History has a way of repeating itself in the platform age. Each new wave of technology creates a new set of giants, which can sometimes erase the startups that helped build the wave in the first place.
When OpenAI rolls out a new capability, it doesn’t just advance the state of AI; it redraws the map for thousands of companies. For founders, the question isn’t whether OpenAI will dominate the next decade. It’s how to build companies that thrive alongside it.
At the 2025 IA Summit, Jason Kwon, OpenAI’s chief strategy officer, who helps steer the company’s partnerships and global strategy, unpacked that tension directly. During our conversation, Jason shared how OpenAI sees its role in an increasingly crowded ecosystem and where founders can still find opportunity rather than competition in OpenAI’s orbit.
Build for the Ecosystem, Not Against It
Jason was direct about OpenAI’s focus: The company exists to build toward Artificial General Intelligence (AGI), meaning systems that can learn, reason, and act across tasks the way humans do. Anything on that “critical path,” he said, is fair game.
Everything else isn’t.
That line might sound narrow, but it leaves vast terrain open for founders, particularly in specialized domains (think manufacturing, healthcare, logistics ), which is not where OpenAI is going to build a full product suite. OpenAI’s energy, talent, and capital are concentrated on the underlying models and the reasoning stack that make them more capable. It will continue to release protocols and APIs that connect reasoning systems with the rest of the digital world. The startups that succeed will be the ones that extend those capabilities by offering trust layers, data integrations, or workflow tools that make the ecosystem stronger.
“Rather than try to over optimize on a particular set of capabilities you see on the models today,” Jason outlined, “bet that the general capability will continue to improve.”
Founders shouldn’t waste cycles trying to compete on model training or generic interfaces. Instead, they should build in the spaces where reasoning systems meet real problems – where the data, domain expertise, and user workflow create a durable moat. The edge won’t come from assuming stagnancy in OpenAI’s models; it will come from embedding those systems meaningfully into products that solve enduring, high-value problems.
Reasoning Is the New Platform
The conversation turned to how OpenAI views the full stack of AI — from compute and data, to foundation models, to the applications and devices built on top. Jason described an increasingly interconnected system where each layer feeds the next, and where new opportunities are emerging in the infrastructure, tooling, and post-training layers that support reasoning. As reasoning capabilities improve, he said, the boundary between these layers blurs.
At inference time, more compute now means better answers. That shift alone will reshape infrastructure: companies will need new tools for orchestration, evaluation, and cost optimization to manage “test-time reasoning” that consumes more power but delivers deeper insight.
The same applies to the application layer. As agents begin to act autonomously, searching, transacting, and collaborating with each other, software will need to adapt to a world where the user is no longer always human. OpenAI’s recent collaboration with Stripe to enable agent-driven commerce is an early glimpse of that future.
For startups, it’s a reminder that the next great platform opportunity isn’t just about building on top of AI models. It’s about enabling the systems that will use those models as building blocks for reasoning, planning, and execution.
The Takeaway for Founders
The future that Jason described isn’t a zero-sum game between startups and the platforms that power them. It’s a co-evolution, one that mirrors the early internet or mobile eras: the infrastructure giants build the substrate, and the most ambitious founders figure out how to make it indispensable to daily life.
OpenAI’s mission may be to reach general intelligence, but the opportunities it creates are anything but general. The next wave of breakout companies will be those that understand the underlying models will likely only get exponentially better and look for the opportunities to apply those models alongside human insight, domain context, and world-class product design.
As Jason put it, “The pace at which all these labs can actually build all these tools and technologies at the various layers is not necessarily going to keep up with the pace at which the ecosystem can build all these things.” For the founders willing to build in that gap, that’s not a threat. It’s an invitation.