Competition among companies seeking to lead the early personal AI agent race, including Instinct and Meta Muse, is heating up. Big tech companies and promising startups are rushing into the personal AI agent market.

How the contest will play out is unpredictable. The current landscape of multiple firms competing could persist, but it could also be reshaped over time into a winner-takes-all structure.

Hunter Rice (헌터 라이스), who built the Sendit app with 250 million users and has had seven apps reach No. 1 in the App Store, drew attention after posting on social media platform X (Twitter) about conditions that will shape the race. He said he had tried more than 10 personal AI agents himself.

He said bluntly that none of the products he tried had the key elements needed to finish first.

The first challenge he cited was the wall users face at first use, or cold start. All the agents he tried had to start from a blank state. "For an agent to be useful, it has to know about me. But I only feel like spending time teaching it about me if it is useful," he said. Delivering an 'aha' moment within the first 30 seconds is important, and that is up to product designers.

He cited what a colleague experienced and pointed to context as the solution.

According to him, the wife of a colleague he used as an example knows that her husband goes to work on Mondays and what foods he likes. So she packs his lunch the night before. The husband does not need to ask or explain. "Everything agent companies are trying to build is contained in this small act: always being by your side, always thinking about us, knowing our preferences, and taking care of things without us having to say it," Rice said.

According to him, context falls into 2 categories.

One is behavioral context. It is information an agent accumulates by inferring preferences based on what users ask and tell it to do. The problem is that it is hard to know a person from a single question. He gave the example of asking, "What day of the week is the Super Bowl?" and said the person could be a lifelong American football fan, or someone who only wants to know when to go to a party.

"Something I asked once out of curiosity gets stored as my preference. Even after I forget I asked, that preference keeps intruding into recommendations. In the end, I get busy fixing the version of me that the AI drew wrong," he said.

The other is identity context. It is information about what kind of person someone is, not what they do. "Where you grew up, your job, the people close to you, and the environment you've lived in go in here. With this foundation, the Super Bowl question also finds its proper place. This is the difference between memory and understanding," Rice said.

Rice said the personal AI agent race will be decided by who captures identity context properly. "These agents show usefulness faster, understand what we say better, and earn trust that makes us willing to keep teaching them. One good conversation lets the next conversation start from a better place," he said. "Small thoughtfulness like a colleague's packed lunch will become the standard for a personal agent. If you can deliver that feeling, you can win," he said.