.

Raj Shamani: If you had flinched at the time, what was the fallback plan?

Sam Altman: The fallback was to make a very conservative, safe, conventional compute investment and slowly advance model iteration. If we had chosen that path