"I think the race to AGI is somewhat overblown," an engineer at my firm wrote to me in 2024. "As we get closer to AGI, the definition itself will diffuse and fragment—not into five or 10 variations, but into thousands of 'not really AGI' solutions."
This sentiment captured something I had struggled to articulate in conversations with founders and fellow investors at the time. Artificial general intelligence (AGI) was defined as a form of AI with the cognitive ability to learn, reason, adapt, and perform any intellectual task on par with humans. But the market for AGI—potentially the biggest shift in human-computer interaction in our lifetime—remains surprisingly small. While AGI's impact on humanity will surely be profound, the real economic opportunity lies in models that are "good enough" at general reasoning and exceptional at specialized tasks.
We can already see this world taking shape through open-weight models, reinforcement-learning environments, and models post-trained on proprietary or sovereign data. Rather than converging on a single universal model, AI is fragmenting into many models adapted to specific contexts and use cases.
This specialization—not general intelligence—is where most of the economics of AI will be captured. The software applications that select the optimal model and point it at a specific use case will become the most important link in the AI value chain. As models become more capable and widely available, builders who can combine them with the right data and environments to create durable software applications will have a strategic edge.
