The future of AI research should focus on understanding the scaling laws and capabilities of reinforcement learning models.

We started building Deep Research over a year ago, driven by progress in reinforcement learning and reasoning models, initially focusing on math, science, and coding tasks.

To develop Deep Research, we created reinforcement learning tasks to teach the model browsing and data analysis capabilities, and provided it access to tools for training.

I think the biggest challenge for OpenAI over the next 12 months is doing the best research and productization of innovation.

The performance of AI models improves significantly with reinforcement learning, and OpenAI aims to scale this approach to enhance reasoning capabilities further, moving towards a future where reinforcement learning will dominate AI development and potentially lead to significant advancements in capabilities.

To maximize improvement in AI, we need to scale existing techniques to their limits while also investing in future innovations that could lead to significant advancements.

If the old scaling laws are beginning to lose their edge, we need to consider what comes next for AI development.

In the coming years, we will see more experiments on how to effectively deploy and manage multiple AI agents for complex tasks.

We need to develop scaling science alongside our compute scaling efforts to ensure effective advancements in AI capabilities.