OpenAI CEO Sam Altman has released a new essay titled “The Gentle Singularity,” outlining his vision for the impact of AI on human experience over the next 15 years.

This essay exemplifies Altman’s characteristic futurism, highlighting the potential of Artificial General Intelligence (AGI) while also downplaying its imminent arrival. As is typical of his writings, Altman sketches a future where AGI revolutionizes concepts of work, energy, and social contracts. Additionally, his essays often contain subtle hints about OpenAI’s upcoming projects.

In his essay, Altman predicts that by 2026, the world will witness the emergence of AI systems capable of devising novel insights. Although somewhat ambiguous, this statement aligns with recent indications from OpenAI executives that the company is focusing on enabling AI models to generate innovative ideas about the world.

During the announcement of OpenAI’s o3 and o4-mini AI reasoning models in April, co-founder and President Greg Brockman noted that these models were the first to be used by scientists to produce novel, helpful ideas.

Altman’s blog post implies that OpenAI may intensify its efforts to develop AI capable of generating novel insights in the coming year. OpenAI is not alone in this pursuit, as several competitors have shifted their focus towards training AI models that can assist scientists in formulating new hypotheses, thereby facilitating novel discoveries.

In May, Google introduced AlphaEvolve, an AI coding agent claimed to have generated novel approaches to complex mathematical problems. Another startup, FutureHouse, backed by former Google CEO Eric Schmidt, asserts that its AI agent tool has achieved a genuine scientific discovery. Furthermore, Anthropic launched a program in May to support scientific research.

If these companies succeed, they could automate a crucial aspect of the scientific process and potentially infiltrate massive industries such as drug discovery, material science, and other fields rooted in science.

This is not the first instance of Altman hinting at OpenAI’s plans through his blog. In January, he wrote an essay suggesting that 2025 would be the year of agents, which was followed by the release of OpenAI’s first three AI agents: Operator, Deep Research, and Codex.

However, creating AI systems that generate novel insights may be more challenging than making them agentic. The broader scientific community remains somewhat skeptical about AI’s ability to produce genuinely original insights.

Earlier this year, Hugging Face’s Chief Science Officer Thomas Wolf argued that modern AI systems are incapable of asking great questions, a crucial aspect of any significant scientific breakthrough. Kenneth Stanley, a former OpenAI research lead, also expressed that current AI models cannot generate novel hypotheses.

Stanley is currently building a team at Lila Sciences, a startup that has raised $200 million to develop an AI-powered laboratory focused on enabling AI models to formulate better hypotheses. According to Stanley, this is a complex problem, as it requires imparting AI models with a sense of creativity and interest.

While it remains to be seen whether OpenAI will successfully create an AI model capable of producing novel insights, Altman’s essay may offer a glimpse into the company’s future direction.