OpenAI is setting an aggressive timeline for AI systems that can participate directly in research work. During a livestream Tuesday, CEO Sam Altman said the company is internally tracking toward an intern-level research assistant by September 2026 and a fully automated "legitimate AI researcher" by 2028.
The claim is not about hiring a person who studies artificial intelligence. Jakub Pachocki, OpenAI's chief scientist, described the target system as one that could autonomously deliver on larger research projects. That framing matters: OpenAI is describing a move from models that answer difficult questions toward systems that can carry out extended research work with less human direction.
What OpenAI Means By An AI Researcher
The phrase "AI researcher" can sound broad, but the source description gives it a specific meaning. Pachocki referred to a system capable of handling larger research projects on its own. In practical terms, that suggests a system that can stay with complex problems over longer periods, reason through multiple steps, and help produce research outputs rather than only respond to short prompts.
OpenAI says its deep learning systems are improving quickly, especially in how fast they can solve complex tasks. Pachocki said current models can handle tasks with a roughly five-hour time horizon. He also said they can match top human performers in competitions like the International Mathematical Olympiad.
The company's next objective is to stretch that time horizon. A model that can work on a five-hour problem is different from a system that can pursue a larger research project. The gap between those two capabilities is where OpenAI is focusing much of its attention.
The Compute Bet Behind The Timeline
OpenAI's plan rests on two main strategies: continued algorithmic innovation and a much larger use of "test time compute." In plain language, test time compute is the computing power a model uses while it is thinking through a problem, not just while it is being trained.
Pachocki argued that models should be able to spend far more computational resources on difficult work. For major scientific breakthroughs, he said it could be worthwhile to dedicate entire data centers' worth of computing power to a single problem.
That is a striking way to frame AI research. It suggests OpenAI sees some future scientific questions as compute-intensive projects, where the value of the answer could justify unusually large amounts of infrastructure. The company is not only talking about smarter models; it is also talking about giving those models much more room to reason.
Altman connected this ambition to OpenAI's infrastructure plans. He said OpenAI has committed to 30 gigawatts of infrastructure, described as a $1.4 trillion financial obligation, over the next few years. In the company's view, that buildout is part of what would allow it to pursue faster scientific progress.
Why The Corporate Restructure Matters
The AI researcher timeline arrived on the same day OpenAI finalized its transition to a public benefit corporation structure. That move shifts the company further away from its non-profit roots and releases it from limitations tied to its non-profit charter. It also creates new opportunities for capital raising.
Altman said the new structure gives OpenAI a framework for pursuing its research assistant timeline while keeping a commitment to responsible AI development. The non-profit OpenAI Foundation will own 26% of the for-profit and will govern the research direction.
The foundation is also tied to a $25 billion commitment to use AI for curing diseases. According to the source, it will help manage AI research and safety initiatives as well. That gives the restructuring both a financial and governance role in OpenAI's plans.
The for-profit arm's ability to raise more funds is central to the strategy Altman described. More capital would support the infrastructure needed for scientific advances, including the computing resources behind longer and more complex model reasoning.
The Superintelligence Context
Pachocki placed the AI researcher target inside a broader expectation about where deep learning systems may be headed. He said, "We believe that it is possible that deep learning systems are less than a decade away from superintelligence." He described superintelligence as systems smarter than humans across a large number of critical actions.
OpenAI's stated research goals are connected to that belief. The company says AI could potentially make discoveries faster than human researchers, address complex problems beyond current human capabilities, and accelerate technological innovation across fields such as medicine, physics, and technology development.
Those are broad ambitions, but the source presents them as part of a single strategy: stronger algorithms, more compute during problem-solving, and a corporate structure designed to support large-scale infrastructure spending. The 2028 AI researcher target is one milestone inside that larger effort.
What To Watch Next
The clearest near-term marker is September 2026, when OpenAI says it is tracking toward an intern-level research assistant. That is a more immediate test of the company's progress than the 2028 goal.
Several questions follow directly from OpenAI's own framing:
- Can models reliably extend beyond the roughly five-hour task horizon Pachocki described?
- Will more test time compute produce better research outcomes on complex problems?
- Can OpenAI scale the infrastructure Altman said is needed for scientific advances?
- How will the OpenAI Foundation's governance role shape research direction and safety initiatives?
For now, OpenAI is making a clear claim: it expects AI research assistants to arrive soon, and it believes a more autonomous research system could follow by 2028. The company is pairing that claim with a major compute strategy and a new corporate structure built to raise the funds required for the effort.