At YC's Winter Demo Day, 199 startups took the stage, and participating in an event of this scale is truly shocking. Just being in the venue, I could see the outline of where the tech industry is headed over the next few years.

The most impressive thing was that AI is no longer just a "tool" but has become "infrastructure." 60% of the participating companies are AI-native, and an additional 26% are AI-enabled. In other words, only 14% are not using AI. But what's important is not just "we use AI," but "how we solve problems that couldn't be achieved with basic models using AI."

Here's where it gets interesting. The core theme for companies is not "co-pilot" but "AI agent." That is, a business model that doesn't support humans but fully replaces high-paying jobs. Beacon Health is replacing pre-approval staff, and Lancer is delegating hotel front desk operations to AI agents. This trend is real.

The overwhelming dominance of B2B was also noticeable. 87% are B2B companies. Only 14 companies target consumers, and of those, only 7 are officially classified as "consumer-facing." Why? Because AI agents are optimized for structured business workflows, making it difficult to respond to vague consumer demands.

It was also fascinating to see how founders found their markets. About 35% of the fastest monetizing companies had founders who personally experienced the problem. In other words, they are selling to former employers or colleagues. The founder of End Close previously worked at Modern Treasury, handling over $1 trillion in payment processing. That experience directly contributed to rapid customer acquisition.

The importance of data flywheels was repeatedly emphasized. LegalOS trained on 12,000 visa application data points and achieved a 100% approval rate. If there’s a system where every customer interaction leads to product improvement, the gap with competitors will widen over time. Conversely, relying only on generic AI wrappers will be overtaken by foundational model providers within weeks.

What surprised me was the revival of hardware. 18% of the startups handle physical products like robots, drones, and wearables. Remy AI and Servo7 are warehouse robots, and GrazeMate is a robotic cowboy managing cattle herds. Many founders come from SpaceX or Tesla, so their ambition for hardware tech is genuinely high.

I also learned that the fields that seem most popular are actually the most risky. There are zero companies in education, consumer social media, mental health, or government tech. This suggests these could be the next frontiers. Historically, legendary companies often emerge from fields with the least funding.

Some common traits among memorable presentations include shocking data, reframing problems, or starting with a background like "I experienced this problem myself." And they end with a concrete, verifiable, and quotable vision that makes investors want to reach for their checkbooks—like "The first AI Oscar will be born at Martini" or "You’ll be able to book a moon hotel by 2032."

Clear failure patterns include undifferentiated agent infrastructure, AI services without data advantage, simple workflow wrappers, or solo tech founders who don’t understand industry jargon and stall. Also, companies that only vaguely describe "AI for XX industry" without addressing specific problems are risky.

The five common traits of the fastest-growing companies are: selling results, founders building relationships with customers before product development, charging from day one (no free tiers), customers being in urgent situations rather than just curious, and MVPs that are unnaturally simple.

Reading this report, I feel there’s no better time to start a company. AI infrastructure is well-developed, foundational models are mature enough, and all that’s left is for founders with deep industry knowledge to identify what customers truly need and solve it simply. There are quite a few people sharing startup trends and industry analysis on Gate Content Plaza, so posts with this perspective might get good engagement. Tracking actual startup movements is useful for investment decisions and understanding overall market trends.