Interview with Sam Altman

At DevDay 2025, a palpable energy buzzed through the air as Rowan Cheung sat down with Sam Altman, the visionary CEO of OpenAI. With ChatGPT already boasting an astounding 800 million weekly active users, Altman offered a rare glimpse into the immediate future of AI, detailing groundbreaking announcements and musing on the profound shifts awaiting society. It was an interview that painted a picture of a world on the cusp of an intelligent revolution, driven by ever-smarter models and an increasingly accessible toolkit for innovation.

DevDay's Unveiling: Supercharging Creation with AI

Altman immediately highlighted his personal excitement for Apps in ChatGPT, a feature he’d "wanted to do for a long time." This new capability, alongside the Agent Builder and Agent Kit, promises to transform ChatGPT into the next great distribution platform. He noted the enthusiasm of builders already exploring the Agent Builder, which marks a significant leap from the GPT Builder launched two years prior. The core breakthrough, Altman explained, lies in the sheer improvement of the models themselves; "the difference in the model capability between then and now, it's like really for 22 months or whatever it's been come a super long way."

The Agent Builder empowers even average knowledge workers to create sophisticated agents with no code, simply by uploading files, linking data sources, and describing desired outcomes. This ease of use suggests a "tectonic shift" in software development, making the creation of impressive applications incredibly fast. Watching a rehearsal, Altman admitted, "I don't think I can come up with ideas fast enough anymore." This acceleration means the volume of software written will "drastically increase," and the time to test and improve ideas will plummet, though the ultimate implications remain to be fully understood. The conversation then delved into the tantalizing prospect of fully autonomous agents. While not yet ready for a "zero-person company," Altman believes week-long autonomous tasks are not far off, especially given the "disappointingly fast" progress of models like Codex.

Key Insights:

• Apps in ChatGPT is set to leverage ChatGPT's massive user base as a new distribution platform.
• The Agent Builder significantly lowers the barrier to entry for agent creation, enabling non-coders to build sophisticated tools.
• Model capabilities have improved dramatically, enabling faster software development and prototyping.
• Truly autonomous agents capable of week-long tasks are on the horizon, though still "years" away for "zero-person companies."

The Evolving Playbook for Builders and Innovators

For aspiring builders and founders, Altman acknowledged the overwhelming opportunity space, but offered unique counsel. He shied away from generic advice, stating, "the best unique advantages... are unique, like you figure it out just for you." Instead, he advocated for a more organic approach: "let tactics become a strategy." This philosophy suggests that by simply building things that work, a durable strategy can emerge. He cited ChatGPT's own journey, where the team wasn't initially confident about their enduring advantages. For example, "memory is a really great competitive advantage for us and a reason people keep using ChatGPT. That was not on our minds at all at the time."

OpenAI's commitment to self-improvement was further highlighted by the discussion around the GDPval benchmark. Despite their GPT-5 model placing second to Claude's Opus, Altman stressed the importance of transparency and humility. "I think it'd be really bad if we didn't if we weren't willing to release things where our model is second," he remarked, emphasizing a culture where admitting when others are better is crucial for growth. This open approach, even when it means acknowledging a competitor's strength in specific areas like enterprise use cases and output formatting, underscores OpenAI's long-term vision.

Key Learnings:

• Unique advantages for startups are highly contextual and must be discovered through building and iterating.
• The philosophy of "let tactics become a strategy" encourages practical action over abstract planning.
• Unexpected features, like memory in ChatGPT, can evolve into significant competitive advantages.
• OpenAI prioritizes a culture of honest self-assessment, releasing benchmarks even when their models aren't first, to drive continuous improvement.

AGI, Society, and the Unpredictable Future

The conversation shifted to the grander vision of AGI, which Altman defined as outperforming humans at "most economically valuable work." However, his personal focus has evolved to "novel discovery" – the ability of AI to expand the total human knowledge base. He noted the "very small" but growing examples of AI making scientific discoveries, calling it "a really big deal" and "the like AGI like thing I care about the most." He drew a parallel to the Turing test: "the thing that had been the AI test forever, just went whooshing by and we all adapted." He believes society will adapt equally quickly to AI making scientific discoveries, echoing his "only weird once" analogy applied to self-driving cars.

The rapid advancement also brings new challenges, such as the "workslop" phenomenon, where AI-generated, polished-looking output creates more human rework. Altman acknowledged this, but contextualized it: "a lot of humans do the equivalent of like work slop as well." He believes the economy will self-correct, favoring those who use tools effectively. Regarding Sora deepfakes, Altman shared his surprisingly calm reaction to seeing hundreds of memes of himself. He views early release with guardrails as a way to "help society with these transitions," noting that "society will adapt to this, of course." He stressed that indistinguishable AI video is not just a goal for its own sake, but a crucial step towards AGI, improving spatial reasoning and world models.

Looking ahead, Altman acknowledged the fear that AI might impact "a billion knowledge worker jobs" before creating new ones, unlike the internet era. He offered a fascinating perspective, suggesting that future jobs might look less like "work" from our current viewpoint, much like a 50-year-ago farmer might view today's desk jobs as "playing a game." While short-term worries exist, he remains deeply optimistic: "man, I'm so willing to bet on human drives being what they are. And, I think we'll find plenty of things to do." He concluded with a call for global policy, urging a "global framework to reduce catastrophic risk" for the very leading edge of powerful AI models. OpenAI's ultimate goal remains building a "really great AI super assistant," not an "everything app," with voice increasingly becoming a natural, intuitive interface.

Key Changes:

• AGI definition emphasis shifting towards AI's capacity for novel scientific discovery.
• Societal adaptation to major AI milestones is happening faster than anticipated ("only weird once").
• The early release of powerful tools like Sora is a deliberate strategy to allow technology and society to "co evolve."
• The nature of "work" is expected to profoundly change, potentially leading to new, currently unimaginable roles.

"[I'm so willing to bet on human drives being what they are. And, I think we'll find plenty of things to do.]" - Sam Altman