When considering the moves and pivots of existing players in the AI space, frontier model providers show little sign of slowing their pace of development. Anthropic released Claude Opus 5.5 on September 22, just 19 days after OpenAI launched GPT-6 Astra on September 3. At the same time, major AI players are moving beyond standalone models toward more autonomous agent products. SpaceXAI launched Grok Bot, a team of persistent AI agents that can work across apps and tools, while Meta introduced Muse, its personal AI agent powered by the Muse Spark model.
As a relatively small service provider compared with these giants, we do not have much time to establish our own position outside a handful of vertical or niche markets. As we have seen in search, which has long been dominated by Google, AI models and agents may eventually consolidate around just a few major players.
As in the case of the American Civil War, a side with overwhelming assets, resources, and technological advantages is unlikely to lose in a prolonged competition. Tactically, a smart David may occasionally defeat a Goliath, but strategically, this game is likely to end with the market divided among a handful of major players. Time is not on the side of smaller players; it favors the giants.
Above all, we do not plan to retreat into a narrow vertical or position ourselves merely as a niche player. Instead, we need to capture the local market before global players establish a firmer grip and consolidate their positions. In this regard, we must accelerate our efforts to address users’ pain points and unmet needs.
For now, many users are not yet loyal to a particular model or agent. Their choices are still heavily influenced by trends, novelty, and momentum. The recent Studio Ghibli-style AI image trend, for example, brought ChatGPT to a much broader mainstream audience by giving ordinary users an immediate and highly shareable reason to try generative AI. Nano Banana then picked up that momentum last year with another wave of consumer attention. Today, competition between ChatGPT and Claude is intensifying, particularly in coding and enterprise use cases.
In Korea, the government is driving a new initiative called “AI for Everyone,” under which three selected consortiums will provide free AI services using government-backed GPU infrastructure. This David-versus-Goliath battle in Korea could become a watershed moment in the AI agent market.
What these consortiums are focusing on today is not necessarily the same as what major foundation model providers are pursuing globally. The key question is how effectively they can differentiate themselves by combining the resources and capabilities of their consortium members with government support. If executed well, this could create a meaningful shift in a market where ChatGPT currently has more than 20 million users in Korea.
To that end, we need to make bold moves by connecting as many relevant MCPs and APIs as possible, enabling our agents to perform a broad range of tasks that global agent providers are not yet aggressively targeting. Local partnerships and technical capabilities such as computer use will also be essential to making this strategy successful.
Ultimately, whoever builds the strongest ecosystem for agent execution and task delegation will have an opportunity to turn the tide in this new era.