Most developers and power users currently experience artificial intelligence as a series of constraints. Whether it is the anxiety of hitting a rate limit on a Tier 1 API key or the monthly friction of a subscription plan, the current relationship with AI is one of rented intelligence. We operate within the boundaries set by a provider, treating the model as a tool we call upon rather than an environment we inhabit. This bottleneck is not just a matter of pricing or server capacity; it is a fundamental limitation of how intelligence is currently distributed and developed.

The March 2028 Automation Engine

OpenAI is moving to break this cycle by fundamentally changing how AI is built. The company has set a target for March 2028 to establish an automated system where a significant portion of its internal research is conducted by AI itself. For decades, the scientific method has relied on a human-centric loop: a researcher forms a hypothesis, designs an experiment, analyzes the data, and iterates. OpenAI intends to shift this core loop into the hands of the AI. In this new paradigm, AI systems will not merely assist in writing code or summarizing papers; they will actively explore alternative hypotheses, execute repetitive experiments, and identify errors in their own logic to find the optimal path toward AGI.

This shift is designed to remove the physical constraints of human time. By automating the research cycle, the speed of technical progress is no longer tied to the number of PhDs in a room but to the compute and algorithmic efficiency of the research agent. This automation is particularly critical for the problem of alignment. Ensuring that an AI's goals remain consistent with human intent is a moving target. OpenAI believes that the only way to achieve a safe AGI is to involve the AI in its own alignment process. By creating a recursive loop where the AI analyzes its own structure and provides real-time feedback on its safety parameters, the company hopes to build a self-correcting mechanism that evolves faster than the risks it creates.

From Product Company to Intelligence Utility

While much of the industry is locked in a race to increase parameter counts and climb benchmark leaderboards, OpenAI is pivoting toward a broader infrastructure play. The company is executing a three-stage evolutionary strategy that redefines its identity. The first stage was the pure research lab, focused on the foundational breakthroughs necessary to reach AGI. The second stage, which the world has witnessed through the explosion of ChatGPT, was the transition into a product company. This phase was about deploying research into the real world and using massive streams of human feedback to refine model performance.

Now, OpenAI is entering the third stage: becoming an AI infrastructure provider. The goal is to move away from selling a high-cost specialized service and toward providing intelligence as a universal utility. In this phase, the priority shifts from increasing frontier model performance by a marginal percentage to increasing the efficiency and accessibility of that intelligence. The objective is to remove the scarcity of intelligence entirely, transforming it into a resource as ubiquitous and affordable as electricity or water.

This strategy culminates in the vision of one AGI per person. OpenAI plans to distribute personalized AGI to every individual on earth, allowing users to possess their own high-performance intelligence tailored to their specific goals. This is a deliberate move toward decentralization. By distributing the power of AGI across billions of individuals and organizations, OpenAI aims to prevent the dangerous concentration of AI power within a few corporations or governments. This distributed structure is intended to increase societal resilience, ensuring that the benefits of AGI are not hoarded but are instead used to empower individual agency.

To manage the risks of this transition, OpenAI supports the creation of an international coordinating body. Such an organization would establish common safety standards to prevent a race to the bottom driven by nationalistic or commercial competition. This body would have the authority to implement coordination measures, including the potential to slow the development of frontier models to ensure that social safety nets and alignment techniques can keep pace with technical capabilities.

As this infrastructure becomes universal, the nature of professional value will shift. When every person has access to a personal AGI, the ability to write a perfect prompt or master a specific AI tool becomes a commodity with zero market value. The competitive advantage will move from execution to architecture. The role of the human professional will evolve into that of a system designer and domain analyst. Instead of performing the task, the human will define the purpose, set the constraints, and manage the trade-offs of the AI system.

This transition requires a new kind of AI resilience. Much like the introduction of the automobile required the invention of seatbelts and traffic laws before it could be safely integrated into society, the deployment of personal AGI requires a social and institutional framework to handle the resulting disruption. For practitioners, this means moving beyond tool proficiency and focusing on the ability to define business problems and design the logical structures that AI will execute. The value will lie in the ability to verify AI outputs through deep domain expertise and to architect the flow of data that leads to a meaningful result.

We are moving from an era of renting intelligence to an era of owning it. By automating the research process and pivoting to a utility model, OpenAI is betting that the winner of the AI race will not be the one with the smartest model, but the one who makes intelligence the most accessible resource in human history.