In hard tech, the iteration cycles are much longer than in software, often due to hardware constraints or regulatory issues, which can affect how quickly you can develop and test new ideas.

When developing new technology, expect that it won't work on the first try or even the tenth; it's crucial to iterate quickly and avoid anything that slows down your progress.

Despite increasing GPU resources, the intelligence improvements from the current generation of LLMs are not meeting expectations, with reports of failed training runs and diminishing returns emerging from major labs. Additionally, the lack of high-quality data to train new models has become a major bottleneck for AI development.

Every failure is a learning experience; as long as you learn from it and iterate quickly, it contributes to your growth and eventual success.

Training large models is challenging because new issues arise that weren't anticipated at the project's inception, requiring ongoing adjustments and problem-solving.

We're building hardware that will be of a very high quality, which takes time, as current computers were designed for a world without AI.

I believe most hardware founders should be 90% focused on operational efficiency and survival, and only 10% on design innovation initially.

Software development is never complete; it must evolve with hardware advancements to remain relevant and effective.

Craft and quality in product development stem from learning from past failures and ensuring that the product solves a real problem for customers.