What We’ve Learned from YC’s 199 Demo Day Companies. Honestly, I didn’t imagine that the “common sense” of startups I thought I already knew had changed this much.

AI is no longer just one category. 60% of the companies are AI-native, and another 26% are AI-enabled. In other words, only 14% are not using AI. The important question isn’t “Are you using AI?” but rather “How did you achieve what was impossible with traditional methods using AI?”

Another way to put it is that the “co-pilot era” is over. The current trend is the “full replacement” by AI agents. Automating jobs that pay $80k a year for a fraction of that salary. Not just support tools, but genuinely taking away jobs. This difference is surprisingly significant.

B2B was overwhelmingly dominant. 87% are B2B, with only 7% targeting consumers. However, some of the consumer-focused companies are actually disguised B2B companies. In the era where AI agents replace knowledge workers’ jobs, B2B is the most strategic battlefield.

The most striking commonality among fast-growing companies is that 40% of founders say, “We are solving a problem we personally experienced.” A risk advisor from McKinsey started a startup and achieved $700k in annual recurring revenue within three weeks. A former marketing firm executive reached $33,000 in monthly recurring revenue in four weeks. The combination of domain knowledge and entrepreneurial spirit is the strongest weapon.

The resurgence of hardware was also impressive. 18% are related to robots, drones, and wearables. Plans include warehouse robots, autonomous drones, and even lunar hotels. The data loops generated by physical products create a competitive advantage that software alone cannot replicate.

The gaps in consumer markets might actually signal opportunities. Fields like education, social media, mental health, and government tech have almost no AI companies. The areas with the least funding have historically yielded the highest returns.

Finally, the most important lesson: successful companies don’t start by “making a product and then selling it,” but by understanding “who they can reach and what they urgently need.” Business cards become distribution channels. YC’s network becomes customers. The original employer network becomes the initial market.

What future entrepreneurs should avoid are un-differentiated agent infrastructure, AI services without data advantage, and generic workflow wrappers. Foundation model providers will incorporate the same features within six months. Companies without defensibility won’t last long, no matter how quickly they monetize.

The most memorable pitch was for the lunar hotel. A $500 million investment intent letter and an invitation from the White House. The atmosphere in the room changed instantly. Half of me thought it was “crazy,” but the other half felt “it might actually be possible.” That difference might be the true quality of an entrepreneur.