Ultimately, success in AI relies on the ability to effectively stitch together existing models into products that users genuinely love, rather than reinventing the wheel.

I wouldn't expect AI models to one-shot entire products yet, especially if they're complicated. I prefer to implement features piece by piece and ensure each step works before proceeding.

A surprising number of AI-native products fail to expose the necessary application primitives to the models, limiting their effectiveness.

In the future, as AI tools proliferate, the obsession with creating excellent products will still require a human touch, as programming an AI to have that level of dedication is a significant challenge.

Startups should focus on building the best products on top of existing AI models, as competition will come from other startups rather than the AI giants themselves.

The challenge in AI is that while the underlying technology is evolving rapidly, it complicates product development because you have to predict where the tech will be in a year.

We're in an interesting time where the capabilities of AI models are far ahead of the products people have built. There's a huge opportunity for new innovations, especially as the cost of AI continues to decrease dramatically.

Design is about problem-solving and requires a deep understanding of user needs, which current AI models struggle to replicate.

Products that start as AI-light and transition to AI-heavy often struggle to integrate AI features seamlessly into their core functionality.