Two Waves of AI

Most people treat AI as one wave. It's actually two distinct waves, and understanding the difference changes how you plan.

Wave 1: LLMs as cheap intelligence via API. The smartest machine you can imagine, available over an API at nearly infinite scale. This is AI added to products. Sustaining innovation. Every SaaS company bolting "AI features" onto their existing product is riding Wave 1.

Wave 2: ~10x productivity in verifiable domains. Coding having 10x output is its own distinct wave. AI added to the development and product creation process changes the entire economic model. Cost assumptions that had been somewhat linear for decades no longer hold.

These waves are fundamentally different. Wave 1 makes existing products smarter. Wave 2 makes the people building products dramatically faster. One is a feature. The other changes the economics of entire markets.

The second wave prediction: usage is not spread evenly. Startups are moving faster than large companies, which was always true, but the gap is widening. Startups adopt new tools in days. Enterprises take quarters.

This matters for how you allocate your time and attention. If you're focused on Wave 1 (adding AI features to your product), you're in a sustaining innovation race. If you're focused on Wave 2 (using AI to build faster), you're changing the game entirely.

I spend time on both waves. At Vercel, we're using AI to build GTM systems that would have taken a full team eighteen months ago and were not even feasible. We shipped them in weeks. Adding AI to our software has made it more powerful and more useful.

When listening to predictions of what's going to happen because of AI, it's cleaner to reason about it as distinct forces. It makes it more clear how the situation will play out.

The tricky part is both waves are impacting us at the same time. We don't know if they cancel each other out or make the crash bigger.