Sam Altman: 99% Said We Were Crazy Back Then, Now Is the Best Time to Start a Company

Summary

In a fireside chat at the AI Startup School in San Francisco on June 16, 2025, Sam Altman delved into the founding journey of OpenAI, the evolution of ChatGPT, and the future direction of AI hardware. He shared his personal journey from the struggles of early entrepreneurship to establishing OpenAI, emphasizing the importance of ambition, responsibility, and the necessity of continuing to build under intense public scrutiny.

Altman pointed out that OpenAI was founded on a seemingly crazy decision—to pursue Artificial General Intelligence (AGI). In 2015, 99% of people thought they were insane, but it was the 1% who resonated with the vision, including top-tier talent, that formed OpenAI's early team. He emphasized that being different is key to attracting talent; if everyone is doing the same thing, it's difficult to gather excellent people and get them to truly believe in a mission. OpenAI's starting point wasn't grand; it began with a few people in a room, focused on writing high-quality research papers. He quoted Vinod Khosla, highlighting the similarity between "zero-million-dollar" and "zero-billion-dollar" startups in their early stages—they are all just trying to get their first product to work.

Regarding ChatGPT and future models, Altman stated that we are currently in an interesting period where product innovation lags behind model capabilities. The cost of models like GPT-4o will drop significantly, and upcoming open-source models will bring surprising capabilities for local execution. He believes reasoning models have not yet reached their full potential, and now is an excellent time to start companies leveraging these new capabilities. Memory is Altman's favorite new feature, signaling that AI will become an "entity" that proactively understands users, connects all their data, and acts on their behalf. Ultimately, through new types of devices and seamless integration, it may become a persistent companion in users' lives. He envisions that future AI interfaces will "melt" into the background, with computers efficiently completing tasks behind the scenes and only interacting with users when necessary, thereby reducing the stress and distraction caused by current digital devices.

For GPT-5 and beyond, Altman's vision is to create a super-model that integrates multiple modalities like reasoning, real-time video generation, and code writing, which will bring a completely new computer interaction experience. He believes that AI with vision, speech, and reasoning capabilities is a key step towards the era of robotics. He even envisions that in the future, users subscribing to the highest tier of ChatGPT might receive a humanoid robot for free. He acknowledges that both the mechanical engineering of robot manufacturing and the AI cognitive aspects are extremely challenging, but believes that within the next few years, robots will begin to perform very useful work. He also discussed the revival of American manufacturing, suggesting that AI and robotics offer new pathways to bring complex industries back onshore.

On startup defensibility, Altman advises startups not to try to replicate OpenAI's core chat assistant, but to focus on filling gaps in the market. He points out that when industry clock cycles change dramatically, startups usually win because they iterate faster and at lower cost. He emphasized the importance of being "contrarian but correct" and shared the difficulty of maintaining conviction in the face of skepticism. He recalled Elon Musk once giving OpenAI a "zero" percent chance of success, which deeply impressed upon him the challenge of sticking to one's beliefs.

Regarding AI agents, Altman believes they will transform workflows, enabling AI to handle short-term, computer-based tasks like a junior employee. He is confident that current models are already capable of supporting many such experiences. On talent recruitment, Altman emphasizes looking for people who are smart, motivated, curious, self-driven, hardworking, have a good track record of achievement, can integrate well into the team, and align with the company's vision. He prefers "young, driven people who can get things done" over seasoned managers with impressive resumes, citing the advice to "hire for slope, not for Y-intercept."

The future direction Altman is personally most excited about is "AI for Science." He believes all long-term, sustainable economic growth stems from the discovery of new science, and AI can greatly accelerate this process. He has long been fascinated by energy and AI, seeing them as key technological levers for achieving "radical abundance." Although he initially thought they were independent, he later realized energy would become a fundamental constraint on AI development. He firmly believes technology can create abundance and finds the current anti-growth sentiment difficult to understand. He thinks that in the next five years, AI technology will maintain the same pace of progress as the past five years, and society will leverage this technology to create immense value, much like the invention of the transistor, leading to significant improvements in quality of life.

Finally, Altman's advice to young entrepreneurs is to recognize the difficulty of starting a company, but also to believe that they will ultimately succeed. He emphasized the importance of belief, resilience, and trusting one's intuition, encouraging people to bravely pursue things that are unpopular but they truly believe in and care about. He likens entrepreneurship to raising a child—the wonderful parts are beyond imagination, the hard parts are far more difficult than anticipated, but the key is perseverance.

Highlights

• 🚀 The Crazy AGI Vision and Talent Attraction: OpenAI was founded on an AGI vision considered crazy by 99% of people at the time, but it was this distinctiveness that attracted top talent and formed the early team.
• 💡 Product Innovation Lags Behind Model Capability: There is a huge opportunity in the AI field right now. Model capabilities have entered a new realm, but product innovation hasn't fully caught up, indicating a wealth of new applications and companies yet to be developed.
• 🧠 Memory and the Future of AI Companions: ChatGPT's memory feature is Altman's favorite new characteristic. It points towards AI becoming a personalized companion that proactively understands users, connects all their data, and acts on their behalf.
• 🤖 Multimodal Super-Models and the Robotics Era: GPT-5 and future models will integrate deep reasoning and multimodal capabilities (like real-time video generation, code writing), and ultimately combine with robotics to automate the physical world.
• 🛡️ Startup Defensibility: Build What's Missing, Don't Replicate the Core: Advises startups not to copy OpenAI's core product, but to focus on building the missing pieces in the market, using AI as a platform for innovation.
• 📈 A Personal Bet on AI for Science: Altman is personally most bullish on AI's application in scientific discovery, believing it will greatly accelerate the birth of new science, leading to massive improvements in human quality of life.
• 💪 Belief and Resilience: The Essence of Entrepreneurship: The entrepreneurial path is full of challenges, requiring firm belief and strong resilience. One must persevere even in the face of skepticism and failure, trusting one's own intuition.

Key Insights

• 💡 The Strategic Appeal of the AGI Vision: From its inception, OpenAI boldly declared its pursuit of AGI. This seemingly crazy vision, in fact, became a powerful magnet for attracting top talent. This demonstrates that in highly competitive fields, a unique and disruptive long-term goal, even if initially misunderstood, can effectively consolidate resources and talent by providing a unique platform for those eager to do "big things." This "counter-mainstream but correct" strategy is one of the cornerstones of its success.
• 💰 The Scissors Gap Between Model Capability and Product Innovation: Sam Altman pointed out that the current capabilities of models (such as GPT-4o's reasoning power and cost reduction) far exceed the development level of existing products. This indicates a massive "product vacuum" in the market, meaning a vast number of potential applications and business models remain untapped. For entrepreneurs, this presents an unprecedented opportunity to leverage existing and upcoming powerful AI models to develop entirely new products and services that can fully utilize AI's reasoning capabilities, going far beyond simple chatbots.
• 🧠 The Profound Impact of AI Memory on User Experience: Altman considers the memory feature his favorite new addition this year because it marks a shift for AI from a passive response tool to an active, personalized companion. This "Her"-style vision—where AI understands the user, connects all their data, and proactively offers assistance—will fundamentally change human-computer interaction. Future AI will no longer be a simple query tool but an intelligent assistant integrated into daily life, heralding a more seamless, personalized, and efficient digital living experience.
• 🤖 The Trend of Multimodal AI Merging with the Physical World: GPT-5 and future models will integrate deep reasoning with multimodal capabilities, such as real-time video generation and code writing, vastly expanding AI's application boundaries. More importantly, Altman explicitly stated that this powerful AI capability will ultimately combine with robotics technology to automate the physical world. This means AI will not only impact the digital realm but also profoundly transform industry, supply chains, and daily life, foreshadowing an era of "radical abundance" driven by intelligent robots.
• 🛡️ A New Paradigm for Startup Defensibility: Differentiation and Platform Leverage: In the AI era, startups face competition from giants like OpenAI. Altman's advice is not to replicate core products but to "build the missing pieces." He emphasizes that when an industry undergoes massive transformation, startups, with their faster iteration speed and lower cost advantages, often surpass large companies. Simultaneously, leveraging OpenAI as a platform—through its APIs, traffic funneling, and future "app stores"—can provide new ventures with unique defensibility and growth opportunities.
• 🔬 AI for Science: The Ultimate Lever for Accelerating Human Progress: Altman views "AI for Science" as his most exciting personal future direction, closely linking it to long-term sustainable economic growth and improvements in human quality of life. He believes scientific discovery is the foundation of all progress, and AI can accelerate this process at an unprecedented pace. This is not just a vision of AI's technical potential but also a profound reflection on the path of human societal development, positioning AI as a key tool for solving major global challenges and achieving widespread prosperity.
• 🧘 The Psychological Resilience and Intuition Cultivation of Entrepreneurs: Reflecting on the entrepreneurial journey, Altman emphasized the importance of conviction, resilience, and trusting one's intuition. He shared difficult moments facing Elon Musk's skepticism, noting that the challenges of entrepreneurship far exceed imagination, yet perseverance is key to success. This offers valuable psychological advice for all entrepreneurs: entrepreneurship is not just a contest of technology and business but also a test of personal will and belief, requiring continuous learning, adaptation, and self-trust.

References

1. Sam Altman: The Future of OpenAI, ChatGPT's Origins, and Building AI Hardware (YouTube) — The original video of the June 16, 2025, San Francisco AI Startup School fireside chat, the primary source for this article.
2. Complete English Transcript of the Fireside Chat (Singju Post) — The full verbatim transcript of this interview, useful for verifying the original quotes cited in the text.