Sequoia recaps the hype cycle of generative AI²³⁹ in the past year and reflects on lessons learned. They argue we are entering "Act 2" where companies must solve real customer problems, not just showcase cool tech demos.

What's going on here?

After an initial explosion of hype and usage, generative AI companies are struggling with poor retention.

What does this mean?

The post argues the main challenge now is proving enduring value to users, not finding demand. Early generative AI products wowed people as novel demos, but they don't yet solve whole problems well enough for most people to use them daily. However, techniques like fine-tuning models on custom data are emerging to close the "expectations vs. reality gap."

Why should I care?

This is a pivotal time for generative AI. The tech has captured people's imagination, but it needs to mature from flashy demos into indispensable tools. As both an AI builder and user, I should care because this period will determine whether generative AI fizzles out as a fad or becomes as foundational as mobile and cloud computing. The playbook for creating lasting value with AI is still being written, so it's an exciting time to get involved.